Vehicle Loan Default Prediction

Financial institutions incur significant losses due to the default of vehicle loans. This has led to the tightening up of vehicle loan underwriting and increased vehicle loan rejection rates. The need for a better credit risk scoring model is also raised by these institutions. This warrants a study to estimate the determinants of vehicle loan default. A financial institution has hired you to accurately predict the probability of loanee/borrower defaulting on a vehicle loan in the first EMI (Equated Monthly Instalments) on the due date. Following Information regarding the loan and loanee are provided in the datasets:

• Loanee Information (Demographic data like age, income, Identity proof etc.)

• Loan Information (Disbursal details, amount, EMI, loan to value ratio etc.)

• Bureau data & history (Bureau score, number of active accounts, the status of other loans, credit history etc.)

Doing so will ensure that clients capable of repayment are not rejected and important determinants can be identified which can be further used for minimising the default rates.

Data Description

The Data Set contains:

• train.csv contains the training data with details on loan as described in the last section

• data_dictionary.csv contains a brief description on each variable provided in the training and test set.
  

• test.csv contains details of all customers and loans for which the participants are to submit probability of default.

• sample_submission.csv contains the submission format for the predictions against the test set. A single csv needs to be submitted as a solution.

Evaluation Metric

Submissions are evaluated on area under the ROC curve between the predicted probability and the observed target.

Functions for Visualizations

R Markdown does not support combining Rmd files. Per the R Markdown website, R Markdown requires a single Rmd file. It does not currently support the embedding of one Rmd file within another Rmd document.

Import Data

library('tidyverse')        # An opinionated collection of R Packages designed for Data Science. 
library('magrittr')         # A Forward-Pipe Operator for R
# install.packages("https://cran.r-project.org/src/contrib/data.table_1.11.8.tar.gz", repos = NULL, type = "source")     # Install package ‘data.table’ version 1.11.8
library('data.table')       # Fast Extension of `data.frame`

## 
## Attaching package: 'data.table'

## The following objects are masked from 'package:dplyr':
## 
##     between, first, last

## The following object is masked from 'package:purrr':
## 
##     transpose

LoadingData <- 'VLDP'

NameMySQLCode <- 'VLDP.sql'
# The 9 Best Features for Set 1
SelectedFeatures <- c('disbursed_amount', 'asset_cost', 'ltv', 'branch_id', 'supplier_id', 'manufacturer_id', 'Current_pincode_ID', 'Employment_Type', 'State_ID', 'Employee_code_ID', 'Aadhar_flag', 'PAN_flag', 'Driving_flag', 'Passport_flag', 'PERFORM_CNS_SCORE', 'PERFORM_CNS_SCORE_DESCRIPTION', 'PRI_NO_OF_ACCTS', 'PRI_ACTIVE_ACCTS', 'PRI_OVERDUE_ACCTS', 'PRI_SANCTIONED_AMOUNT', 'PRI_DISBURSED_AMOUNT', 'SEC_CURRENT_BALANCE', 'SEC_SANCTIONED_AMOUNT', 'PRIMARY_INSTAL_AMT', 'NEW_ACCTS_IN_LAST_SIX_MONTHS', 'DELINQUENT_ACCTS_IN_LAST_SIX_M', 'NO_OF_INQUIRIES', 'Age', 'YearsSinceDisbursment', 'AverageLoanTenure', 'TimeSinceFirstLoan') 

ModelFilename = 'VLDP_model.rds'

# Check It is Microsoft R Server working
isMicrosoftRServer <- 'RevoScaleR' %in% rownames(installed.packages())
if( isMicrosoftRServer ) { 
    writeLines(paste('Microsoft R Server (Machine Learning)', getNamespaceVersion('RevoScaleR') ))
} else writeLines(paste('Common', version$version.string ))

## Microsoft R Server (Machine Learning) 9.4.7

# To Import file with Data for Application Scoring
DT1 <- data.table::fread(unzip(zipfile = 'VehicleLoanDefaultPrediction.zip', files = 'train.csv'))
data.table::setnames(DT1, old = c('loan_default'), new = c('GB_flag'))

DT2 <- data.table::fread(unzip(zipfile = 'VehicleLoanDefaultPrediction.zip', files = 'test.csv'))

# Makes One Data.Table From A List Of Many filling missing columns, and match by col names
DT <- list(DT1, DT2) %>% 
  data.table::rbindlist(., use.names = TRUE, fill = TRUE)

# To Import file with Data Dictionary for Application Scoring
DD <- readxl::read_excel(unzip(zipfile = 'VehicleLoanDefaultPrediction.zip', files = 'Data Dictionary.xlsx')) %>%   dplyr::mutate(`Variable Name` = stringr::str_replace(string = `Variable Name`, pattern = 'loan_default',
                                                       replacement = 'GB_flag'))

## New names:
## * `` -> `..3`

# create a list of 70% of the rows in the original training dataset we can use for training
# Randomly Split the data into two sets: training (inTrain) and testing (inTest) 
seed <- 2019
set.seed(seed)

inTrain1 <- rep(FALSE, nrow(DT1))
inTrain1[sample(nrow(DT1), 9/10 * nrow(DT1))] <- TRUE

inTrain = c( inTrain1, rep( FALSE, times = dim(DT2)[1] ) )
inTest = c( !inTrain1, rep( FALSE, times = dim(DT2)[1] ) )
inProblem = c( rep( FALSE, times = dim(DT1)[1] ), rep( TRUE, times = dim(DT2)[1] ) )

# Adding Attributes by Variables into DT from Data Dictionary
attr(DT, "variable.labels") <- tibble(`Variable.Name` = colnames(DT)) %>%
  dplyr::left_join(DD, by = c('Variable.Name' = 'Variable Name')) %>%
    dplyr::select(Description) %>%
      pull

# gsub() is replacing '.' (dot)
names(DT) %<>% gsub('[.]', '_', .)
# DT$PRI_CURRENT_BALANCE <- DT$PRI_CURRENT_BALANCE + 6678296
# DT$PRI_CURRENT_BALANCE <- log(DT$PRI_CURRENT_BALANCE)

# Renaming All Variables with Name's Length > 31
oldnames = c('DELINQUENT_ACCTS_IN_LAST_SIX_MONTHS')

newnames = c('DELINQUENT_ACCTS_IN_LAST_SIX_M')

data.table::setnames(DT, old = oldnames, new = newnames)

# Get The Date of Birthday and convert into 'Age' at '2019-01-01'
Finish_Date <- lubridate::parse_date_time2('2019-01-01', orders = '%Y-%m-%d', tz = 'Asia/Dhaka')
DT[, Date_of_Birth := paste0(substr( DT$Date_of_Birth, 1, 6 ),
                             ifelse(substr( DT$Date_of_Birth, 7, 8 ) %>% as.integer > 40, '19', '20'),
                             substr( DT$Date_of_Birth, 7, 8 )) %>%
     as.Date(., '%d-%m-%Y') ]
DT[, Age := difftime(time1 = Finish_Date, time2 = Date_of_Birth, units = 'days') %>%
          as.integer(.) / 365.24 ]

# Get The Date of Disbursal and convert into 'YearsSinceDisbursment'
DT[, DisbursalDate := paste0(substr( DT$DisbursalDate, 1, 6 ),
                             ifelse(substr( DT$DisbursalDate, 7, 8 ) %>% as.integer > 40, '19', '20'),
                             substr( DT$DisbursalDate, 7, 8 )) %>%
    as.Date( ., '%d-%m-%Y') ]
DT[, YearsSinceDisbursment := difftime(time1 = Finish_Date, time2 = DisbursalDate, units = 'days') %>%
      as.integer(.) / 365.24 ]

# Get The Average loan tenure and convert into 'AverageLoanTenure'
DT[, AverageLoanTenure := 
     stringr::str_split_fixed(DT$AVERAGE_ACCT_AGE, ' ', 2) %>%
       data.frame(., stringsAsFactors = FALSE) %>%
         setNames(c('Years', 'Months')) %>%
           dplyr::mutate(Years = readr::parse_number(Years), Months = readr::parse_number(Months) / 12 ) %>%
             dplyr::mutate(Len = Years +  Months) %>%
               dplyr::select(Len) %>% pull ]

# Get The Time since first loan and convert into 'TimeSinceFirstLoan'
DT[, TimeSinceFirstLoan := 
     stringr::str_split_fixed(DT$CREDIT_HISTORY_LENGTH, ' ', 2) %>%
       data.frame(., stringsAsFactors = FALSE) %>%
         setNames(c('Years', 'Months')) %>%
           dplyr::mutate(Years = readr::parse_number(Years), Months = readr::parse_number(Months) / 12 ) %>%
             dplyr::mutate(Len = Years +  Months) %>%
               dplyr::select(Len) %>% pull ]

DT[, ':=' (
  YearsOnLoan = difftime(time1 = DisbursalDate, time2 = Date_of_Birth, units = 'days') %>%
          as.integer(.) / 365.24,
  DisAsDiff = asset_cost - disbursed_amount,
  DisAsShare = asset_cost / disbursed_amount,
  # DiffLTV = (asset_cost - disbursed_amount - ltv),
  
  
  Qrt = lubridate::quarter(DisbursalDate) %>% as.factor(.), # weekdays
  Day = format(DisbursalDate, '%d') %>% as.integer,
  
  OutstandingNow = disbursed_amount + PRI_CURRENT_BALANCE,
  DisbursedTotal = PRI_DISBURSED_AMOUNT + disbursed_amount,
  ShareOverdue = DELINQUENT_ACCTS_IN_LAST_SIX_M - NEW_ACCTS_IN_LAST_SIX_MONTHS,
  # OutstandingNow2Dsbrsd = OutstandingNow / DisbursedTotal
  
  SEC_OverdueShareSec = SEC_OVERDUE_ACCTS / SEC_NO_OF_ACCTS,
  PRI_OverdueShare = PRI_OVERDUE_ACCTS / PRI_NO_OF_ACCTS
) ]

library('Hmisc')            # Harrell Miscellaneous

## Loading required package: survival

## 
## Attaching package: 'survival'

## The following object is masked from 'package:rpart':
## 
##     solder

## Loading required package: Formula

## 
## Attaching package: 'Hmisc'

## The following objects are masked from 'package:dplyr':
## 
##     src, summarize

## The following objects are masked from 'package:base':
## 
##     format.pval, units

DT <- Hmisc::upData(DT, labels = c(Age = 'Age in Years', 
                                   YearsSinceDisbursment = 'Years Since Disbursment',
                                   AverageLoanTenure = 'Average loan tenure in Years (AVERAGE_ACCT_AGE)',
                                   TimeSinceFirstLoan = 'Years since First Loan (CREDIT_HISTORY_LENGTH)',
                                   YearsOnLoan = 'Years On Loan',
                                   DisAsDiff = 'Difference of asset_cost from disbursed_amount',
                                   DisAsShare = 'Ratio of asset_cost to disbursed_amount',
                                   # DiffLTV = 'Difference of asset_cost from disbursed_amount and ltv',
                                   Qrt = 'Quarter of DisbursalDate',
                                   Day = 'Months`s Day of DisbursalDate',
                                   OutstandingNow = 'Difference of disbursed_amount from PRI_CURRENT_BALANCE',
                                   DisbursedTotal = 'Summ of PRI_DISBURSED_AMOUNT and disbursed_amount',
                                   ShareOverdue = 'Difference of DELINQUENT_ACCTS_IN_LAST_SIX_M from NEW_ACCTS_IN_LAST_SIX_MONTHS',
                                   SEC_OverdueShareSec = 'Ratio of SEC_OVERDUE_ACCTS to SEC_NO_OF_ACCTS',
                                   PRI_OverdueShare = 'Ratio of PRI_OVERDUE_ACCTS / PRI_NO_OF_ACCTS'))

## Input object size:    96804992 bytes;     55 variables    345546 observations
## New object size: 96812608 bytes; 55 variables    345546 observations

# Hmisc::contents(DT)

# Convert *some* column classes in data.table
factor_cols  <- c('branch_id', 'supplier_id', 'manufacturer_id', 'Current_pincode_ID', 'Employment_Type', 'State_ID', 'Employee_code_ID', 'MobileNo_Avl_Flag', 'Aadhar_flag', 'PAN_flag', 'VoterID_flag', 'Driving_flag', 'Passport_flag', 'PERFORM_CNS_SCORE_DESCRIPTION')

DT[, (factor_cols) := lapply(.SD, factor), .SDcols = factor_cols]

delcols <- c('UniqueID', 'Date_of_Birth', 'DisbursalDate', 'PRI_CURRENT_BALANCE', 'AVERAGE_ACCT_AGE', 'CREDIT_HISTORY_LENGTH') # , grep('_flag', names(DT), value = TRUE), grep('SEC_', names(DT), value = TRUE)

# Remove outcome variable 'GB_flag' from 'delcols' vector to due '_flag'
delcols <- delcols[! delcols %in% c('GB_flag')]

if ( isMicrosoftRServer ) {
# for Microsoft Machine Learning Server 9.3
# Get a subset rows and columns from the data frame
  DF1 <- RevoScaleR::rxDataStep(inData = DT, maxRowsByCols = 2e9,
      varsToDrop = delcols 
      # varsToKeep = c('x', 'w', 'z'),
      # rowSelection = z > 0
      )
  out <- rxGetInfo(DF1, getVarInfo = TRUE)
} else {
  # for CRAN R version
  delColumns <- delcols
  DF1 <- copy(DT) %>% 
    .[, (delColumns) := NULL] %>% 
      data.table::setDF(.)
}

## Rows Read: 345546, Total Rows Processed: 345546, Total Chunk Time: 0.746 seconds

attr(DF1, 'variable.labels') <- attr(DF1, 'variable.labels')[names(DF1)]

I use package embed that contains extra steps for the recipes package for embedding predictors into one or more numeric columns. All of the preprocessing methods are supervised.

# Encoding of Multi-Level Categorical Variables (Factors) into Numeric Predictors
for (var in c('branch_id', 'supplier_id', 'manufacturer_id', 'Current_pincode_ID', 'Employment_Type', 'State_ID', 'Employee_code_ID', 'PERFORM_CNS_SCORE_DESCRIPTION')) {
  print(var)
  
  DF1 <- EncodingMultiLevelFactors(X = DF1, Y = var, Z = inTrain)
}

## [1] "branch_id"
## start par. =  0.1123205 fn =  221335.9 
## At return
## eval:  19 fn:      221335.56 par: 0.106959
## [1] "supplier_id"
## start par. =  0.2075279 fn =  220404.6 
## At return
## eval:  19 fn:      220349.07 par: 0.172799
## [1] "manufacturer_id"
## start par. =  0.04441674 fn =  223266.8 
## At return
## eval:  18 fn:      223266.33 par: 0.0533252
## [1] "Current_pincode_ID"
## start par. =  0.245132 fn =  221670.7 
## At return
## eval:  16 fn:      221350.34 par: 0.166798
## [1] "Employment_Type"
## start par. =  0.02972829 fn =  216219.5 
## At return
## eval:  28 fn:      216219.22 par: 0.0422908
## [1] "State_ID"
## start par. =  0.08355545 fn =  222283.1 
## At return
## eval:  18 fn:            222283. par: 0.0889640
## [1] "Employee_code_ID"
## start par. =  0.2213827 fn =  220258.2 
## At return
## eval:  19 fn:      220127.69 par: 0.175705
## [1] "PERFORM_CNS_SCORE_DESCRIPTION"
## start par. =  0.098095 fn =  221735 
## At return
## eval:  17 fn:      221733.42 par: 0.119980

# DF1 <- DF1[, c('GB_flag', SelectedFeatures)]

DF0 <- DF1[inTrain, ]   # Testing Set after Discretization (Binning) of Numeric Variables

Max_Vars <- 32
Max_Levels <- 48

remove(DT1, DT2, DD, inTrain1, delcols, factor_cols, oldnames, newnames)

Inspection Data Frame using Automate Data Exploration and Inspection packages: DataExplorer, summarytools.

# Inspection  Data Frame

library('DataExplorer')     # Automate Data Exploration and Treatment 
library('summarytools')     # Tools to Quickly and Neatly Summarize Data

## 
## Attaching package: 'summarytools'

## The following objects are masked from 'package:Hmisc':
## 
##     label, label<-

## The following object is masked from 'package:tibble':
## 
##     view

dfName <- 'DT'
dfCols <- colnames(DT) %>% length()

data.frame(Variables = sapply(DT, class) %>%  unlist) %>%
  dplyr::filter(Variables != 'labelled') %>% 
    ggplot2::ggplot(aes(x = Variables, fill = Variables)) +
      ggplot2::geom_bar( stat = 'count') +
      ggplot2::geom_text(stat='count', aes(label = ..count..), vjust = 2, size = 5, color = 'white') +
      ggplot2::theme(legend.position = 'none') +
      ggplot2::labs(title = paste(dfName, 'Column Types'), subtitle = sprintf('Data.frame has %g columns.', dfCols),
                    x = NULL, y = 'Number of columns') 

# Summarise Missing values in dataframe
DataExplorer::plot_intro(DT)

DataExplorer::plot_missing(DT, geom_label_args = list("size" = 2, "label.padding" = unit(0.1, "lines")))

# DataExplorer::profile_missing(DT)

# Data frame Summary by Variable
print(summarytools::dfSummary(DT, graph.magnif = 0.75), method = 'render')

Data Frame Summary

DT

N: 345546

No	Variable	Label	Stats / Values	Freqs (% of Valid)	Graph	Valid	Missing
1	UniqueID [integer]		mean (sd) : 593106.04 (101481.56) min < med < max : 417428 < 592918.5 < 769909 IQR (CV) : 175115.5 (0.17)	345546 distinct values		345546 (100%)	0 (0%)
2	disbursed_amount [integer]		mean (sd) : 54916.38 (13045.96) min < med < max : 11613 < 54303 < 990572 IQR (CV) : 13302 (0.24)	29271 distinct values		345546 (100%)	0 (0%)
3	asset_cost [integer]		mean (sd) : 76294.84 (18738.64) min < med < max : 37000 < 71541 < 1628992 IQR (CV) : 13323 (0.25)	53158 distinct values		345546 (100%)	0 (0%)
4	ltv [numeric]		mean (sd) : 74.93 (11.32) min < med < max : 10.03 < 77.14 < 95 IQR (CV) : 14.41 (0.15)	6819 distinct values		345546 (100%)	0 (0%)
5	branch_id [factor]		1. 1 2. 2 3. 3 4. 5 5. 7 6. 8 7. 9 8. 10 9. 11 10. 13 [ 72 others ]	8337 (2.4%) 20527 (5.9%) 14881 (4.3%) 10276 (3.0%) 4323 (1.3%) 4472 (1.3%) 3891 (1.1%) 5685 (1.6%) 5873 (1.7%) 4170 (1.2%) 263111 (76.1%)		345546 (100%)	0 (0%)
6	supplier_id [factor]		1. 10524 2. 12311 3. 12312 4. 12374 5. 12441 6. 12456 7. 12500 8. 12534 9. 12539 10. 12797 [ 3079 others ]	7 (0.0%) 6 (0.0%) 57 (0.0%) 146 (0.0%) 62 (0.0%) 101 (0.0%) 79 (0.0%) 73 (0.0%) 11 (0.0%) 87 (0.0%) 344917 (99.8%)		345546 (100%)	0 (0%)
7	manufacturer_id [factor]		1. 45 2. 48 3. 49 4. 51 5. 67 6. 86 7. 120 8. 145 9. 152 10. 153 [ 2 others ]	87053 (25.2%) 22964 (6.6%) 14812 (4.3%) 40927 (11.8%) 3364 (1.0%) 161203 (46.7%) 14049 (4.1%) 1138 (0.3%) 9 (0.0%) 25 (0.0%) 2 (0.0%)		345546 (100%)	0 (0%)
8	Current_pincode_ID [factor]		1. 1 2. 2 3. 3 4. 4 5. 5 6. 6 7. 7 8. 8 9. 9 10. 10 [ 7086 others ]	44 (0.0%) 118 (0.0%) 87 (0.0%) 153 (0.0%) 331 (0.1%) 162 (0.0%) 152 (0.0%) 74 (0.0%) 56 (0.0%) 8 (0.0%) 344361 (99.7%)		345546 (100%)	0 (0%)
9	Date_of_Birth [Date]		min : 1949-09-15 med : 1986-01-01 max : 2000-11-29 range : 51y 2m 14d	15888 distinct val.		345546 (100%)	0 (0%)
10	Employment_Type [factor]		1. 2. Salaried 3. Self employed	11104 (3.2%) 147013 (42.5%) 187429 (54.2%)		345546 (100%)	0 (0%)
11	DisbursalDate [Date]		min : 2018-08-01 med : 2018-10-20 max : 2018-11-30 range : 3m 29d	111 distinct val.		345546 (100%)	0 (0%)
12	State_ID [factor]		1. 1 2. 2 3. 3 4. 4 5. 5 6. 6 7. 7 8. 8 9. 9 10. 10 [ 12 others ]	14351 (4.2%) 7258 (2.1%) 47868 (13.9%) 70438 (20.4%) 14304 (4.1%) 48903 (14.2%) 10628 (3.1%) 20047 (5.8%) 21459 (6.2%) 5564 (1.6%) 84726 (24.5%)		345546 (100%)	0 (0%)
13	Employee_code_ID [factor]		1. 1 2. 3 3. 4 4. 5 5. 7 6. 9 7. 10 8. 11 9. 12 10. 15 [ 3388 others ]	106 (0.0%) 192 (0.1%) 96 (0.0%) 133 (0.0%) 221 (0.1%) 77 (0.0%) 44 (0.0%) 111 (0.0%) 162 (0.0%) 146 (0.0%) 344258 (99.6%)		345546 (100%)	0 (0%)
14	MobileNo_Avl_Flag [factor]		1. 1	345546 (100.0%)		345546 (100%)	0 (0%)
15	Aadhar_flag [factor]		1. 0 2. 1	51883 (15.0%) 293663 (85.0%)		345546 (100%)	0 (0%)
16	PAN_flag [factor]		1. 0 2. 1	306392 (88.7%) 39154 (11.3%)		345546 (100%)	0 (0%)
17	VoterID_flag [factor]		1. 0 2. 1	298155 (86.3%) 47391 (13.7%)		345546 (100%)	0 (0%)
18	Driving_flag [factor]		1. 0 2. 1	338249 (97.9%) 7297 (2.1%)		345546 (100%)	0 (0%)
19	Passport_flag [factor]		1. 0 2. 1	344835 (99.8%) 711 (0.2%)		345546 (100%)	0 (0%)
20	PERFORM_CNS_SCORE [integer]		mean (sd) : 289.03 (338.84) min < med < max : 0 < 0 < 890 IQR (CV) : 679 (1.17)	574 distinct values		345546 (100%)	0 (0%)
21	PERFORM_CNS_SCORE_DESCRIPTION [factor]		1. A-Very Low Risk 2. B-Very Low Risk 3. C-Very Low Risk 4. D-Very Low Risk 5. E-Low Risk 6. F-Low Risk 7. G-Low Risk 8. H-Medium Risk 9. I-Medium Risk 10. J-High Risk [ 10 others ]	21683 (6.3%) 13696 (4.0%) 23870 (6.9%) 16472 (4.8%) 8393 (2.4%) 12176 (3.5%) 5795 (1.7%) 10142 (2.9%) 8260 (2.4%) 5526 (1.6%) 219533 (63.5%)		345546 (100%)	0 (0%)
22	PRI_NO_OF_ACCTS [integer]		mean (sd) : 2.37 (5.01) min < med < max : 0 < 0 < 453 IQR (CV) : 3 (2.11)	114 distinct values		345546 (100%)	0 (0%)
23	PRI_ACTIVE_ACCTS [integer]		mean (sd) : 1 (1.88) min < med < max : 0 < 0 < 144 IQR (CV) : 1 (1.87)	42 distinct values		345546 (100%)	0 (0%)
24	PRI_OVERDUE_ACCTS [integer]		mean (sd) : 0.16 (0.54) min < med < max : 0 < 0 < 25 IQR (CV) : 0 (3.5)	23 distinct values		345546 (100%)	0 (0%)
25	PRI_CURRENT_BALANCE [integer]		mean (sd) : 160270.2 (925345.66) min < med < max : -6678296 < 0 < 96524920 IQR (CV) : 31364.5 (5.77)	97465 distinct values		345546 (100%)	0 (0%)
26	PRI_SANCTIONED_AMOUNT [integer]		mean (sd) : 209650.86 (2043865.78) min < med < max : -481500 < 0 < 1e+09 IQR (CV) : 59416.75 (9.75)	60681 distinct values		345546 (100%)	0 (0%)
27	PRI_DISBURSED_AMOUNT [integer]		mean (sd) : 209560.79 (2047482.84) min < med < max : 0 < 0 < 1e+09 IQR (CV) : 57645.75 (9.77)	65673 distinct values		345546 (100%)	0 (0%)
28	SEC_NO_OF_ACCTS [integer]		mean (sd) : 0.05 (0.56) min < med < max : 0 < 0 < 57 IQR (CV) : 0 (11.82)	40 distinct values		345546 (100%)	0 (0%)
29	SEC_ACTIVE_ACCTS [integer]		mean (sd) : 0.02 (0.28) min < med < max : 0 < 0 < 36 IQR (CV) : 0 (12.47)	23 distinct values		345546 (100%)	0 (0%)
30	SEC_OVERDUE_ACCTS [integer]		mean (sd) : 0.01 (0.1) min < med < max : 0 < 0 < 8 IQR (CV) : 0 (16.97)	0 : 343928 (99.5%) 1 : 1358 (0.4%) 2 : 166 (0.0%) 3 : 54 (0.0%) 4 : 22 (0.0%) 5 : 8 (0.0%) 6 : 6 (0.0%) 7 : 2 (0.0%) 8 : 2 (0.0%)		345546 (100%)	0 (0%)
31	SEC_CURRENT_BALANCE [integer]		mean (sd) : 4565.3 (161202.6) min < med < max : -574647 < 0 < 36032852 IQR (CV) : 0 (35.31)	3947 distinct values		345546 (100%)	0 (0%)
32	SEC_SANCTIONED_AMOUNT [integer]		mean (sd) : 6133.3 (189342.66) min < med < max : 0 < 0 < 57945000 IQR (CV) : 0 (30.87)	2631 distinct values		345546 (100%)	0 (0%)
33	SEC_DISBURSED_AMOUNT [integer]		mean (sd) : 6038.72 (188911.37) min < med < max : 0 < 0 < 57945000 IQR (CV) : 0 (31.28)	3031 distinct values		345546 (100%)	0 (0%)
34	PRIMARY_INSTAL_AMT [integer]		mean (sd) : 12497.73 (199754.5) min < med < max : 0 < 0 < 85262329 IQR (CV) : 1946 (15.98)	34330 distinct values		345546 (100%)	0 (0%)
35	SEC_INSTAL_AMT [integer]		mean (sd) : 272.74 (16261.26) min < med < max : 0 < 0 < 5390000 IQR (CV) : 0 (59.62)	2295 distinct values		345546 (100%)	0 (0%)
36	NEW_ACCTS_IN_LAST_SIX_MONTHS [integer]		mean (sd) : 0.36 (0.92) min < med < max : 0 < 0 < 35 IQR (CV) : 0 (2.56)	26 distinct values		345546 (100%)	0 (0%)
37	DELINQUENT_ACCTS_IN_LAST_SIX_M [integer]		mean (sd) : 0.1 (0.38) min < med < max : 0 < 0 < 20 IQR (CV) : 0 (4.01)	16 distinct values		345546 (100%)	0 (0%)
38	AVERAGE_ACCT_AGE [character]		1. 0yrs 0mon 2. 0yrs 6mon 3. 0yrs 7mon 4. 0yrs 11mon 5. 0yrs 10mon 6. 1yrs 0mon 7. 0yrs 9mon 8. 0yrs 8mon 9. 1yrs 1mon 10. 0yrs 5mon [ 190 others ]	177481 (51.4%) 9325 (2.7%) 8167 (2.4%) 7665 (2.2%) 7587 (2.2%) 7447 (2.2%) 7353 (2.1%) 7224 (2.1%) 6680 (1.9%) 6458 (1.9%) 100159 (29.0%)		345546 (100%)	0 (0%)
39	CREDIT_HISTORY_LENGTH [character]		1. 0yrs 0mon 2. 0yrs 6mon 3. 2yrs 1mon 4. 0yrs 7mon 5. 2yrs 0mon 6. 1yrs 0mon 7. 1yrs 1mon 8. 0yrs 11mon 9. 0yrs 8mon 10. 0yrs 9mon [ 297 others ]	177178 (51.3%) 7456 (2.2%) 6932 (2.0%) 6243 (1.8%) 5762 (1.7%) 5153 (1.5%) 4542 (1.3%) 3925 (1.1%) 3753 (1.1%) 3572 (1.0%) 121030 (35.0%)		345546 (100%)	0 (0%)
40	NO_OF_INQUIRIES [integer]		mean (sd) : 0.21 (0.72) min < med < max : 0 < 0 < 36 IQR (CV) : 0 (3.37)	26 distinct values		345546 (100%)	0 (0%)
41	GB_flag [integer]		mean (sd) : 0.22 (0.41) min < med < max : 0 < 0 < 1 IQR (CV) : 0 (1.9)	0 : 182543 (78.3%) 1 : 50611 (21.7%)		233154 (67.47%)	112392 (32.53%)
42	Age [labelled, numeric]	Age in Years	mean (sd) : 34.81 (9.86) min < med < max : 18.09 < 33 < 69.29 IQR (CV) : 15.28 (0.28)	15888 distinct values		345546 (100%)	0 (0%)
43	YearsSinceDisbursment [labelled, numeric]	Years Since Disbursment	mean (sd) : 0.22 (0.1) min < med < max : 0.08 < 0.2 < 0.42 IQR (CV) : 0.16 (0.43)	111 distinct values		345546 (100%)	0 (0%)
44	AverageLoanTenure [labelled, numeric]	Average loan tenure in Years (AVERAGE_ACCT_AGE)	mean (sd) : 0.74 (1.26) min < med < max : 0 < 0 < 30.75 IQR (CV) : 1.08 (1.69)	200 distinct values		345546 (100%)	0 (0%)
45	TimeSinceFirstLoan [labelled, numeric]	Years since First Loan (CREDIT_HISTORY_LENGTH)	mean (sd) : 1.33 (2.35) min < med < max : 0 < 0 < 39 IQR (CV) : 1.92 (1.76)	307 distinct values		345546 (100%)	0 (0%)
46	YearsOnLoan [labelled, numeric]	Years On Loan	mean (sd) : 34.59 (9.86) min < med < max : 18 < 32.81 < 69.13 IQR (CV) : 15.23 (0.28)	16252 distinct values		345546 (100%)	0 (0%)
47	DisAsDiff [labelled, integer]	Difference of asset_cost from disbursed_amount	mean (sd) : 21378.46 (12308.93) min < med < max : 3917 < 17901 < 638420 IQR (CV) : 13187 (0.58)	48676 distinct values		345546 (100%)	0 (0%)
48	DisAsShare [labelled, numeric]	Ratio of asset_cost to disbursed_amount	mean (sd) : 1.42 (0.31) min < med < max : 1.07 < 1.34 < 10.57 IQR (CV) : 0.26 (0.22)	299901 distinct values		345546 (100%)	0 (0%)
49	Qrt [labelled, factor]	Quarter of DisbursalDate	1. 3 2. 4	134790 (39.0%) 210756 (61.0%)		345546 (100%)	0 (0%)
50	Day [labelled, integer]	Months`s Day of DisbursalDate	mean (sd) : 19.3 (7.86) min < med < max : 1 < 20 < 31 IQR (CV) : 13 (0.41)	31 distinct values		345546 (100%)	0 (0%)
51	OutstandingNow [labelled, integer]	Difference of disbursed_amount from PRI_CURRENT_BALANCE	mean (sd) : 215186.57 (925615.45) min < med < max : -6608979 < 60379.5 < 96583433 IQR (CV) : 40649.75 (4.3)	118719 distinct values		345546 (100%)	0 (0%)
52	DisbursedTotal [labelled, integer]	Summ of PRI_DISBURSED_AMOUNT and disbursed_amount	mean (sd) : 264477.16 (2047611.56) min < med < max : 11613 < 62639.5 < 1000047773 IQR (CV) : 63776.5 (7.74)	121005 distinct values		345546 (100%)	0 (0%)
53	ShareOverdue [labelled, integer]	Difference of DELINQUENT_ACCTS_IN_LAST_SIX_M from NEW_ACCTS_IN_LAST_SIX_MONTHS	mean (sd) : -0.26 (0.93) min < med < max : -30 < 0 < 17 IQR (CV) : 0 (-3.52)	38 distinct values		345546 (100%)	0 (0%)
54	SEC_OverdueShareSec [labelled, numeric]	Ratio of SEC_OVERDUE_ACCTS to SEC_NO_OF_ACCTS	mean (sd) : 0.16 (0.33) min < med < max : 0 < 0 < 1 IQR (CV) : 0 (2.11)	66 distinct values		7108 (2.06%)	338438 (97.94%)
55	PRI_OverdueShare [labelled, numeric]	Ratio of PRI_OVERDUE_ACCTS / PRI_NO_OF_ACCTS	mean (sd) : 0.09 (0.23) min < med < max : 0 < 0 < 1 IQR (CV) : 0 (2.49)	324 distinct values		170703 (49.4%)	174843 (50.6%)

Generated by summarytools 0.8.8 (R version 3.5.2)
2019-08-31

remove( dfName, dfCols )

# Binning and Fine Classing Factor and Numeric (Scale) Variables
# https://support.sas.com/documentation/cdl/en/emcsgs/66008/PDF/default/emcsgs.pdf - SAS Enterprise Miner 12.1

# install.packages("https://cran.microsoft.com/src/contrib/smbinning_0.9.tar.gz", repos = NULL, type = "source")
library('smbinning')         # Scoring Modeling and Optimal Binning for GLM Model from Herman Jopia (Chile)

## Loading required package: sqldf

## Loading required package: gsubfn

## Loading required package: proto

## Loading required package: RSQLite

## Loading required package: partykit

## Loading required package: grid

## Loading required package: libcoin

## Loading required package: mvtnorm

# install.packages("https://cran.r-project.org/src/contrib/woeBinning_0.1.6.tar.gz", repos = NULL, type = "source")
library('woeBinning')        # Supervised Weight of Evidence Binning of Numeric Variables and Factors
library('openxlsx')          # Read, Write and Edit Miscosoft XLSX Files 

isobin <- function(data, y, x) { # Second Variant - Finer Monotonic Binning Based on Isotonic Regression
# WenSui Liu, is leading a team of quantitative analysts developing operational risk models for American Bank.
# https://statcompute.wordpress.com/2017/06/15/finer-monotonic-binning-based-on-isotonic-regression/
  d1 <- data[c(y, x)]
  d2 <- d1[!is.na(d1[x]), ]
  c <- cor(d2[, 2], d2[, 1], method = 'spearman', use = 'complete.obs')
  reg <- isoreg(d2[, 2], c / abs(c) * d2[, 1])
  k <- knots(as.stepfun(reg))
  sm1 <- smbinning::smbinning.custom(d1, y, x, k)
  c1 <- subset(sm1$ivtable, subset = CntGood * CntBad > 0, select = Cutpoint)
  c2 <- suppressWarnings(as.numeric(unlist(strsplit(c1$Cutpoint, ' '))))
  c3 <- c2[!is.na(c2)]
  return(smbinning::smbinning.custom(d1, y, x, c3[-length(c3)]))
}

tree_bin<- function(data, y, x) {
# Thilo Eichenberg for `woeBinning` package
  binning <- woeBinning::woe.tree.binning(df = data, target.var =  y, pred.var =  x,
                           min.perc.total = 0.05, min.perc.class = 0, stop.limit = 0.01,
                           abbrev.fact.levels = 200, event.class = 1)
  
  if (class(binning) == 'list') {
    z <- c()
    
    if ( class(data[, x]) ==  'factor') {
      for (variable in binning[[2]]$Group.2 %>% levels()) {
        df <- binning[[2]]
        fac_vec <- df[binning[[2]]$Group.2 == variable, 'Group.1'] %>%
          as.character 
        chr_vec <- paste0('\'', paste0(fac_vec,  sep = "\'", collapse = ', \''))
        z <- c(z, chr_vec)
      }  
    } else { 
      df <- binning[[2]]
      z <- df$cutpoints.final %>%
        .[c(-1, -nrow(df))] %>%
          as.vector()
      } # End of 'Numeric' class
    } else { # ERROR !!!
      z <- c()
  }
  
  return( z )
}

# Exploratory Data Analysis (EDA)
if (Max_Vars >= ncol(DF0) )
  smbinning.eda(DF0)$eda

# Convert double after round into integer for smbinning()
DF0 %<>% # dplyr::mutate_if(is.double, round) %>%
  # dplyr::mutate_if(is.double, as.integer) %>% 
    dplyr::mutate( GB_flag = ifelse(GB_flag == 1, 0, 1) )   # Flag for Optimal Binning
# Convert integer features with unique < 5 into factor for smbinning()
DF0 %<>% dplyr::mutate_at( dplyr::select_if(., ~ is.integer(.) & unique(.) %>% length(.) < 5 ) %>%
  dplyr::select( -dplyr::one_of('GB_flag') ) %>% colnames,
    as.factor )

# Create MS Excel File for Output
openxlsx::addWorksheet(wb <- openxlsx::createWorkbook(), sheetName = 'IV Table', 
                       gridLines = FALSE, tabColour = 'olivedrab')
openxlsx::addWorksheet(wb, sheetName = 'Scorecard', gridLines = FALSE, tabColour = 'brown')

NamesOfVariables <- DF0 %>%
  dplyr::select( -dplyr::one_of('GB_flag', dplyr::select_if(., ~ is.factor(.) & nlevels(.) == 1) %>%
                                  colnames, # Levels > 1
                         # At least 5 different values for Numeric variables
                dplyr::select_if(., ~ is.numeric(.) & unique(.) %>% length() < 5) %>% colnames)) %>%
    colnames

binning.df <- cbind(Variable = NamesOfVariables,
                    `IV-Finish` = rep(NA_real_, times = length(NamesOfVariables)),
                    data.frame(matrix(data = rep(NA_real_, length(NamesOfVariables) * 8),
                                      nrow = length(NamesOfVariables), ncol = 8))) %>%
                      setNames(c('Variable', 'IV', 'IV-RPart', 'N-RPart', 'IV-Decile', 'N-Decile',
                                 'IV-Iso', 'N-Iso', 'IV-Tree', 'N-Tree'))
binning.df$Method <- rep(x = '', times = length(NamesOfVariables))

TotalBinning.sql <- ''

for (i in 1:length(NamesOfVariables)) {
  val <- NamesOfVariables[i]
  writeLines(paste(i, '-', val))

  if (DF0[, val] %>% is.factor) {  #  Generate a binning table for all the categories of a given factor variable (A factor variable with at least 2 different values. Labels with commas are not allowed)
    
    result.smb <- switch(val,
      `education`    = smbinning.factor.custom(DF0, x = val, y = 'GB_flag',
                          c("'Высшее'",                                    # 'Высшее'
                            "'Начальное','Средне-специальное'",            # 'Начальное','Средне-специальное'
                            "'Среднее'")),                                 # 'Среднее'
      if (levels(DF0[, val]) %>% length() > Max_Levels) { # Multi levels `factor` variables
        chr_vec <- tree_bin(DF0, 'GB_flag', val)   # Combine levels of factor variable by Badrate into some bins
        smbinning.factor.custom(DF0, x = val, y = 'GB_flag', chr_vec)
      }
      else {
        smbinning.factor(DF0, x = val, y = 'GB_flag', maxcat = levels(DF0[, val]) %>% length() + 1)  
      }
    )

    if (class(result.smb) == 'list') {
      binning.df[i, 'IV']      <- result.smb$iv
      binning.df[i, 'IV-RPart'] <- result.smb$iv
      binning.df[i, 'N-RPart']  <- result.smb$ivtable %>% nrow - 1 # with Missing Vales
      binning.df[i, 'Method']  <- ifelse(length(result.smb$groups) == 0, 'IV-By Levels', 'IV-By Groups')[1]
      
        # IV Table Supplement
      result.smb$ivtable <- result.smb$ivtable %>%
      dplyr::mutate(G_Dis = CntGood / table(DF0$GB_flag)[2],
           B_Dis = CntBad / table(DF0$GB_flag)[1],
           `G/B Index` = ifelse(G_Dis > B_Dis, G_Dis / B_Dis, B_Dis / G_Dis),
           `0=Good, 1=Bad` = ifelse(G_Dis > B_Dis, 0, 1),
           Bin = c(1:(nrow(result.smb$ivtable) - 1), NA),
           # Min = (c(NA, result.smb$cuts, NA, NA)),
           # Max = (c(result.smb$cuts, NA, NA, NA))
           Min = rep(NA, times = result.smb$ivtable %>% nrow),
           Max = rep(NA, times = result.smb$ivtable %>% nrow)
      )
    }

  } else 
    {                                                                  # Numeric Class
    # Optimal Binning for Scoring Modeling from package `smbinning`
    # This process, also known as supervised discretization, utilizes Recursive Partitioning to categorize the numeric characteristic. The especific algorithm is Conditional Inference Trees which initially excludes missing values (NA) to compute the cutpoints, adding them back later in the process for the calculation of the Information Value.
    result1.smb <- smbinning(DF0, 'GB_flag', val)
  
    if (class(result1.smb) == 'list') {
      binning.df[i, 'IV-RPart'] <- result1.smb$iv
      binning.df[i, 'N-RPart']  <- result1.smb$bands %>% length
    }
  
    # Custom Binning Based by cutpoints using percentiles (10% each)
    if (length(NamesOfVariables) <= Max_Vars || class(result1.smb) != 'list') {
      cbs1cuts <- as.vector(quantile(DF0[, val], probs=seq(0, 1, 0.1), na.rm=TRUE)) # Quantiles by 10%
      cbs1cuts <- cbs1cuts[2:(length(cbs1cuts) - 1)] # Remove first (min) and last (max) values
      result2.smb <- smbinning.custom(df=DF0, y = 'GB_flag', x = val, cuts = cbs1cuts)
      binning.df[i, 'IV-Decile'] <- result2.smb$iv
      binning.df[i, 'N-Decile']  <- result2.smb$bands %>% length
    } else {
      binning.df[i, 'IV-Decile'] <- 0
      binning.df[i, 'N-Decile']  <- ncol(DF0) + 1
    }
      
    if (length(NamesOfVariables) <= Max_Vars ) { # & !isMicrosoftRServer
      # Finer Monotonic Binning Based on Isotonic Regression - Do Not working with Microsoft R Server 9.3.0
      result3.smb <- isobin(DF0, 'GB_flag', val)
      binning.df[i, 'IV-Iso'] <- result3.smb$iv
      binning.df[i, 'N-Iso']  <- result3.smb$bands %>% length
    
      # Generates a supervised tree-like segmentation of numeric variables with respect to a binary target outcome
      # result4.smb <- tree_chimergebin(DF0, 'GB_flag', val)
      cbs1cuts <- tree_bin(DF0, 'GB_flag', val)   # Binning via Tree-Like Segmentation
      result4.smb <- smbinning.custom(df = DF0, x = val, y = 'GB_flag', cuts = cbs1cuts)
      if (class(result4.smb) == 'list') {  
        binning.df[i, 'IV-Tree'] <- result4.smb$iv
        binning.df[i, 'N-Tree']  <- result4.smb$bands %>% length
      } else {  # 'Not Meaningful (IV<0.1)' or 'Uniques values < 5' case
        binning.df[i, 'IV-Tree'] <- 0
        binning.df[i, 'N-Tree']  <- ncol(DF0)
      }
    } else {
      binning.df[i, 'IV-Iso'] <- 0
      binning.df[i, 'N-Iso']  <- ncol(DF0)
      
      binning.df[i, 'IV-Bad'] <- 0
      binning.df[i, 'N-Bad']  <- ncol(DF0)
    }
    
    # Selection of the Optimal Binning Method
    if (if_else(is.na(binning.df[i, 'IV-RPart']) == TRUE, 0, binning.df[i, 'IV-RPart'] * 1.1) >
          binning.df[i, 'IV-Decile']) {
                        binning.df[i, 'Method'] <- 'IV-RPart'
    } else 
      {
      if ( ( (binning.df[i, 'IV-Iso'] > binning.df[i, 'IV-Decile']) & 
             (binning.df[i, 'N-Iso'] / 1.1 < binning.df[i, 'N-Decile']) ) |
           ( (binning.df[i, 'IV-Iso'] * 1.1 > binning.df[i, 'IV-Decile']) & 
             (binning.df[i, 'N-Iso'] * 2 < binning.df[i, 'N-Decile']) ) ) {
                        binning.df[i, 'Method'] <- 'IV-Iso'
      } else {
        if (binning.df[i, 'IV-Decile'] >= binning.df[i, 'IV-Iso']) { 
                        binning.df[i, 'Method'] <- 'IV-Decile'
              } else { 
                  binning.df[i, 'Method'] <- 'IV-Iso' } 
              }
      
      }  # End Else If

    type <- binning.df[i, 'Method']
    result.smb <- 
      switch(type,
             `IV-RPart` = result1.smb,
             `IV-Decile` = result2.smb,
             `IV-Iso` = result3.smb,
             `IV-Bad` = result4.smb
             )
    binning.df[i, 'IV'] <- result.smb$iv
  
    # IV Table Supplement
    result.smb$ivtable <- result.smb$ivtable %>%
      mutate(G_Dis = CntGood / table(DF0$GB_flag)[2],
             B_Dis = CntBad / table(DF0$GB_flag)[1],
             `G/B Index` = if_else(G_Dis > B_Dis, G_Dis / B_Dis, B_Dis / G_Dis),
             `0=Good, 1=Bad` = if_else(G_Dis > B_Dis, 0, 1),
             Bin = c(1:(nrow(result.smb$ivtable) - 1), NA),
             Min = c(NA, result.smb$cuts, NA, NA),
             Max = c(result.smb$cuts, NA, NA, NA)
      )

  } #  End else for numeric class 
  
  # Prepare MySQL-code for Binning and Fine Classing
  # Sys.setloc <- Sys.setlocale(locale = 'Russian') # set locale to `Russian`
  binning.sql <- capture.output(smbinning.sql(result.smb)) %>% 
    gsub("then '0", "then '", .) %>% 
      gsub('TableName', 'DF', .) %>%
        stringr::str_replace('NewCharName', paste0(val, '_fct')) 
  
  val_fct <- paste0('  \'', val, '_fct', '\'', '  from data.frame \'DF0\'')
  binning.sql <- c('select *,', paste0('      /*   Inserting the new factor variable', val_fct, '  */'),
                    binning.sql[-c(1:3)] , paste0('  \'', val, '_fct', '\'', ' from \'DF1\''))

  # Truncation of the Name of the Gradation in a Complex Set of levels
  theBest <- TRUE
  for (j in 1:(nrow(result.smb[['ivtable']]) - 2)) {
    if (str_length(binning.sql[j + 3]) > 999) {
      gradation_str <- str_split( string = binning.sql[j + 3], pattern = sprintf('%s: %s ', j, val) ) %>% 
        unlist
      binning.sql[j + 3] <- 
        paste0(gradation_str[1], ifelse( theBest, sprintf('%s: %s the Best\'', j, val), 
                                          sprintf('%s: %s the Worst\'', j, val) ) )
      theBest <= FALSE
    } else {
      # print('empty')
    }
  }
  
  # Appending binning.sql into TotalBinning.sql
  if (i == 1) {
    binning.sql[length(binning.sql)] = paste0('  \'', val, '_fct\',')
    TotalBinning.sql <- binning.sql
  } else {
    if (i == length(NamesOfVariables)) {
      TotalBinning.sql <- c(TotalBinning.sql, '', binning.sql[-1] )
    } else {
      TotalBinning.sql <- c(TotalBinning.sql, '', binning.sql[-c(1, length(binning.sql))], 
                            paste0('  \'', val, '_fct\','))
    }
  } # End if (i == 1)
  # TotalBinning.sql <- c(TotalBinning.sql, '', binning.sql)
  
  # Preparing a Data.Frame with VI Table for Export into MS Excel
  addWorksheet(wb, val, gridLines = FALSE, tabColour = ifelse(result.smb$iv >= 0.05, 'chartreuse',
                                                              ifelse(result.smb$iv >= 0.03, 'khaki', 'white')))
  result.smb$ivtable[ is.na( result.smb$ivtable ) ] <- NA  # Dealing with NaN's in data frames
  N <- 3:(nrow(result.smb$ivtable) + 2)
  
  result.smb$ivtable %>% 
    dplyr::select(Cutpoint, Bin, Min, Max, CntRec, CntGood, CntBad,  G_Dis, B_Dis, Share = PctRec, 
                  BadRate, WoE, IV,`G/B Index`, `0=Good, 1=Bad`) %>% 
      writeDataTable(wb, sheet = val, x = ., tableStyle = 'TableStyleMedium2', startCol = 'A',
                     startRow = 2, tableName = val, firstColumn = TRUE, lastColumn = FALSE, bandedRows = TRUE)
  # Set Columns widths
  setColWidths(wb, sheet = val, cols = 1:4, widths = c(32, 7, 10, 10))
  
  # # Set Row heights
  # setRowHeights(wb, sheet = 1, rows = 1, heights = 45)
  
  # Set Styles & Conditional Formattings in Columns
  addStyle(wb, sheet = val, style = createStyle(wrapText = TRUE, halign = 'center', valign = 'center'),
                                              cols = 1:ncol(result.smb$ivtable), rows = 2)
  addStyle(wb, sheet = val, cols = 1, rows = 1, style = createStyle(fontSize = 16, textDecoration = 'bold'))
  addStyle(wb, sheet = val, cols = 1:ncol(result.smb$ivtable), rows = (nrow(result.smb$ivtable) + 2), 
           style = createStyle(textDecoration = 'bold'))
  addStyle(wb, sheet = val, cols = 5:7, rows = N, style = createStyle(numFmt = 'COMMA'), gridExpand = TRUE)
  addStyle(wb, sheet = val, cols = 8:10, rows = N, style = createStyle(numFmt = '0%'), gridExpand = TRUE)
  addStyle(wb, sheet = val, cols = 11, rows = N, style = createStyle(numFmt = paste0('0', options()$OutDec, '0%'), textDecoration = 'bold'))
  conditionalFormatting(wb, sheet = val, cols = 13, rows = 3:(nrow(result.smb$ivtable) + 1), type = 'databar',
                        border = FALSE, style = c('red', 'chartreuse'))
  addStyle(wb, sheet = val, cols = 4, rows = N, style = createStyle(border = 'right', borderColour = '#4F81BC'))
  addStyle(wb, sheet = val, cols = 7, rows = N, style = createStyle(numFmt = 'COMMA', border = 'right',
                                                                    borderColour = '#4F81BC'))
  addStyle(wb, sheet = val, cols = 10, rows = N, style = createStyle(numFmt = '0%', border = 'right',
                                                                     borderColour = '#4F81BC'))

  conditionalFormatting(wb, sheet = val, cols = 15, rows = 3:(nrow(result.smb$ivtable) + 1), rule ='$O3=0',
      style = createStyle(fontColour = 'red', halign = 'center', valign = 'center', textDecoration = 'bold'))
  conditionalFormatting(wb, sheet = val, cols = 15, rows = 3:(nrow(result.smb$ivtable) + 1), rule ='$O3>0',
      style = createStyle(fontColour = 'black', halign = 'center', valign = 'center', textDecoration = 'bold'))

  writeData(wb, sheet = val, val, startCol = 'A', startRow = 1)
  writeData(wb, sheet = val, data.frame(binning.sql), colNames = FALSE, rowNames = FALSE,
            startCol = 'A', startRow = nrow(result.smb$ivtable) + 4)

  writeFormula(wb, sheet = val, startCol = 'A', 
               startRow = nrow(result.smb$ivtable) + 5 + length(binning.sql), 
               x = makeHyperlinkString(sheet = 'IV Table', row = i + 2, col = 1,
                                       text = 'Link to IV Table'))
}     # End next i

## 1 - disbursed_amount
## 2 - asset_cost
## 3 - ltv
## 4 - branch_id
## 5 - supplier_id
## 6 - manufacturer_id
## 7 - Current_pincode_ID
## 8 - Employment_Type
## 9 - State_ID
## 10 - Employee_code_ID
## 11 - Aadhar_flag
## 12 - PAN_flag
## 13 - VoterID_flag
## 14 - Driving_flag
## 15 - Passport_flag
## 16 - PERFORM_CNS_SCORE
## 17 - PERFORM_CNS_SCORE_DESCRIPTION
## 18 - PRI_NO_OF_ACCTS
## 19 - PRI_ACTIVE_ACCTS
## 20 - PRI_OVERDUE_ACCTS
## 21 - PRI_SANCTIONED_AMOUNT
## 22 - PRI_DISBURSED_AMOUNT
## 23 - SEC_NO_OF_ACCTS
## 24 - SEC_ACTIVE_ACCTS
## 25 - SEC_OVERDUE_ACCTS
## 26 - SEC_CURRENT_BALANCE
## 27 - SEC_SANCTIONED_AMOUNT
## 28 - SEC_DISBURSED_AMOUNT
## 29 - PRIMARY_INSTAL_AMT
## 30 - SEC_INSTAL_AMT
## 31 - NEW_ACCTS_IN_LAST_SIX_MONTHS
## 32 - DELINQUENT_ACCTS_IN_LAST_SIX_M
## 33 - NO_OF_INQUIRIES
## 34 - Age
## 35 - YearsSinceDisbursment
## 36 - AverageLoanTenure
## 37 - TimeSinceFirstLoan
## 38 - YearsOnLoan
## 39 - DisAsDiff
## 40 - DisAsShare
## 41 - Qrt
## 42 - Day
## 43 - OutstandingNow
## 44 - DisbursedTotal
## 45 - ShareOverdue
## 46 - SEC_OverdueShareSec
## 47 - PRI_OverdueShare

# Flag Recovery after Optimal Binning
DF0 %<>% dplyr::mutate( GB_flag = ifelse(GB_flag == 1, 0, 1) )

# Write MySQL code for Coarse Classing Selected Variables
write_lines(x = TotalBinning.sql, path = NameMySQLCode, na = "NA", append = FALSE)

N <- 3:(nrow(binning.df) + 2)
# writeDataTable(wb, sheet = 'IV Table', x = binning.df, tableStyle = 'TableStyleMedium4', startCol = 'A',
#               startRow = 2, tableName = 'IVTable', firstColumn = FALSE, lastColumn = TRUE, bandedRows = TRUE)

# Set Columns widths
setColWidths(wb, sheet = 'IV Table', cols = 1:2, widths = c(15, 12))
setColWidths(wb, sheet = 'IV Table', cols = 11, widths = c(12))

conditionalFormatting(wb, sheet = 'IV Table', cols = 2, rows = N, type = 'databar',
                      border = FALSE, style = c('red', 'royalblue'))
addStyle(wb, sheet = 'IV Table', cols = 2, rows = N, 
         style = createStyle(border = 'right', borderColour = '#9CB95C'))
addStyle(wb, sheet = 'IV Table', cols = 10, rows = N, 
         style = createStyle(border = 'right', borderColour = '#9CB95C'))

writeData(wb, sheet = 'IV Table', 'IV Table', startCol = 'A', startRow = 1)
addStyle(wb, sheet = 'IV Table', cols = 1, rows = 1, 
         style = createStyle(fontSize = 16, textDecoration = 'bold'))

for (i in 1:nrow(binning.df)) {
  ## Internal - Text to display
  val = binning.df[i, 'Variable']
  writeFormula(wb, sheet = 'IV Table', startCol = 'A', startRow = i + 2, 
    x = makeHyperlinkString(sheet = val, row = 1, col = 1, text = val))
}

# # Open MS Excel
# openXL(wb)

 remove(NamesOfVariables, result1.smb, result2.smb, result3.smb, result4.smb, result.smb, chr_vec, # binning.df,
        j, TotalBinning.sql, theBest, binning.sql, val, val_fct, i, type, cbs1cuts, N)

Coarse Classing Factor and Numeric Variables

1. Fine Classing

Create 10/20 bins/groups for a continuous independent variable and then calculates WOE and IV of the variable

2. Coarse Classing

Combine adjacent categories with similar WOE scores

Rules related to WOE

• Each category (bin) should have at least 5% of the observations.

• Each category (bin) should be non-zero for both non-events and events.

• The WOE should be distinct for each category. Similar groups should be aggregated.

• The WOE should be monotonic, i.e. either growing or decreasing with the groupings.

Missing values are binned separately.

# use library('sqlrutils') from Microsoft Corporation
library('sqldf') # Manipulate R Data Frames Using SQL

# Prepare MySQL-code for Coarse Classing
DF <- sqldf::sqldf(x = read_file( NameMySQLCode ), method = 'name__class') %>%  # PRIORITY_DAYS - 'difftime'
  dplyr::select(c(one_of('GB_flag'), ends_with('_fct'))) %>% 
    dplyr::mutate_all(as.factor)

remove(DF0, DF1)
# 
# SelectedFeatures <- c('ltv_fct', 'branch_id_fct', 'supplier_id_fct', 'Current_pincode_ID_fct', 'State_ID_fct', 'Employee_code_ID_fct', 'Passport_flag_fct', 'PERFORM_CNS_SCORE_DESCRIPTION_fct', 'PRI_OVERDUE_ACCTS_fct', 'PRI_DISBURSED_AMOUNT_fct', 'SEC_CURRENT_BALANCE_fct', 'NO_OF_INQUIRIES_fct', 'Age_fct', 'YearsSinceDisbursment_fct', 'TimeSinceFirstLoan_fct')
# 
# DF <- DF[, c('GB_flag', SelectedFeatures)]

Create Outcome - Dependent Binary Factor

Let’s form an array of independent factors (predictors) and a outcome - dependent binary factor.

# Сonverting the Integer Variables into Factor
X <- DF %>% 
  dplyr::select(-GB_flag)

Y <- DF %$%                                                # Pull out a single variable into a vector throu %$%
  ifelse(GB_flag == '1', 'Bad', 'Good') %>%               
    factor(levels = c('Bad', 'Good'))                      # Correct levels for Scoring Model

Split Train and Testing Sets

It is a good idea to use a validation hold out set. This is a sample of the data that we hold back from our analysis and modeling. We use it right at the end of our project to confirm the accuracy of our final model. It is a smoke test that we can use to see if we messed up and to give us confidence on our estimates of accuracy on unseen data.

# Use train / test data set

table(Y[inTrain]) ## significant Class Imbalance Problem

## 
##    Bad   Good 
##  45557 164281

# Weights of cases to resolve Class Imbalances Problem
weight.cases <- as.numeric( Y[inTrain]  )
for(val in unique(weight.cases)) {weight.cases[weight.cases==val]=
    1/sum(weight.cases==val)*length(weight.cases)/2} # normalized to sum to length(samples)

# # To list all columns with missing values you might write: 
# is.na(X) %>% colSums()
#
# # To list all rows with missing values you might write: 
# X[!complete.cases(X), ]

Population Proportion – Sample Size

\[ \displaystyle \large X = \frac{Z_{\alpha/2}^2 × p × (1\ - \ p)}{\beta^2} \hspace{.5 in} [1]\]

where, \(Z_{\alpha/2}\) is the critical value of the Normal distribution at \(\alpha\) (e.g. for a confidence level of 98%, \({\alpha}\) (Type I Error or False Positive) is 0.02, so \(alpha/2\) = 0.01 and \(Z_{\alpha/2}\) - the critical value is 2.326) and \({\beta}\) for calculating the Margin of Error (Type II Error or False Negative) is 0.05,

• \(\beta\) is the margin of error by (1 - Power),

• p is the sample proportion,

and X is the population size.

# Population Proportion – Minimal Sample Size
# Source: Daniel  WW.  Biostatistics:  A  Foundation for   Analysis   in   the   Health   Sciences.   7th edition. New York: John Wiley & Sons. (1999)
# https://select-statistics.co.uk/calculators/sample-size-calculator-population-proportion/

yName <- 'GB_flag'  # Name of Outcome
yTarget <- 1        # Name of Outcome's Positive Level

writeLines(sprintf('Sampling into two unequal sample sizes with Fraction (%0.2f%%) of Training & Test Sets. \n',
                   length(Y[inTrain]) / length(Y) * 100))

## Sampling into two unequal sample sizes with Fraction (60.73%) of Training & Test Sets.

writeLines(sprintf('Full Dataset has Real Sample Size is equal %g observations.
  Therefore the Training Dateset might be %g obs. and the Test Dateset - %g obs. \n',
                   length(Y), round(length(Y) * 0.70, -4), length(Y) - round(length(Y) * 0.70, -4)))

## Full Dataset has Real Sample Size is equal 345546 observations.
##   Therefore the Training Dateset might be 240000 obs. and the Test Dateset - 105546 obs.

alpha = 0.05        # Type I Error or False Positive - α probability (Significance level) fro Two tailed
beta =  0.10        # Type II Error or False Negative - β probability level

BadRate <- table(Y[inTrain]) %>% .['Bad'] / length(Y[inTrain])
# MSS <- (qnorm(1 - alpha/2)^2 * BadRate * (1 - BadRate)) / (beta)^2       # BadRate *

# *********************************************************************************************************** #
# 
# Power Analysis for two Proportions (different sample sizes) of The Binomial Distributions
#
# *********************************************************************************************************** #

# Fisher's exact test is a statistical significance test used for Contingency Tables (Binomial Distribution)
matrix(c(table(Y[inTrain]), table(Y[inTest])),
       nrow = 2,
       dimnames = list(Subjects = c('Bad', 'Good'),
                           Samples = c('inTrain', 'inTest'))) %T>%
    print() %>% 
stats::fisher.test(., alternative = 'two', conf.level = 1-alpha)

##         Samples
## Subjects inTrain inTest
##     Bad    45557   5054
##     Good  164281  18262

## 
##  Fisher's Exact Test for Count Data
## 
## data:  .
## p-value = 0.9067
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##  0.9695789 1.0356963
## sample estimates:
## odds ratio 
##    1.00203

# # Null Hypothesis Significance Testing
# pB <-
#   ShowNullHypothesisSignificanceTesting(pA = BadRate,
#                                         pT = table(Y[inTest]) %>% .['Bad'] / length(Y[inTest]),
#                                         nA = length(Y[inTrain]),
#                                         nB = length(Y[inTest]),
#                                         alpha = alpha, 
#                                         beta = beta)

# Example of using `uniroot` function - https://rpubs.com/chidungkt/408264

frac <- length(Y[inTrain]) / (length(Y[inTrain]) + length(Y[inTest]))
# Set First functions: Power and Sample Size for Two-Sample Binomial Test
func1 <- function(x) { sizes <- Hmisc::bsamsize(p1 = BadRate, p2 = BadRate + x, 
                                       fraction = frac, alpha = alpha, power = (1 - beta))
                       return(sizes[2]) }  # B (Control) Sample Size

# Equilibrium quantity: Threshold for h - Effect Size (Cohen’s d)
pThreshold <- stats::uniroot( function(x) func1(x) - (length(Y[inTest]) * 0.98), c(0, BadRate) )$root

# Estimating Sample size(s) for Two unequivalent Binomial Sample 

library('pwr')              # Basic Functions for Power Analysis

( estimate <- pwr::pwr.2p2n.test( h = pwr::ES.h(p1 = BadRate,   # Treatment Probability of Success
                                    p2 = BadRate + pThreshold ) # Hypothesized control probability of success 
                                 , n1 = length(Y[inTrain])      # Sample Size for Training or A (Treatment) Set
                                 , n2 = NULL                    # Sample Size for Test or B (Control) Set
                                 , sig.level = alpha            # Type I Error or False Positive
                                 , power = (1 - beta)           # Power of Test (1 - Type II Error)
                                 , alternative = 'two.sided') )

## 
##      difference of proportion power calculation for binomial distribution (arcsine transformation) 
## 
##               h = 0.02259551
##              n1 = 209838
##              n2 = 22818.26
##       sig.level = 0.05
##           power = 0.9
##     alternative = two.sided
## 
## NOTE: different sample sizes

MSS <- round(estimate$n2 + 0.5)                                 # Minimal Sample Size for Test (Control) Set

pwr::plot.power.htest(estimate) + 
    ggplot2::annotate(geom = 'text', x = 0, y = 0, size = 3.5, hjust = 'left', 
                      label = sprintf('Badrate in { %0.2f%%, %0.2f%% }',
                                      (BadRate - pThreshold) * 100, (BadRate + pThreshold) * 100))

# MSS <- Hmisc::bsamsize(p1 = table(Y[inTrain]) %>% .['Bad'] / length(Y[inTrain]),
#                        p2 = table(Y[inTrain]) %>% .['Bad'] / length(Y[inTrain]) - 0.005,
#                        fraction = 0.7, alpha = alpha, power = (1 - beta)) %>%
#   round(.[2] + 0.5)

writeLines(sprintf('With Treatment Probability of Success (Bad Rate on Training dataset = %.2f%%) and Bias = %0.2f%%, Significance level (1 - alpha = %.1f%%) & Power (1 - beta = %.1f%%) Minimal Sample Size of the Binomial distribution should be %g obs., but Test set is %g obs. \n',
                   BadRate * 100, pThreshold * 100, (1 - alpha) * 100, (1- beta) * 100, MSS, length(Y[inTest]) ))

## With Treatment Probability of Success (Bad Rate on Training dataset = 21.71%) and Bias = 0.94%, Significance level (1 - alpha = 95.0%) & Power (1 - beta = 90.0%) Minimal Sample Size of the Binomial distribution should be 22819 obs., but Test set is 23316 obs.

# Weights of cases to resolve Class Imbalances Problem
weight.cases <- DT[inTrain, ] %>%
  dplyr::pull(!!yName) %>%
    as.double()
for(val in unique(weight.cases)) {weight.cases[weight.cases==val]=
    1/sum(weight.cases==val)*length(weight.cases)/2} # normalized to sum to length(samples)

remove(alpha, beta, BadRate, MSS, estimate, pThreshold, frac)

Unimodal Data Visualizations: Information Values & Distrubutions

Let’s look at visualizations of individual attributes. It is often useful to look at your data using multiple different visualizations in order to spark ideas. Let’s look at histograms of each attribute to get a sense of the data distributions.

# Density plot for each Features vs `Y`

ShowUnimodalVisualizations(X[inTrain, ], Y[inTrain], isMicrosoftRServer)

## There are no any Numeric feature in `X` data.frame.
## There are no any Factor or Character feature in `X` data.frame.

Important Points

1. Information value increases as bins / groups increases for an independent variable. Be careful when there are more than 20 bins as some bins may have a very few number of events and non-events.

2. Information value should not be used as a feature selection method when you are building a classification model other than binary logistic regression (for example, random forest or SVM) as it’s designed for binary logistic regression model only.

Multimodal Data Visualizations - The Correlation Matrix

Let’s look at some visualizations of the interactions between outcome and predictors. The best place to start is a scatter plot matrix.

## 
## Attaching package: 'plotly'

## The following object is masked from 'package:Hmisc':
## 
##     subplot

## The following object is masked from 'package:ggplot2':
## 
##     last_plot

## The following object is masked from 'package:stats':
## 
##     filter

## The following object is masked from 'package:graphics':
## 
##     layout

## Loading required package: caret

## 
## Attaching package: 'caret'

## The following object is masked from 'package:survival':
## 
##     cluster

## The following object is masked from 'package:purrr':
## 
##     lift

## Loading required package: reshape2

## 
## Attaching package: 'reshape2'

## The following objects are masked from 'package:data.table':
## 
##     dcast, melt

## The following object is masked from 'package:tidyr':
## 
##     smiths

##   Var1                                                Var2 Spearman
## 1    Y     Employee_code_ID_fct.2..Employee_code_ID....153   0.0743
## 2    Y                  DisAsDiff_fct.5..DisAsDiff...19822   0.0679
## 3    Y Current_pincode_ID_fct.3..Current_pincode_ID....174   0.0673
## 4    Y               DisAsShare_fct.7..DisAsShare...1.7649   0.0655
## 5    Y     Employee_code_ID_fct.3..Employee_code_ID....170   0.0645
## 6    Y   PERFORM_CNS_SCORE_fct.6..PERFORM_CNS_SCORE....824   0.0644
## 7    Y               supplier_id_fct.3..supplier_id....165   0.0573
##   Var1                                                 Var2 Spearman
## 1    Y  Current_pincode_ID_fct.12..Current_pincode_ID...291  -0.1252
## 2    Y      Employee_code_ID_fct.15..Employee_code_ID...319  -0.1147
## 3    Y                supplier_id_fct.13..supplier_id...314  -0.1057
## 4    Y       OutstandingNow_fct.3..OutstandingNow....171384  -0.0907
## 5    Y Current_pincode_ID_fct.11..Current_pincode_ID....291  -0.0839
## 6    Y     Employee_code_ID_fct.14..Employee_code_ID....319  -0.0823
## 7    Y               supplier_id_fct.12..supplier_id....314  -0.0809

This helps point out the skew in many distributions so much so that data looks like outliers (e.g. beyond the whisker of the plots).

Select Features

Feature Selection (removing correlated attributes or reduce the quality of classification), but Transformations (Box-Cox or YeoJohnson) could not apply to factors.

#
#       Filtering predictors - Remove Redundant Variables
#
library('caret')            # Classification and Regression Training 

# # 1a. Imputing Missing Value, Centering, Scaling and Transformation, for example 'YeoJohnson'
# X <- cbind(X %>% select_if(., is.numeric) %>%
#             predict(preProcess( ., method = c( 'bagImpute' ) ), newdata = . ) #,
# 
# # 1b. Convert factor (nominal and ordered) variables into a full set of dummy integer variables without linear dependencies induced between these predictors
#            X %>% select_if(., is.factor) %>%
#              data.frame(predict(dummyVars(' ~ .', data = ., fullRank = TRUE), newdata = .)) %>%
#                select_if(., is.numeric)
#       )

# 2. Dropping zero variance predictors: the cutoff for the ratio of the first most common value to the second value. See https://www.mql5.com/ru/articles/2029
nzv <- caret::nearZeroVar( X[inTrain, ], freqCut = 999/1 , saveMetrics= TRUE )
nzv[nzv$nzv,]

## [1] freqRatio     percentUnique zeroVar       nzv          
## <0 rows> (or 0-length row.names)

zv_cols = caret::nearZeroVar( X[inTrain, ], freqCut = 999/1, saveMetrics = FALSE )
print( sprintf('Dropping %d zero variance predictors from %d (fraction=%10.6f)',
        length(zv_cols), dim(X[inTrain, ])[2], length(zv_cols)/dim(X[inTrain, ])[2]) )

## [1] "Dropping 0 zero variance predictors from 47 (fraction=  0.000000)"

zv_cols

## integer(0)

# class(X) <- 'data.frame'
if ( length(zv_cols) != 0 )  {
      X <- X[, -zv_cols]
      }

# # 3. Centering, Scaling and Transformation or Principal Components # , 'pca'
# X[is.na(X)] <- -1        #  replace the <NA>s with zeros in number columns
# X <- X %>% predict(preProcess( X[inTrain, ], method = c( 'medianImpute', 'center', 'scale', 'YeoJohnson', 'pca' ), thresh = 0.99 ), newdata = . ) # %>% # Define all logical fields into numeric type
#         # spatialSign %>% data.frame

# 4. Remove NUMERIC variables with high correlation (> .70) to others (multicollinearity)

  cor.matrix <- cor( sapply( X[inTrain, ], function(x)
                    { as.numeric(x) } ) )
  cor.high   <- caret::findCorrelation(cor.matrix, cutoff = 0.80, verbose = TRUE,  names = FALSE, exact = TRUE)

## Compare row 18  and column  19 with corr  0.869 
##   Means:  0.239 vs 0.104 so flagging column 18 
## Compare row 37  and column  36 with corr  0.843 
##   Means:  0.211 vs 0.098 so flagging column 37 
## Compare row 44  and column  43 with corr  0.871 
##   Means:  0.206 vs 0.094 so flagging column 44 
## Compare row 36  and column  47 with corr  0.846 
##   Means:  0.186 vs 0.088 so flagging column 36 
## Compare row 47  and column  16 with corr  0.883 
##   Means:  0.162 vs 0.084 so flagging column 47 
## Compare row 21  and column  22 with corr  0.99 
##   Means:  0.132 vs 0.081 so flagging column 21 
## Compare row 31  and column  45 with corr  0.949 
##   Means:  0.134 vs 0.079 so flagging column 31 
## Compare row 23  and column  24 with corr  0.803 
##   Means:  0.143 vs 0.075 so flagging column 23 
## Compare row 24  and column  27 with corr  0.989 
##   Means:  0.129 vs 0.072 so flagging column 24 
## Compare row 27  and column  28 with corr  0.995 
##   Means:  0.106 vs 0.069 so flagging column 27 
## Compare row 28  and column  26 with corr  0.933 
##   Means:  0.081 vs 0.068 so flagging column 28 
## Compare row 40  and column  39 with corr  0.812 
##   Means:  0.102 vs 0.067 so flagging column 40 
## Compare row 34  and column  38 with corr  0.996 
##   Means:  0.082 vs 0.065 so flagging column 34 
## Compare row 11  and column  13 with corr  0.869 
##   Means:  0.093 vs 0.064 so flagging column 11 
## Compare row 35  and column  41 with corr  0.824 
##   Means:  0.064 vs 0.063 so flagging column 35 
## All correlations <= 0.8

  high.cor.remove <- row.names(cor.matrix)[cor.high]
  print( sprintf('Dropping %d predictors due to high correlation to others (multicollinearity) %d (fraction=%10.6f)',
                 length(high.cor.remove), dim(X[inTrain, ])[2],  length(high.cor.remove)/dim(X[inTrain, ])[2]) );

## [1] "Dropping 15 predictors due to high correlation to others (multicollinearity) 47 (fraction=  0.319149)"

  high.cor.remove

##  [1] "PRI_NO_OF_ACCTS_fct"              "TimeSinceFirstLoan_fct"          
##  [3] "DisbursedTotal_fct"               "AverageLoanTenure_fct"           
##  [5] "PRI_OverdueShare_fct"             "PRI_SANCTIONED_AMOUNT_fct"       
##  [7] "NEW_ACCTS_IN_LAST_SIX_MONTHS_fct" "SEC_NO_OF_ACCTS_fct"             
##  [9] "SEC_ACTIVE_ACCTS_fct"             "SEC_SANCTIONED_AMOUNT_fct"       
## [11] "SEC_DISBURSED_AMOUNT_fct"         "DisAsShare_fct"                  
## [13] "Age_fct"                          "Aadhar_flag_fct"                 
## [15] "YearsSinceDisbursment_fct"

  if (length(high.cor.remove) != 0) {
        X <- X[, -cor.high]
     }

# Remove some other features that do not add useful information for Machine Learning
remove(nzv, zv_cols, cor.matrix, cor.high, high.cor.remove)

# Remaining Variables
print( sprintf('Remaining %d Variables', dim(X)[2]) )

## [1] "Remaining 32 Variables"

names( X )

##  [1] "disbursed_amount_fct"              
##  [2] "asset_cost_fct"                    
##  [3] "ltv_fct"                           
##  [4] "branch_id_fct"                     
##  [5] "supplier_id_fct"                   
##  [6] "manufacturer_id_fct"               
##  [7] "Current_pincode_ID_fct"            
##  [8] "Employment_Type_fct"               
##  [9] "State_ID_fct"                      
## [10] "Employee_code_ID_fct"              
## [11] "PAN_flag_fct"                      
## [12] "VoterID_flag_fct"                  
## [13] "Driving_flag_fct"                  
## [14] "Passport_flag_fct"                 
## [15] "PERFORM_CNS_SCORE_fct"             
## [16] "PERFORM_CNS_SCORE_DESCRIPTION_fct" 
## [17] "PRI_ACTIVE_ACCTS_fct"              
## [18] "PRI_OVERDUE_ACCTS_fct"             
## [19] "PRI_DISBURSED_AMOUNT_fct"          
## [20] "SEC_OVERDUE_ACCTS_fct"             
## [21] "SEC_CURRENT_BALANCE_fct"           
## [22] "PRIMARY_INSTAL_AMT_fct"            
## [23] "SEC_INSTAL_AMT_fct"                
## [24] "DELINQUENT_ACCTS_IN_LAST_SIX_M_fct"
## [25] "NO_OF_INQUIRIES_fct"               
## [26] "YearsOnLoan_fct"                   
## [27] "DisAsDiff_fct"                     
## [28] "Qrt_fct"                           
## [29] "Day_fct"                           
## [30] "OutstandingNow_fct"                
## [31] "ShareOverdue_fct"                  
## [32] "SEC_OverdueShareSec_fct"

Entropy-Based Feature Selection

Different methods for calculating the feature importance are built into FSelectorRcpp’s function information_gain(). I recommend using a fast but effective method FSelectorRcpp_information.gain written in C++ from package FSelectorRcpp.

if (ncol(X) <= Max_Vars) {

  library('FSelectorRcpp')    # 'Rcpp' Implementation of 'FSelector' Entropy-Based Feature Selection Algorithms with a Sparse Matrix Support
  
  library('parallel')         # Support for Parallel computation in R
  # Support for Parallel computation in R
  ncore <- parallel::detectCores()
  (cl = parallel::makeCluster(ncore))
  
  # Entropy-Based Feature Selection Algorithms with a Sparse Matrix Support
  Entropy_Based_Features <- 
    FSelectorRcpp::information_gain(        # Calculate the score for each attribute
        formula = as.formula('Y ~ .'),      # that is on the right side of the formula.
        data = cbind(Y = Y, X)[inTrain, ],  # Attributes must exist in the passed data.
        type  = 'infogain',                 # Choose the type of a score to be calculated.
        threads = ncore                     # Set number of threads in a parallel backend.
    ) %>% 
    dplyr::arrange(-importance) %>%   # Sort of Entropy-based Importance by Descending.
      dplyr::slice(1:Max_Vars)        # Selection the First `Max_Vars` features.
    
  knitr::kable(Entropy_Based_Features, caption = sprintf('Selection of the %g Important Variables vs. `%s` outcome', Max_Vars, yName))    
  # utils::writeClipboard(Entropy_Based_Features[['attributes']])
  
  X <- X[, Entropy_Based_Features$attributes]
  
  if(!is.null(cl)) {
      parallel::stopCluster(cl)
      cl = NULL
  }
  
  remove(Entropy_Based_Features)

}

Simple Logit-Model

The Logit in logistic regression is a special case of a link function in a Generalized Linear Model (GLM): it is the canonical link function for the Bernoulli distribution.

The logistic model is usually represented as:

\[ \displaystyle \large \pi(Y)=\frac{\exp(\beta_0+\beta_1X)}{1+\exp(\beta_0+\beta_1X)} \hspace{.5 in} [2]\] or going into the common linear regression model:

\[ \displaystyle \large \ln\left(\frac{\pi(Y)}{1-\pi(Y)}\right)=\beta_0+\beta_1X \hspace{.5 in} [3]\]

Therefore, logit itself is obtained:

\[ \displaystyle \large \Pr(Y=1 \mid X) = [1 + e^{-X'\beta}]^{-1} \hspace{.5 in} [4]\]

# Train Logistic Regression Model with limited memory Broyden-FletcherGoldfarb-Shanno (L-BFGS) optimization 

# Microsoft Machine Learning Example - https://docs.microsoft.com/en-us/machine-learning-server/r/sample-solutions#loan-credit-risk
writeLines('\n Logit-Model on Training Set')

## 
##  Logit-Model on Training Set

# The named list containing components upper and lower, both formulae, defining the range of models to be examined in the stepwise search
scope <- list(
   lower = ~ supplier_id_fct,
   upper = as.formula( paste( '~', paste(names(X), collapse = ' + ') ) ) )

## rxLogit / variableSelection
varsel <- 
  RevoScaleR::rxStepControl(method = 'stepwise'
            , keepStepCoefs = TRUE  # flag specifying whether or not to keep the model coefficients at each step
            , scope = scope
            , stepCriterion = 'SigLevel' ) # significance level, the traditional stepwise approach in SAS

# # Ranking Features by Importance using rx (Random Forest)
# rf_model <-
#   RevoScaleR::rxDForest(  formula = paste0('Y ~ ', paste(colnames(X), collapse = '+'))
#                         , seed = seed
#                         , importance = TRUE
#                         , pweights = 'Pweights'
#                         , data = cbind(Y = Y[inTrain], X[inTrain, ], Pweights = weight.cases) )
# View(rf_model$importance )

system.time(
  rxLogitFit <- # Create a Logistic Regression Model
    RevoScaleR::rxLogit(
      formula = paste('Y ~ ', paste(names(X), collapse = '+')) 
      # , variableSelection = varsel
      , pweights = 'Pweights'
      , data = cbind(Y = Y[inTrain], X[inTrain, ], Pweights = weight.cases)
      , reportProgress = 1 # the number of processed rows is printed and updated
      , verbose = 1)
)

## **** Computing starting values:
## 
Rows Processed: 209838 
## 
Rows Processed: 209838 
## **** Scoring iteration #3:
## 
## Deviance: 201058.0735
## 
Rows Processed: 209838 
## **** Scoring iteration #4:
## 
## Deviance: 199961.8653
## 
Rows Processed: 209838 
## **** Scoring iteration #5:
## 
## Deviance: 199949.6311
## 
Rows Processed: 209838 
## **** Scoring iteration #6:
## 
## Deviance: 199949.6240
## 
Rows Processed: 209838 
## 
## 
## Logistic Regression Results for: Y~Employee_code_ID_fct+Current_pincode_ID_fct+supplier_id_fct+branch_id_fct+ltv_fct+PERFORM_CNS_SCORE_fct+disbursed_amount_fct+OutstandingNow_fct+PERFORM_CNS_SCORE_DESCRIPTION_fct+State_ID_fct+DisAsDiff_fct+PRI_DISBURSED_AMOUNT_fct+PRI_OVERDUE_ACCTS_fct+manufacturer_id_fct+VoterID_flag_fct+ShareOverdue_fct+PRI_ACTIVE_ACCTS_fct+Day_fct+NO_OF_INQUIRIES_fct+DELINQUENT_ACCTS_IN_LAST_SIX_M_fct+Qrt_fct+YearsOnLoan_fct+Employment_Type_fct+PRIMARY_INSTAL_AMT_fct+asset_cost_fct+SEC_OverdueShareSec_fct+SEC_CURRENT_BALANCE_fct+Passport_flag_fct+Driving_flag_fct+SEC_INSTAL_AMT_fct+PAN_flag_fct+SEC_OVERDUE_ACCTS_fct
## ************************************************************************************************************************
## Dependent Variable: Y
## Total independent variables: 163 (Including number dropped: 33)
## Number of valid observations: 209838
## -2*LogLikelihood: 199950 (Residual Deviance on 209708 degrees of freedom)
## Row        Coeffs.                                                                                 Value      Std. Error         t Value        Pr(>|t|)
## [   1,.]   (Intercept)                                                                           0.7391          0.2000          3.6958          0.0002
## [   2,.]   Employee_code_ID_fct=1: Employee_code_ID <= 134                                       0.7657          0.0427         17.9335          0.0000
## [   3,.]   Employee_code_ID_fct=10: Employee_code_ID <= 242                                     -0.0770          0.0303         -2.5387          0.0111
## [   4,.]   Employee_code_ID_fct=11: Employee_code_ID <= 254                                     -0.1015          0.0296         -3.4293          0.0006
## [   5,.]   Employee_code_ID_fct=12: Employee_code_ID <= 270                                     -0.1436          0.0297         -4.8288          0.0000
## [   6,.]   Employee_code_ID_fct=13: Employee_code_ID <= 289                                     -0.2340          0.0293         -7.9857          0.0000
## [   7,.]   Employee_code_ID_fct=14: Employee_code_ID <= 319                                     -0.2577          0.0286         -9.0069          0.0000
## [   8,.]   Employee_code_ID_fct=15: Employee_code_ID > 319                                      -0.4823          0.0320        -15.0504          0.0000
## [   9,.]   Employee_code_ID_fct=2: Employee_code_ID <= 153                                       0.5028          0.0339         14.8256          0.0000
## [  10,.]   Employee_code_ID_fct=3: Employee_code_ID <= 170                                       0.4005          0.0298         13.4511          0.0000
## [  11,.]   Employee_code_ID_fct=4: Employee_code_ID <= 179                                       0.2856          0.0337          8.4796          0.0000
## [  12,.]   Employee_code_ID_fct=5: Employee_code_ID <= 188                                       0.2518          0.0318          7.9085          0.0000
## [  13,.]   Employee_code_ID_fct=6: Employee_code_ID <= 199                                       0.1600          0.0297          5.3933          0.0000
## [  14,.]   Employee_code_ID_fct=7: Employee_code_ID <= 211                                       0.0937          0.0291          3.2206          0.0013
## [  15,.]   Employee_code_ID_fct=8: Employee_code_ID <= 221                                       0.0284          0.0298          0.9526          0.3408
## [  16,.]   Employee_code_ID_fct=9: Employee_code_ID <= 233                                      Dropped         Dropped         Dropped          0.0000
## [  17,.]   Current_pincode_ID_fct=1: Current_pincode_ID <= 143                                   0.8572          0.0388         22.0715          0.0000
## [  18,.]   Current_pincode_ID_fct=10: Current_pincode_ID <= 257                                 -0.1809          0.0245         -7.3807          0.0000
## [  19,.]   Current_pincode_ID_fct=11: Current_pincode_ID <= 291                                 -0.2728          0.0236        -11.5684          0.0000
## [  20,.]   Current_pincode_ID_fct=12: Current_pincode_ID > 291                                  -0.3917          0.0259        -15.1180          0.0000
## [  21,.]   Current_pincode_ID_fct=2: Current_pincode_ID <= 158                                   0.7512          0.0370         20.3062          0.0000
## [  22,.]   Current_pincode_ID_fct=3: Current_pincode_ID <= 174                                   0.5712          0.0278         20.5606          0.0000
## [  23,.]   Current_pincode_ID_fct=4: Current_pincode_ID <= 188                                   0.4876          0.0269         18.1216          0.0000
## [  24,.]   Current_pincode_ID_fct=5: Current_pincode_ID <= 201                                   0.3985          0.0244         16.3081          0.0000
## [  25,.]   Current_pincode_ID_fct=6: Current_pincode_ID <= 212                                   0.3153          0.0253         12.4573          0.0000
## [  26,.]   Current_pincode_ID_fct=7: Current_pincode_ID <= 219                                   0.1621          0.0295          5.4903          0.0000
## [  27,.]   Current_pincode_ID_fct=8: Current_pincode_ID <= 225                                   0.1075          0.0296          3.6285          0.0003
## [  28,.]   Current_pincode_ID_fct=9: Current_pincode_ID <= 238                                  Dropped         Dropped         Dropped          0.0000
## [  29,.]   supplier_id_fct=1: supplier_id <= 133                                                 0.3345          0.0425          7.8740          0.0000
## [  30,.]   supplier_id_fct=10: supplier_id <= 253                                               -0.0792          0.0270         -2.9305          0.0034
## [  31,.]   supplier_id_fct=11: supplier_id <= 275                                               -0.0973          0.0248         -3.9244          0.0001
## [  32,.]   supplier_id_fct=12: supplier_id <= 314                                               -0.1325          0.0259         -5.1236          0.0000
## [  33,.]   supplier_id_fct=13: supplier_id > 314                                                -0.2318          0.0306         -7.5686          0.0000
## [  34,.]   supplier_id_fct=2: supplier_id <= 149                                                 0.3415          0.0373          9.1609          0.0000
## [  35,.]   supplier_id_fct=3: supplier_id <= 165                                                 0.2308          0.0311          7.4297          0.0000
## [  36,.]   supplier_id_fct=4: supplier_id <= 178                                                 0.1855          0.0288          6.4352          0.0000
## [  37,.]   supplier_id_fct=5: supplier_id <= 196                                                 0.1510          0.0256          5.9035          0.0000
## [  38,.]   supplier_id_fct=6: supplier_id <= 206                                                 0.1296          0.0287          4.5106          0.0000
## [  39,.]   supplier_id_fct=7: supplier_id <= 214                                                 0.1039          0.0304          3.4171          0.0006
## [  40,.]   supplier_id_fct=8: supplier_id <= 225                                                 0.0651          0.0260          2.5088          0.0121
## [  41,.]   supplier_id_fct=9: supplier_id <= 240                                                Dropped         Dropped         Dropped          0.0000
## [  42,.]   branch_id_fct=1: branch_id <= 153                                                    -0.4503          0.0410        -10.9721          0.0000
## [  43,.]   branch_id_fct=10: branch_id <= 284                                                    0.0550          0.0304          1.8089          0.0705
## [  44,.]   branch_id_fct=11: branch_id > 284                                                     0.0527          0.0414          1.2718          0.2034
## [  45,.]   branch_id_fct=2: branch_id <= 174                                                    -0.4923          0.0384        -12.8176          0.0000
## [  46,.]   branch_id_fct=3: branch_id <= 184                                                    -0.3828          0.0317        -12.0901          0.0000
## [  47,.]   branch_id_fct=4: branch_id <= 198                                                    -0.3148          0.0281        -11.2160          0.0000
## [  48,.]   branch_id_fct=5: branch_id <= 214                                                    -0.2090          0.0319         -6.5462          0.0000
## [  49,.]   branch_id_fct=6: branch_id <= 222                                                    -0.1907          0.0332         -5.7400          0.0000
## [  50,.]   branch_id_fct=7: branch_id <= 233                                                    -0.0964          0.0316         -3.0471          0.0023
## [  51,.]   branch_id_fct=8: branch_id <= 261                                                    -0.2644          0.0302         -8.7606          0.0000
## [  52,.]   branch_id_fct=9: branch_id <= 276                                                    Dropped         Dropped         Dropped          0.0000
## [  53,.]   ltv_fct=1: ltv <= 55.63                                                               0.6741          0.0483         13.9558          0.0000
## [  54,.]   ltv_fct=10: ltv <= 84.57                                                             -0.1191          0.0296         -4.0282          0.0001
## [  55,.]   ltv_fct=11: ltv <= 85                                                                -0.2560          0.0299         -8.5656          0.0000
## [  56,.]   ltv_fct=12: ltv <= 87.8                                                              -0.1109          0.0328         -3.3790          0.0007
## [  57,.]   ltv_fct=13: ltv <= 89.3                                                              -0.2617          0.0335         -7.8184          0.0000
## [  58,.]   ltv_fct=14: ltv > 89.3                                                               -0.3100          0.0333         -9.3079          0.0000
## [  59,.]   ltv_fct=2: ltv <= 62.22                                                               0.5654          0.0424         13.3198          0.0000
## [  60,.]   ltv_fct=3: ltv <= 68.34                                                               0.3978          0.0373         10.6611          0.0000
## [  61,.]   ltv_fct=4: ltv <= 72.9301                                                             0.2696          0.0334          8.0815          0.0000
## [  62,.]   ltv_fct=5: ltv <= 74.31                                                               0.1395          0.0346          4.0289          0.0001
## [  63,.]   ltv_fct=6: ltv <= 75                                                                  0.0904          0.0336          2.6889          0.0072
## [  64,.]   ltv_fct=7: ltv <= 77.39                                                               0.1953          0.0306          6.3806          0.0000
## [  65,.]   ltv_fct=8: ltv <= 78.92                                                               0.0634          0.0287          2.2089          0.0272
## [  66,.]   ltv_fct=9: ltv <= 83.34                                                              Dropped         Dropped         Dropped          0.0000
## [  67,.]   PERFORM_CNS_SCORE_fct=1: PERFORM_CNS_SCORE <= 0                                      -0.1592          0.0722         -2.2036          0.0276
## [  68,.]   PERFORM_CNS_SCORE_fct=2: PERFORM_CNS_SCORE <= 18                                     -0.0610          0.0589         -1.0345          0.3009
## [  69,.]   PERFORM_CNS_SCORE_fct=3: PERFORM_CNS_SCORE <= 441                                    -0.2494          0.0714         -3.4913          0.0005
## [  70,.]   PERFORM_CNS_SCORE_fct=4: PERFORM_CNS_SCORE <= 643                                    -0.1389          0.0618         -2.2491          0.0245
## [  71,.]   PERFORM_CNS_SCORE_fct=5: PERFORM_CNS_SCORE <= 738                                    -0.0788          0.0409         -1.9274          0.0539
## [  72,.]   PERFORM_CNS_SCORE_fct=6: PERFORM_CNS_SCORE <= 824                                     0.1750          0.0449          3.8987          0.0001
## [  73,.]   PERFORM_CNS_SCORE_fct=7: PERFORM_CNS_SCORE > 824                                     Dropped         Dropped         Dropped          0.0000
## [  74,.]   disbursed_amount_fct=1: disbursed_amount <= 39134                                     0.0393          0.0477          0.8234          0.4103
## [  75,.]   disbursed_amount_fct=2: disbursed_amount <= 43615                                     0.0841          0.0422          1.9918          0.0464
## [  76,.]   disbursed_amount_fct=3: disbursed_amount <= 48555                                     0.0864          0.0309          2.7959          0.0052
## [  77,.]   disbursed_amount_fct=4: disbursed_amount <= 51908                                     0.0959          0.0248          3.8738          0.0001
## [  78,.]   disbursed_amount_fct=5: disbursed_amount <= 55400                                     0.0522          0.0191          2.7370          0.0062
## [  79,.]   disbursed_amount_fct=6: disbursed_amount > 55400                                     Dropped         Dropped         Dropped          0.0000
## [  80,.]   OutstandingNow_fct=1: OutstandingNow <= 44402                                        -0.0596          0.0582         -1.0234          0.3061
## [  81,.]   OutstandingNow_fct=2: OutstandingNow <= 50314                                        -0.1223          0.0545         -2.2443          0.0248
## [  82,.]   OutstandingNow_fct=3: OutstandingNow <= 171384                                       -0.1834          0.0485         -3.7823          0.0002
## [  83,.]   OutstandingNow_fct=4: OutstandingNow <= 324324                                       -0.0857          0.0429         -1.9988          0.0456
## [  84,.]   OutstandingNow_fct=5: OutstandingNow <= 746271                                       -0.1138          0.0390         -2.9159          0.0035
## [  85,.]   OutstandingNow_fct=6: OutstandingNow > 746271                                        Dropped         Dropped         Dropped          0.0000
## [  86,.]   PERFORM_CNS_SCORE_DESCRIPTION_fct=1: PERFORM_CNS_SCORE_DESCRIPTION <= 150             0.3537          0.0632          5.5963          0.0000
## [  87,.]   PERFORM_CNS_SCORE_DESCRIPTION_fct=2: PERFORM_CNS_SCORE_DESCRIPTION <= 172             0.3311          0.0616          5.3773          0.0000
## [  88,.]   PERFORM_CNS_SCORE_DESCRIPTION_fct=3: PERFORM_CNS_SCORE_DESCRIPTION <= 205             0.2451          0.0500          4.9071          0.0000
## [  89,.]   PERFORM_CNS_SCORE_DESCRIPTION_fct=4: PERFORM_CNS_SCORE_DESCRIPTION <= 231             0.2198          0.0622          3.5352          0.0004
## [  90,.]   PERFORM_CNS_SCORE_DESCRIPTION_fct=5: PERFORM_CNS_SCORE_DESCRIPTION <= 256             0.0049          0.0358          0.1370          0.8910
## [  91,.]   PERFORM_CNS_SCORE_DESCRIPTION_fct=6: PERFORM_CNS_SCORE_DESCRIPTION > 256             Dropped         Dropped         Dropped          0.0000
## [  92,.]   State_ID_fct=1: State_ID <= 183                                                      -0.0597          0.0494         -1.2078          0.2271
## [  93,.]   State_ID_fct=2: State_ID <= 188                                                      -0.0293          0.0464         -0.6329          0.5268
## [  94,.]   State_ID_fct=3: State_ID <= 206                                                      -0.0022          0.0432         -0.0499          0.9602
## [  95,.]   State_ID_fct=4: State_ID <= 214                                                       0.0796          0.0425          1.8719          0.0612
## [  96,.]   State_ID_fct=5: State_ID <= 220                                                       0.1009          0.0497          2.0304          0.0423
## [  97,.]   State_ID_fct=6: State_ID <= 229                                                       0.0639          0.0512          1.2496          0.2115
## [  98,.]   State_ID_fct=7: State_ID <= 272                                                       0.0847          0.0445          1.9050          0.0568
## [  99,.]   State_ID_fct=8: State_ID > 272                                                       Dropped         Dropped         Dropped          0.0000
## [ 100,.]   DisAsDiff_fct=1: DisAsDiff <= 13554                                                  -0.1302          0.0396         -3.2854          0.0010
## [ 101,.]   DisAsDiff_fct=2: DisAsDiff <= 15670                                                  -0.0822          0.0342         -2.4020          0.0163
## [ 102,.]   DisAsDiff_fct=3: DisAsDiff <= 16661                                                   0.0062          0.0359          0.1722          0.8633
## [ 103,.]   DisAsDiff_fct=4: DisAsDiff <= 19822                                                  -0.0175          0.0254         -0.6869          0.4921
## [ 104,.]   DisAsDiff_fct=5: DisAsDiff > 19822                                                   Dropped         Dropped         Dropped          0.0000
## [ 105,.]   PRI_DISBURSED_AMOUNT_fct=1: PRI_DISBURSED_AMOUNT <= 218581                           -0.2560          0.0387         -6.6095          0.0000
## [ 106,.]   PRI_DISBURSED_AMOUNT_fct=2: PRI_DISBURSED_AMOUNT > 218581                            Dropped         Dropped         Dropped          0.0000
## [ 107,.]   PRI_OVERDUE_ACCTS_fct=1: PRI_OVERDUE_ACCTS <= 0                                       0.1444          0.0287          5.0372          0.0000
## [ 108,.]   PRI_OVERDUE_ACCTS_fct=2: PRI_OVERDUE_ACCTS > 0                                       Dropped         Dropped         Dropped          0.0000
## [ 109,.]   manufacturer_id_fct=1: manufacturer_id <= 210                                         0.1232          0.0239          5.1589          0.0000
## [ 110,.]   manufacturer_id_fct=2: manufacturer_id <= 221                                        -0.0035          0.0278         -0.1266          0.8993
## [ 111,.]   manufacturer_id_fct=3: manufacturer_id <= 228                                         0.0977          0.0261          3.7512          0.0002
## [ 112,.]   manufacturer_id_fct=4: manufacturer_id > 228                                         Dropped         Dropped         Dropped          0.0000
## [ 113,.]   VoterID_flag_fct=1: VoterID_flag = 0                                                  0.0935          0.0195          4.8004          0.0000
## [ 114,.]   VoterID_flag_fct=2: VoterID_flag = 1                                                 Dropped         Dropped         Dropped          0.0000
## [ 115,.]   ShareOverdue_fct=1: ShareOverdue <= -2                                                0.0923          0.0318          2.8997          0.0037
## [ 116,.]   ShareOverdue_fct=2: ShareOverdue <= -1                                                0.1288          0.0233          5.5417          0.0000
## [ 117,.]   ShareOverdue_fct=3: ShareOverdue > -1                                                Dropped         Dropped         Dropped          0.0000
## [ 118,.]   PRI_ACTIVE_ACCTS_fct=1: PRI_ACTIVE_ACCTS <= 0                                        -0.2271          0.0443         -5.1314          0.0000
## [ 119,.]   PRI_ACTIVE_ACCTS_fct=2: PRI_ACTIVE_ACCTS <= 1                                        -0.2429          0.0334         -7.2669          0.0000
## [ 120,.]   PRI_ACTIVE_ACCTS_fct=3: PRI_ACTIVE_ACCTS <= 3                                        -0.2096          0.0288         -7.2732          0.0000
## [ 121,.]   PRI_ACTIVE_ACCTS_fct=4: PRI_ACTIVE_ACCTS > 3                                         Dropped         Dropped         Dropped          0.0000
## [ 122,.]   Day_fct=1: Day <= 28                                                                  0.2494          0.0213         11.7055          0.0000
## [ 123,.]   Day_fct=2: Day <= 30                                                                  0.1283          0.0260          4.9301          0.0000
## [ 124,.]   Day_fct=3: Day > 30                                                                  Dropped         Dropped         Dropped          0.0000
## [ 125,.]   NO_OF_INQUIRIES_fct=1: NO_OF_INQUIRIES <= 0                                           0.2732          0.0170         16.0551          0.0000
## [ 126,.]   NO_OF_INQUIRIES_fct=2: NO_OF_INQUIRIES > 0                                           Dropped         Dropped         Dropped          0.0000
## [ 127,.]   DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=1: DELINQUENT_ACCTS_IN_LAST_SIX_M <= 0             0.2820          0.0244         11.5364          0.0000
## [ 128,.]   DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=2: DELINQUENT_ACCTS_IN_LAST_SIX_M > 0             Dropped         Dropped         Dropped          0.0000
## [ 129,.]   Qrt_fct=1: Qrt = 3                                                                    0.2191          0.0118         18.5775          0.0000
## [ 130,.]   Qrt_fct=2: Qrt = 4                                                                   Dropped         Dropped         Dropped          0.0000
## [ 131,.]   YearsOnLoan_fct=1: YearsOnLoan <= 22.8918                                            -0.3841          0.0306        -12.5451          0.0000
## [ 132,.]   YearsOnLoan_fct=2: YearsOnLoan <= 28.8496                                            -0.2784          0.0266        -10.4683          0.0000
## [ 133,.]   YearsOnLoan_fct=3: YearsOnLoan <= 38.8321                                            -0.1799          0.0259         -6.9531          0.0000
## [ 134,.]   YearsOnLoan_fct=4: YearsOnLoan <= 51.8208                                            -0.0930          0.0265         -3.5125          0.0004
## [ 135,.]   YearsOnLoan_fct=5: YearsOnLoan > 51.8208                                             Dropped         Dropped         Dropped          0.0000
## [ 136,.]   Employment_Type_fct=1: Employment_Type = 203                                          0.1524          0.0123         12.4431          0.0000
## [ 137,.]   Employment_Type_fct=2: Employment_Type = 215                                          0.2166          0.0347          6.2474          0.0000
## [ 138,.]   Employment_Type_fct=3: Employment_Type = 227                                         Dropped         Dropped         Dropped          0.0000
## [ 139,.]   PRIMARY_INSTAL_AMT_fct=1: PRIMARY_INSTAL_AMT <= 1564                                 -0.0180          0.0325         -0.5519          0.5810
## [ 140,.]   PRIMARY_INSTAL_AMT_fct=2: PRIMARY_INSTAL_AMT <= 2832                                 -0.2560          0.0375         -6.8204          0.0000
## [ 141,.]   PRIMARY_INSTAL_AMT_fct=3: PRIMARY_INSTAL_AMT <= 5033                                 -0.1442          0.0387         -3.7226          0.0002
## [ 142,.]   PRIMARY_INSTAL_AMT_fct=4: PRIMARY_INSTAL_AMT <= 25326                                -0.1437          0.0325         -4.4164          0.0000
## [ 143,.]   PRIMARY_INSTAL_AMT_fct=5: PRIMARY_INSTAL_AMT > 25326                                 Dropped         Dropped         Dropped          0.0000
## [ 144,.]   asset_cost_fct=1: asset_cost <= 60098                                                 0.0726          0.0457          1.5873          0.1124
## [ 145,.]   asset_cost_fct=2: asset_cost <= 70561                                                 0.1494          0.0296          5.0467          0.0000
## [ 146,.]   asset_cost_fct=3: asset_cost <= 85738                                                 0.1776          0.0224          7.9199          0.0000
## [ 147,.]   asset_cost_fct=4: asset_cost > 85738                                                 Dropped         Dropped         Dropped          0.0000
## [ 148,.]   SEC_OverdueShareSec_fct=1: SEC_OverdueShareSec <= 0                                  -0.0599          0.0932         -0.6423          0.5206
## [ 149,.]   SEC_OverdueShareSec_fct=11: SEC_OverdueShareSec Is Null                              -0.0752          0.1024         -0.7339          0.4630
## [ 150,.]   SEC_OverdueShareSec_fct=8: SEC_OverdueShareSec <= 0.2                                 0.2864          0.2454          1.1672          0.2431
## [ 151,.]   SEC_OverdueShareSec_fct=9: SEC_OverdueShareSec <= 1                                  Dropped         Dropped         Dropped          0.0000
## [ 152,.]   SEC_CURRENT_BALANCE_fct=1: SEC_CURRENT_BALANCE <= 0                                  -0.0239          0.0779         -0.3071          0.7588
## [ 153,.]   SEC_CURRENT_BALANCE_fct=10: SEC_CURRENT_BALANCE > 0                                  Dropped         Dropped         Dropped          0.0000
## [ 154,.]   Passport_flag_fct=1: Passport_flag = 0                                               -0.1880          0.1371         -1.3710          0.1704
## [ 155,.]   Passport_flag_fct=2: Passport_flag = 1                                               Dropped         Dropped         Dropped          0.0000
## [ 156,.]   Driving_flag_fct=1: Driving_flag = 0                                                  0.0159          0.0384          0.4145          0.6785
## [ 157,.]   Driving_flag_fct=2: Driving_flag = 1                                                 Dropped         Dropped         Dropped          0.0000
## [ 158,.]   SEC_INSTAL_AMT_fct=1: SEC_INSTAL_AMT <= 0                                             0.0609          0.0778          0.7835          0.4334
## [ 159,.]   SEC_INSTAL_AMT_fct=10: SEC_INSTAL_AMT > 0                                            Dropped         Dropped         Dropped          0.0000
## [ 160,.]   PAN_flag_fct=1: PAN_flag = 0                                                         -0.0460          0.0230         -1.9976          0.0458
## [ 161,.]   PAN_flag_fct=2: PAN_flag = 1                                                         Dropped         Dropped         Dropped          0.0000
## [ 162,.]   SEC_OVERDUE_ACCTS_fct=1: SEC_OVERDUE_ACCTS <= 0                                      Dropped         Dropped         Dropped          0.0000
## [ 163,.]   SEC_OVERDUE_ACCTS_fct=10: SEC_OVERDUE_ACCTS > 0                                      Dropped         Dropped         Dropped          0.0000
## Condition number of final VC matrix: 1997.7639
## ************************************************************************************************************************
## 

##    user  system elapsed 
##    0.14    0.01    3.95

summary(rxLogitFit)

## Call:
## RevoScaleR::rxLogit(formula = paste("Y ~ ", paste(names(X), collapse = "+")), 
##     data = cbind(Y = Y[inTrain], X[inTrain, ], Pweights = weight.cases), 
##     pweights = "Pweights", reportProgress = 1, verbose = 1)
## 
## Logistic Regression Results for: Y ~
##     Employee_code_ID_fct+Current_pincode_ID_fct+supplier_id_fct+branch_id_fct+ltv_fct+PERFORM_CNS_SCORE_fct+disbursed_amount_fct+OutstandingNow_fct+PERFORM_CNS_SCORE_DESCRIPTION_fct+State_ID_fct+DisAsDiff_fct+PRI_DISBURSED_AMOUNT_fct+PRI_OVERDUE_ACCTS_fct+manufacturer_id_fct+VoterID_flag_fct+ShareOverdue_fct+PRI_ACTIVE_ACCTS_fct+Day_fct+NO_OF_INQUIRIES_fct+DELINQUENT_ACCTS_IN_LAST_SIX_M_fct+Qrt_fct+YearsOnLoan_fct+Employment_Type_fct+PRIMARY_INSTAL_AMT_fct+asset_cost_fct+SEC_OverdueShareSec_fct+SEC_CURRENT_BALANCE_fct+Passport_flag_fct+Driving_flag_fct+SEC_INSTAL_AMT_fct+PAN_flag_fct+SEC_OVERDUE_ACCTS_fct
## Data: cbind(Y = Y[inTrain], X[inTrain, ], Pweights = weight.cases)
## Dependent variable(s): Y
## Total independent variables: 163 (Including number dropped: 33)
## Number of valid observations: 209838
## Number of missing observations: 0 
## -2*LogLikelihood: 199949.624 (Residual deviance on 209708 degrees of freedom)
##  
## Coefficients:
##                                                                            Estimate
## (Intercept)                                                                0.739128
## Employee_code_ID_fct=1: Employee_code_ID <= 134                            0.765665
## Employee_code_ID_fct=10: Employee_code_ID <= 242                          -0.077044
## Employee_code_ID_fct=11: Employee_code_ID <= 254                          -0.101471
## Employee_code_ID_fct=12: Employee_code_ID <= 270                          -0.143575
## Employee_code_ID_fct=13: Employee_code_ID <= 289                          -0.234039
## Employee_code_ID_fct=14: Employee_code_ID <= 319                          -0.257740
## Employee_code_ID_fct=15: Employee_code_ID > 319                           -0.482348
## Employee_code_ID_fct=2: Employee_code_ID <= 153                            0.502839
## Employee_code_ID_fct=3: Employee_code_ID <= 170                            0.400538
## Employee_code_ID_fct=4: Employee_code_ID <= 179                            0.285620
## Employee_code_ID_fct=5: Employee_code_ID <= 188                            0.251824
## Employee_code_ID_fct=6: Employee_code_ID <= 199                            0.160031
## Employee_code_ID_fct=7: Employee_code_ID <= 211                            0.093653
## Employee_code_ID_fct=8: Employee_code_ID <= 221                            0.028377
## Employee_code_ID_fct=9: Employee_code_ID <= 233                             Dropped
## Current_pincode_ID_fct=1: Current_pincode_ID <= 143                        0.857166
## Current_pincode_ID_fct=10: Current_pincode_ID <= 257                      -0.180906
## Current_pincode_ID_fct=11: Current_pincode_ID <= 291                      -0.272803
## Current_pincode_ID_fct=12: Current_pincode_ID > 291                       -0.391739
## Current_pincode_ID_fct=2: Current_pincode_ID <= 158                        0.751176
## Current_pincode_ID_fct=3: Current_pincode_ID <= 174                        0.571246
## Current_pincode_ID_fct=4: Current_pincode_ID <= 188                        0.487634
## Current_pincode_ID_fct=5: Current_pincode_ID <= 201                        0.398476
## Current_pincode_ID_fct=6: Current_pincode_ID <= 212                        0.315290
## Current_pincode_ID_fct=7: Current_pincode_ID <= 219                        0.162143
## Current_pincode_ID_fct=8: Current_pincode_ID <= 225                        0.107461
## Current_pincode_ID_fct=9: Current_pincode_ID <= 238                         Dropped
## supplier_id_fct=1: supplier_id <= 133                                      0.334523
## supplier_id_fct=10: supplier_id <= 253                                    -0.079193
## supplier_id_fct=11: supplier_id <= 275                                    -0.097272
## supplier_id_fct=12: supplier_id <= 314                                    -0.132506
## supplier_id_fct=13: supplier_id > 314                                     -0.231812
## supplier_id_fct=2: supplier_id <= 149                                      0.341500
## supplier_id_fct=3: supplier_id <= 165                                      0.230836
## supplier_id_fct=4: supplier_id <= 178                                      0.185501
## supplier_id_fct=5: supplier_id <= 196                                      0.150963
## supplier_id_fct=6: supplier_id <= 206                                      0.129583
## supplier_id_fct=7: supplier_id <= 214                                      0.103929
## supplier_id_fct=8: supplier_id <= 225                                      0.065137
## supplier_id_fct=9: supplier_id <= 240                                       Dropped
## branch_id_fct=1: branch_id <= 153                                         -0.450331
## branch_id_fct=10: branch_id <= 284                                         0.055049
## branch_id_fct=11: branch_id > 284                                          0.052659
## branch_id_fct=2: branch_id <= 174                                         -0.492323
## branch_id_fct=3: branch_id <= 184                                         -0.382755
## branch_id_fct=4: branch_id <= 198                                         -0.314757
## branch_id_fct=5: branch_id <= 214                                         -0.208958
## branch_id_fct=6: branch_id <= 222                                         -0.190663
## branch_id_fct=7: branch_id <= 233                                         -0.096355
## branch_id_fct=8: branch_id <= 261                                         -0.264450
## branch_id_fct=9: branch_id <= 276                                           Dropped
## ltv_fct=1: ltv <= 55.63                                                    0.674092
## ltv_fct=10: ltv <= 84.57                                                  -0.119055
## ltv_fct=11: ltv <= 85                                                     -0.256004
## ltv_fct=12: ltv <= 87.8                                                   -0.110868
## ltv_fct=13: ltv <= 89.3                                                   -0.261671
## ltv_fct=14: ltv > 89.3                                                    -0.310022
## ltv_fct=2: ltv <= 62.22                                                    0.565406
## ltv_fct=3: ltv <= 68.34                                                    0.397794
## ltv_fct=4: ltv <= 72.9301                                                  0.269590
## ltv_fct=5: ltv <= 74.31                                                    0.139500
## ltv_fct=6: ltv <= 75                                                       0.090372
## ltv_fct=7: ltv <= 77.39                                                    0.195296
## ltv_fct=8: ltv <= 78.92                                                    0.063365
## ltv_fct=9: ltv <= 83.34                                                     Dropped
## PERFORM_CNS_SCORE_fct=1: PERFORM_CNS_SCORE <= 0                           -0.159161
## PERFORM_CNS_SCORE_fct=2: PERFORM_CNS_SCORE <= 18                          -0.060971
## PERFORM_CNS_SCORE_fct=3: PERFORM_CNS_SCORE <= 441                         -0.249354
## PERFORM_CNS_SCORE_fct=4: PERFORM_CNS_SCORE <= 643                         -0.138944
## PERFORM_CNS_SCORE_fct=5: PERFORM_CNS_SCORE <= 738                         -0.078764
## PERFORM_CNS_SCORE_fct=6: PERFORM_CNS_SCORE <= 824                          0.175027
## PERFORM_CNS_SCORE_fct=7: PERFORM_CNS_SCORE > 824                            Dropped
## disbursed_amount_fct=1: disbursed_amount <= 39134                          0.039263
## disbursed_amount_fct=2: disbursed_amount <= 43615                          0.084088
## disbursed_amount_fct=3: disbursed_amount <= 48555                          0.086379
## disbursed_amount_fct=4: disbursed_amount <= 51908                          0.095899
## disbursed_amount_fct=5: disbursed_amount <= 55400                          0.052241
## disbursed_amount_fct=6: disbursed_amount > 55400                            Dropped
## OutstandingNow_fct=1: OutstandingNow <= 44402                             -0.059569
## OutstandingNow_fct=2: OutstandingNow <= 50314                             -0.122291
## OutstandingNow_fct=3: OutstandingNow <= 171384                            -0.183387
## OutstandingNow_fct=4: OutstandingNow <= 324324                            -0.085655
## OutstandingNow_fct=5: OutstandingNow <= 746271                            -0.113832
## OutstandingNow_fct=6: OutstandingNow > 746271                               Dropped
## PERFORM_CNS_SCORE_DESCRIPTION_fct=1: PERFORM_CNS_SCORE_DESCRIPTION <= 150  0.353738
## PERFORM_CNS_SCORE_DESCRIPTION_fct=2: PERFORM_CNS_SCORE_DESCRIPTION <= 172  0.331067
## PERFORM_CNS_SCORE_DESCRIPTION_fct=3: PERFORM_CNS_SCORE_DESCRIPTION <= 205  0.245147
## PERFORM_CNS_SCORE_DESCRIPTION_fct=4: PERFORM_CNS_SCORE_DESCRIPTION <= 231  0.219787
## PERFORM_CNS_SCORE_DESCRIPTION_fct=5: PERFORM_CNS_SCORE_DESCRIPTION <= 256  0.004902
## PERFORM_CNS_SCORE_DESCRIPTION_fct=6: PERFORM_CNS_SCORE_DESCRIPTION > 256    Dropped
## State_ID_fct=1: State_ID <= 183                                           -0.059697
## State_ID_fct=2: State_ID <= 188                                           -0.029348
## State_ID_fct=3: State_ID <= 206                                           -0.002158
## State_ID_fct=4: State_ID <= 214                                            0.079577
## State_ID_fct=5: State_ID <= 220                                            0.100859
## State_ID_fct=6: State_ID <= 229                                            0.063927
## State_ID_fct=7: State_ID <= 272                                            0.084702
## State_ID_fct=8: State_ID > 272                                              Dropped
## DisAsDiff_fct=1: DisAsDiff <= 13554                                       -0.130242
## DisAsDiff_fct=2: DisAsDiff <= 15670                                       -0.082229
## DisAsDiff_fct=3: DisAsDiff <= 16661                                        0.006175
## DisAsDiff_fct=4: DisAsDiff <= 19822                                       -0.017471
## DisAsDiff_fct=5: DisAsDiff > 19822                                          Dropped
## PRI_DISBURSED_AMOUNT_fct=1: PRI_DISBURSED_AMOUNT <= 218581                -0.255990
## PRI_DISBURSED_AMOUNT_fct=2: PRI_DISBURSED_AMOUNT > 218581                   Dropped
## PRI_OVERDUE_ACCTS_fct=1: PRI_OVERDUE_ACCTS <= 0                            0.144360
## PRI_OVERDUE_ACCTS_fct=2: PRI_OVERDUE_ACCTS > 0                              Dropped
## manufacturer_id_fct=1: manufacturer_id <= 210                              0.123239
## manufacturer_id_fct=2: manufacturer_id <= 221                             -0.003519
## manufacturer_id_fct=3: manufacturer_id <= 228                              0.097721
## manufacturer_id_fct=4: manufacturer_id > 228                                Dropped
## VoterID_flag_fct=1: VoterID_flag = 0                                       0.093489
## VoterID_flag_fct=2: VoterID_flag = 1                                        Dropped
## ShareOverdue_fct=1: ShareOverdue <= -2                                     0.092295
## ShareOverdue_fct=2: ShareOverdue <= -1                                     0.128849
## ShareOverdue_fct=3: ShareOverdue > -1                                       Dropped
## PRI_ACTIVE_ACCTS_fct=1: PRI_ACTIVE_ACCTS <= 0                             -0.227117
## PRI_ACTIVE_ACCTS_fct=2: PRI_ACTIVE_ACCTS <= 1                             -0.242919
## PRI_ACTIVE_ACCTS_fct=3: PRI_ACTIVE_ACCTS <= 3                             -0.209621
## PRI_ACTIVE_ACCTS_fct=4: PRI_ACTIVE_ACCTS > 3                                Dropped
## Day_fct=1: Day <= 28                                                       0.249400
## Day_fct=2: Day <= 30                                                       0.128288
## Day_fct=3: Day > 30                                                         Dropped
## NO_OF_INQUIRIES_fct=1: NO_OF_INQUIRIES <= 0                                0.273240
## NO_OF_INQUIRIES_fct=2: NO_OF_INQUIRIES > 0                                  Dropped
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=1: DELINQUENT_ACCTS_IN_LAST_SIX_M <= 0  0.281950
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=2: DELINQUENT_ACCTS_IN_LAST_SIX_M > 0    Dropped
## Qrt_fct=1: Qrt = 3                                                         0.219055
## Qrt_fct=2: Qrt = 4                                                          Dropped
## YearsOnLoan_fct=1: YearsOnLoan <= 22.8918                                 -0.384106
## YearsOnLoan_fct=2: YearsOnLoan <= 28.8496                                 -0.278357
## YearsOnLoan_fct=3: YearsOnLoan <= 38.8321                                 -0.179941
## YearsOnLoan_fct=4: YearsOnLoan <= 51.8208                                 -0.092981
## YearsOnLoan_fct=5: YearsOnLoan > 51.8208                                    Dropped
## Employment_Type_fct=1: Employment_Type = 203                               0.152449
## Employment_Type_fct=2: Employment_Type = 215                               0.216629
## Employment_Type_fct=3: Employment_Type = 227                                Dropped
## PRIMARY_INSTAL_AMT_fct=1: PRIMARY_INSTAL_AMT <= 1564                      -0.017958
## PRIMARY_INSTAL_AMT_fct=2: PRIMARY_INSTAL_AMT <= 2832                      -0.256020
## PRIMARY_INSTAL_AMT_fct=3: PRIMARY_INSTAL_AMT <= 5033                      -0.144240
## PRIMARY_INSTAL_AMT_fct=4: PRIMARY_INSTAL_AMT <= 25326                     -0.143675
## PRIMARY_INSTAL_AMT_fct=5: PRIMARY_INSTAL_AMT > 25326                        Dropped
## asset_cost_fct=1: asset_cost <= 60098                                      0.072598
## asset_cost_fct=2: asset_cost <= 70561                                      0.149356
## asset_cost_fct=3: asset_cost <= 85738                                      0.177618
## asset_cost_fct=4: asset_cost > 85738                                        Dropped
## SEC_OverdueShareSec_fct=1: SEC_OverdueShareSec <= 0                       -0.059896
## SEC_OverdueShareSec_fct=11: SEC_OverdueShareSec Is Null                   -0.075173
## SEC_OverdueShareSec_fct=8: SEC_OverdueShareSec <= 0.2                      0.286437
## SEC_OverdueShareSec_fct=9: SEC_OverdueShareSec <= 1                         Dropped
## SEC_CURRENT_BALANCE_fct=1: SEC_CURRENT_BALANCE <= 0                       -0.023913
## SEC_CURRENT_BALANCE_fct=10: SEC_CURRENT_BALANCE > 0                         Dropped
## Passport_flag_fct=1: Passport_flag = 0                                    -0.187963
## Passport_flag_fct=2: Passport_flag = 1                                      Dropped
## Driving_flag_fct=1: Driving_flag = 0                                       0.015928
## Driving_flag_fct=2: Driving_flag = 1                                        Dropped
## SEC_INSTAL_AMT_fct=1: SEC_INSTAL_AMT <= 0                                  0.060927
## SEC_INSTAL_AMT_fct=10: SEC_INSTAL_AMT > 0                                   Dropped
## PAN_flag_fct=1: PAN_flag = 0                                              -0.045962
## PAN_flag_fct=2: PAN_flag = 1                                                Dropped
## SEC_OVERDUE_ACCTS_fct=1: SEC_OVERDUE_ACCTS <= 0                             Dropped
## SEC_OVERDUE_ACCTS_fct=10: SEC_OVERDUE_ACCTS > 0                             Dropped
##                                                                           Std. Error
## (Intercept)                                                                 0.199992
## Employee_code_ID_fct=1: Employee_code_ID <= 134                             0.042695
## Employee_code_ID_fct=10: Employee_code_ID <= 242                            0.030348
## Employee_code_ID_fct=11: Employee_code_ID <= 254                            0.029589
## Employee_code_ID_fct=12: Employee_code_ID <= 270                            0.029733
## Employee_code_ID_fct=13: Employee_code_ID <= 289                            0.029307
## Employee_code_ID_fct=14: Employee_code_ID <= 319                            0.028616
## Employee_code_ID_fct=15: Employee_code_ID > 319                             0.032049
## Employee_code_ID_fct=2: Employee_code_ID <= 153                             0.033917
## Employee_code_ID_fct=3: Employee_code_ID <= 170                             0.029777
## Employee_code_ID_fct=4: Employee_code_ID <= 179                             0.033683
## Employee_code_ID_fct=5: Employee_code_ID <= 188                             0.031842
## Employee_code_ID_fct=6: Employee_code_ID <= 199                             0.029672
## Employee_code_ID_fct=7: Employee_code_ID <= 211                             0.029080
## Employee_code_ID_fct=8: Employee_code_ID <= 221                             0.029788
## Employee_code_ID_fct=9: Employee_code_ID <= 233                              Dropped
## Current_pincode_ID_fct=1: Current_pincode_ID <= 143                         0.038836
## Current_pincode_ID_fct=10: Current_pincode_ID <= 257                        0.024511
## Current_pincode_ID_fct=11: Current_pincode_ID <= 291                        0.023582
## Current_pincode_ID_fct=12: Current_pincode_ID > 291                         0.025912
## Current_pincode_ID_fct=2: Current_pincode_ID <= 158                         0.036992
## Current_pincode_ID_fct=3: Current_pincode_ID <= 174                         0.027784
## Current_pincode_ID_fct=4: Current_pincode_ID <= 188                         0.026909
## Current_pincode_ID_fct=5: Current_pincode_ID <= 201                         0.024434
## Current_pincode_ID_fct=6: Current_pincode_ID <= 212                         0.025310
## Current_pincode_ID_fct=7: Current_pincode_ID <= 219                         0.029533
## Current_pincode_ID_fct=8: Current_pincode_ID <= 225                         0.029616
## Current_pincode_ID_fct=9: Current_pincode_ID <= 238                          Dropped
## supplier_id_fct=1: supplier_id <= 133                                       0.042484
## supplier_id_fct=10: supplier_id <= 253                                      0.027024
## supplier_id_fct=11: supplier_id <= 275                                      0.024787
## supplier_id_fct=12: supplier_id <= 314                                      0.025862
## supplier_id_fct=13: supplier_id > 314                                       0.030628
## supplier_id_fct=2: supplier_id <= 149                                       0.037278
## supplier_id_fct=3: supplier_id <= 165                                       0.031069
## supplier_id_fct=4: supplier_id <= 178                                       0.028826
## supplier_id_fct=5: supplier_id <= 196                                       0.025572
## supplier_id_fct=6: supplier_id <= 206                                       0.028729
## supplier_id_fct=7: supplier_id <= 214                                       0.030414
## supplier_id_fct=8: supplier_id <= 225                                       0.025963
## supplier_id_fct=9: supplier_id <= 240                                        Dropped
## branch_id_fct=1: branch_id <= 153                                           0.041043
## branch_id_fct=10: branch_id <= 284                                          0.030432
## branch_id_fct=11: branch_id > 284                                           0.041404
## branch_id_fct=2: branch_id <= 174                                           0.038410
## branch_id_fct=3: branch_id <= 184                                           0.031658
## branch_id_fct=4: branch_id <= 198                                           0.028063
## branch_id_fct=5: branch_id <= 214                                           0.031921
## branch_id_fct=6: branch_id <= 222                                           0.033216
## branch_id_fct=7: branch_id <= 233                                           0.031622
## branch_id_fct=8: branch_id <= 261                                           0.030186
## branch_id_fct=9: branch_id <= 276                                            Dropped
## ltv_fct=1: ltv <= 55.63                                                     0.048302
## ltv_fct=10: ltv <= 84.57                                                    0.029555
## ltv_fct=11: ltv <= 85                                                       0.029887
## ltv_fct=12: ltv <= 87.8                                                     0.032811
## ltv_fct=13: ltv <= 89.3                                                     0.033469
## ltv_fct=14: ltv > 89.3                                                      0.033307
## ltv_fct=2: ltv <= 62.22                                                     0.042449
## ltv_fct=3: ltv <= 68.34                                                     0.037313
## ltv_fct=4: ltv <= 72.9301                                                   0.033359
## ltv_fct=5: ltv <= 74.31                                                     0.034625
## ltv_fct=6: ltv <= 75                                                        0.033609
## ltv_fct=7: ltv <= 77.39                                                     0.030608
## ltv_fct=8: ltv <= 78.92                                                     0.028686
## ltv_fct=9: ltv <= 83.34                                                      Dropped
## PERFORM_CNS_SCORE_fct=1: PERFORM_CNS_SCORE <= 0                             0.072227
## PERFORM_CNS_SCORE_fct=2: PERFORM_CNS_SCORE <= 18                            0.058937
## PERFORM_CNS_SCORE_fct=3: PERFORM_CNS_SCORE <= 441                           0.071422
## PERFORM_CNS_SCORE_fct=4: PERFORM_CNS_SCORE <= 643                           0.061776
## PERFORM_CNS_SCORE_fct=5: PERFORM_CNS_SCORE <= 738                           0.040865
## PERFORM_CNS_SCORE_fct=6: PERFORM_CNS_SCORE <= 824                           0.044894
## PERFORM_CNS_SCORE_fct=7: PERFORM_CNS_SCORE > 824                             Dropped
## disbursed_amount_fct=1: disbursed_amount <= 39134                           0.047683
## disbursed_amount_fct=2: disbursed_amount <= 43615                           0.042217
## disbursed_amount_fct=3: disbursed_amount <= 48555                           0.030895
## disbursed_amount_fct=4: disbursed_amount <= 51908                           0.024756
## disbursed_amount_fct=5: disbursed_amount <= 55400                           0.019087
## disbursed_amount_fct=6: disbursed_amount > 55400                             Dropped
## OutstandingNow_fct=1: OutstandingNow <= 44402                               0.058208
## OutstandingNow_fct=2: OutstandingNow <= 50314                               0.054488
## OutstandingNow_fct=3: OutstandingNow <= 171384                              0.048486
## OutstandingNow_fct=4: OutstandingNow <= 324324                              0.042853
## OutstandingNow_fct=5: OutstandingNow <= 746271                              0.039039
## OutstandingNow_fct=6: OutstandingNow > 746271                                Dropped
## PERFORM_CNS_SCORE_DESCRIPTION_fct=1: PERFORM_CNS_SCORE_DESCRIPTION <= 150   0.063209
## PERFORM_CNS_SCORE_DESCRIPTION_fct=2: PERFORM_CNS_SCORE_DESCRIPTION <= 172   0.061567
## PERFORM_CNS_SCORE_DESCRIPTION_fct=3: PERFORM_CNS_SCORE_DESCRIPTION <= 205   0.049958
## PERFORM_CNS_SCORE_DESCRIPTION_fct=4: PERFORM_CNS_SCORE_DESCRIPTION <= 231   0.062171
## PERFORM_CNS_SCORE_DESCRIPTION_fct=5: PERFORM_CNS_SCORE_DESCRIPTION <= 256   0.035776
## PERFORM_CNS_SCORE_DESCRIPTION_fct=6: PERFORM_CNS_SCORE_DESCRIPTION > 256     Dropped
## State_ID_fct=1: State_ID <= 183                                             0.049425
## State_ID_fct=2: State_ID <= 188                                             0.046370
## State_ID_fct=3: State_ID <= 206                                             0.043220
## State_ID_fct=4: State_ID <= 214                                             0.042510
## State_ID_fct=5: State_ID <= 220                                             0.049674
## State_ID_fct=6: State_ID <= 229                                             0.051160
## State_ID_fct=7: State_ID <= 272                                             0.044464
## State_ID_fct=8: State_ID > 272                                               Dropped
## DisAsDiff_fct=1: DisAsDiff <= 13554                                         0.039643
## DisAsDiff_fct=2: DisAsDiff <= 15670                                         0.034234
## DisAsDiff_fct=3: DisAsDiff <= 16661                                         0.035854
## DisAsDiff_fct=4: DisAsDiff <= 19822                                         0.025433
## DisAsDiff_fct=5: DisAsDiff > 19822                                           Dropped
## PRI_DISBURSED_AMOUNT_fct=1: PRI_DISBURSED_AMOUNT <= 218581                  0.038731
## PRI_DISBURSED_AMOUNT_fct=2: PRI_DISBURSED_AMOUNT > 218581                    Dropped
## PRI_OVERDUE_ACCTS_fct=1: PRI_OVERDUE_ACCTS <= 0                             0.028659
## PRI_OVERDUE_ACCTS_fct=2: PRI_OVERDUE_ACCTS > 0                               Dropped
## manufacturer_id_fct=1: manufacturer_id <= 210                               0.023889
## manufacturer_id_fct=2: manufacturer_id <= 221                               0.027798
## manufacturer_id_fct=3: manufacturer_id <= 228                               0.026050
## manufacturer_id_fct=4: manufacturer_id > 228                                 Dropped
## VoterID_flag_fct=1: VoterID_flag = 0                                        0.019475
## VoterID_flag_fct=2: VoterID_flag = 1                                         Dropped
## ShareOverdue_fct=1: ShareOverdue <= -2                                      0.031829
## ShareOverdue_fct=2: ShareOverdue <= -1                                      0.023251
## ShareOverdue_fct=3: ShareOverdue > -1                                        Dropped
## PRI_ACTIVE_ACCTS_fct=1: PRI_ACTIVE_ACCTS <= 0                               0.044260
## PRI_ACTIVE_ACCTS_fct=2: PRI_ACTIVE_ACCTS <= 1                               0.033428
## PRI_ACTIVE_ACCTS_fct=3: PRI_ACTIVE_ACCTS <= 3                               0.028821
## PRI_ACTIVE_ACCTS_fct=4: PRI_ACTIVE_ACCTS > 3                                 Dropped
## Day_fct=1: Day <= 28                                                        0.021306
## Day_fct=2: Day <= 30                                                        0.026022
## Day_fct=3: Day > 30                                                          Dropped
## NO_OF_INQUIRIES_fct=1: NO_OF_INQUIRIES <= 0                                 0.017019
## NO_OF_INQUIRIES_fct=2: NO_OF_INQUIRIES > 0                                   Dropped
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=1: DELINQUENT_ACCTS_IN_LAST_SIX_M <= 0   0.024440
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=2: DELINQUENT_ACCTS_IN_LAST_SIX_M > 0     Dropped
## Qrt_fct=1: Qrt = 3                                                          0.011791
## Qrt_fct=2: Qrt = 4                                                           Dropped
## YearsOnLoan_fct=1: YearsOnLoan <= 22.8918                                   0.030618
## YearsOnLoan_fct=2: YearsOnLoan <= 28.8496                                   0.026590
## YearsOnLoan_fct=3: YearsOnLoan <= 38.8321                                   0.025879
## YearsOnLoan_fct=4: YearsOnLoan <= 51.8208                                   0.026471
## YearsOnLoan_fct=5: YearsOnLoan > 51.8208                                     Dropped
## Employment_Type_fct=1: Employment_Type = 203                                0.012252
## Employment_Type_fct=2: Employment_Type = 215                                0.034675
## Employment_Type_fct=3: Employment_Type = 227                                 Dropped
## PRIMARY_INSTAL_AMT_fct=1: PRIMARY_INSTAL_AMT <= 1564                        0.032539
## PRIMARY_INSTAL_AMT_fct=2: PRIMARY_INSTAL_AMT <= 2832                        0.037537
## PRIMARY_INSTAL_AMT_fct=3: PRIMARY_INSTAL_AMT <= 5033                        0.038747
## PRIMARY_INSTAL_AMT_fct=4: PRIMARY_INSTAL_AMT <= 25326                       0.032532
## PRIMARY_INSTAL_AMT_fct=5: PRIMARY_INSTAL_AMT > 25326                         Dropped
## asset_cost_fct=1: asset_cost <= 60098                                       0.045737
## asset_cost_fct=2: asset_cost <= 70561                                       0.029595
## asset_cost_fct=3: asset_cost <= 85738                                       0.022427
## asset_cost_fct=4: asset_cost > 85738                                         Dropped
## SEC_OverdueShareSec_fct=1: SEC_OverdueShareSec <= 0                         0.093246
## SEC_OverdueShareSec_fct=11: SEC_OverdueShareSec Is Null                     0.102431
## SEC_OverdueShareSec_fct=8: SEC_OverdueShareSec <= 0.2                       0.245402
## SEC_OverdueShareSec_fct=9: SEC_OverdueShareSec <= 1                          Dropped
## SEC_CURRENT_BALANCE_fct=1: SEC_CURRENT_BALANCE <= 0                         0.077865
## SEC_CURRENT_BALANCE_fct=10: SEC_CURRENT_BALANCE > 0                          Dropped
## Passport_flag_fct=1: Passport_flag = 0                                      0.137099
## Passport_flag_fct=2: Passport_flag = 1                                       Dropped
## Driving_flag_fct=1: Driving_flag = 0                                        0.038427
## Driving_flag_fct=2: Driving_flag = 1                                         Dropped
## SEC_INSTAL_AMT_fct=1: SEC_INSTAL_AMT <= 0                                   0.077767
## SEC_INSTAL_AMT_fct=10: SEC_INSTAL_AMT > 0                                    Dropped
## PAN_flag_fct=1: PAN_flag = 0                                                0.023009
## PAN_flag_fct=2: PAN_flag = 1                                                 Dropped
## SEC_OVERDUE_ACCTS_fct=1: SEC_OVERDUE_ACCTS <= 0                              Dropped
## SEC_OVERDUE_ACCTS_fct=10: SEC_OVERDUE_ACCTS > 0                              Dropped
##                                                                           z value
## (Intercept)                                                                 3.696
## Employee_code_ID_fct=1: Employee_code_ID <= 134                            17.934
## Employee_code_ID_fct=10: Employee_code_ID <= 242                           -2.539
## Employee_code_ID_fct=11: Employee_code_ID <= 254                           -3.429
## Employee_code_ID_fct=12: Employee_code_ID <= 270                           -4.829
## Employee_code_ID_fct=13: Employee_code_ID <= 289                           -7.986
## Employee_code_ID_fct=14: Employee_code_ID <= 319                           -9.007
## Employee_code_ID_fct=15: Employee_code_ID > 319                           -15.050
## Employee_code_ID_fct=2: Employee_code_ID <= 153                            14.826
## Employee_code_ID_fct=3: Employee_code_ID <= 170                            13.451
## Employee_code_ID_fct=4: Employee_code_ID <= 179                             8.480
## Employee_code_ID_fct=5: Employee_code_ID <= 188                             7.909
## Employee_code_ID_fct=6: Employee_code_ID <= 199                             5.393
## Employee_code_ID_fct=7: Employee_code_ID <= 211                             3.221
## Employee_code_ID_fct=8: Employee_code_ID <= 221                             0.953
## Employee_code_ID_fct=9: Employee_code_ID <= 233                           Dropped
## Current_pincode_ID_fct=1: Current_pincode_ID <= 143                        22.072
## Current_pincode_ID_fct=10: Current_pincode_ID <= 257                       -7.381
## Current_pincode_ID_fct=11: Current_pincode_ID <= 291                      -11.568
## Current_pincode_ID_fct=12: Current_pincode_ID > 291                       -15.118
## Current_pincode_ID_fct=2: Current_pincode_ID <= 158                        20.306
## Current_pincode_ID_fct=3: Current_pincode_ID <= 174                        20.561
## Current_pincode_ID_fct=4: Current_pincode_ID <= 188                        18.122
## Current_pincode_ID_fct=5: Current_pincode_ID <= 201                        16.308
## Current_pincode_ID_fct=6: Current_pincode_ID <= 212                        12.457
## Current_pincode_ID_fct=7: Current_pincode_ID <= 219                         5.490
## Current_pincode_ID_fct=8: Current_pincode_ID <= 225                         3.629
## Current_pincode_ID_fct=9: Current_pincode_ID <= 238                       Dropped
## supplier_id_fct=1: supplier_id <= 133                                       7.874
## supplier_id_fct=10: supplier_id <= 253                                     -2.930
## supplier_id_fct=11: supplier_id <= 275                                     -3.924
## supplier_id_fct=12: supplier_id <= 314                                     -5.124
## supplier_id_fct=13: supplier_id > 314                                      -7.569
## supplier_id_fct=2: supplier_id <= 149                                       9.161
## supplier_id_fct=3: supplier_id <= 165                                       7.430
## supplier_id_fct=4: supplier_id <= 178                                       6.435
## supplier_id_fct=5: supplier_id <= 196                                       5.903
## supplier_id_fct=6: supplier_id <= 206                                       4.511
## supplier_id_fct=7: supplier_id <= 214                                       3.417
## supplier_id_fct=8: supplier_id <= 225                                       2.509
## supplier_id_fct=9: supplier_id <= 240                                     Dropped
## branch_id_fct=1: branch_id <= 153                                         -10.972
## branch_id_fct=10: branch_id <= 284                                          1.809
## branch_id_fct=11: branch_id > 284                                           1.272
## branch_id_fct=2: branch_id <= 174                                         -12.818
## branch_id_fct=3: branch_id <= 184                                         -12.090
## branch_id_fct=4: branch_id <= 198                                         -11.216
## branch_id_fct=5: branch_id <= 214                                          -6.546
## branch_id_fct=6: branch_id <= 222                                          -5.740
## branch_id_fct=7: branch_id <= 233                                          -3.047
## branch_id_fct=8: branch_id <= 261                                          -8.761
## branch_id_fct=9: branch_id <= 276                                         Dropped
## ltv_fct=1: ltv <= 55.63                                                    13.956
## ltv_fct=10: ltv <= 84.57                                                   -4.028
## ltv_fct=11: ltv <= 85                                                      -8.566
## ltv_fct=12: ltv <= 87.8                                                    -3.379
## ltv_fct=13: ltv <= 89.3                                                    -7.818
## ltv_fct=14: ltv > 89.3                                                     -9.308
## ltv_fct=2: ltv <= 62.22                                                    13.320
## ltv_fct=3: ltv <= 68.34                                                    10.661
## ltv_fct=4: ltv <= 72.9301                                                   8.081
## ltv_fct=5: ltv <= 74.31                                                     4.029
## ltv_fct=6: ltv <= 75                                                        2.689
## ltv_fct=7: ltv <= 77.39                                                     6.381
## ltv_fct=8: ltv <= 78.92                                                     2.209
## ltv_fct=9: ltv <= 83.34                                                   Dropped
## PERFORM_CNS_SCORE_fct=1: PERFORM_CNS_SCORE <= 0                            -2.204
## PERFORM_CNS_SCORE_fct=2: PERFORM_CNS_SCORE <= 18                           -1.035
## PERFORM_CNS_SCORE_fct=3: PERFORM_CNS_SCORE <= 441                          -3.491
## PERFORM_CNS_SCORE_fct=4: PERFORM_CNS_SCORE <= 643                          -2.249
## PERFORM_CNS_SCORE_fct=5: PERFORM_CNS_SCORE <= 738                          -1.927
## PERFORM_CNS_SCORE_fct=6: PERFORM_CNS_SCORE <= 824                           3.899
## PERFORM_CNS_SCORE_fct=7: PERFORM_CNS_SCORE > 824                          Dropped
## disbursed_amount_fct=1: disbursed_amount <= 39134                           0.823
## disbursed_amount_fct=2: disbursed_amount <= 43615                           1.992
## disbursed_amount_fct=3: disbursed_amount <= 48555                           2.796
## disbursed_amount_fct=4: disbursed_amount <= 51908                           3.874
## disbursed_amount_fct=5: disbursed_amount <= 55400                           2.737
## disbursed_amount_fct=6: disbursed_amount > 55400                          Dropped
## OutstandingNow_fct=1: OutstandingNow <= 44402                              -1.023
## OutstandingNow_fct=2: OutstandingNow <= 50314                              -2.244
## OutstandingNow_fct=3: OutstandingNow <= 171384                             -3.782
## OutstandingNow_fct=4: OutstandingNow <= 324324                             -1.999
## OutstandingNow_fct=5: OutstandingNow <= 746271                             -2.916
## OutstandingNow_fct=6: OutstandingNow > 746271                             Dropped
## PERFORM_CNS_SCORE_DESCRIPTION_fct=1: PERFORM_CNS_SCORE_DESCRIPTION <= 150   5.596
## PERFORM_CNS_SCORE_DESCRIPTION_fct=2: PERFORM_CNS_SCORE_DESCRIPTION <= 172   5.377
## PERFORM_CNS_SCORE_DESCRIPTION_fct=3: PERFORM_CNS_SCORE_DESCRIPTION <= 205   4.907
## PERFORM_CNS_SCORE_DESCRIPTION_fct=4: PERFORM_CNS_SCORE_DESCRIPTION <= 231   3.535
## PERFORM_CNS_SCORE_DESCRIPTION_fct=5: PERFORM_CNS_SCORE_DESCRIPTION <= 256   0.137
## PERFORM_CNS_SCORE_DESCRIPTION_fct=6: PERFORM_CNS_SCORE_DESCRIPTION > 256  Dropped
## State_ID_fct=1: State_ID <= 183                                            -1.208
## State_ID_fct=2: State_ID <= 188                                            -0.633
## State_ID_fct=3: State_ID <= 206                                            -0.050
## State_ID_fct=4: State_ID <= 214                                             1.872
## State_ID_fct=5: State_ID <= 220                                             2.030
## State_ID_fct=6: State_ID <= 229                                             1.250
## State_ID_fct=7: State_ID <= 272                                             1.905
## State_ID_fct=8: State_ID > 272                                            Dropped
## DisAsDiff_fct=1: DisAsDiff <= 13554                                        -3.285
## DisAsDiff_fct=2: DisAsDiff <= 15670                                        -2.402
## DisAsDiff_fct=3: DisAsDiff <= 16661                                         0.172
## DisAsDiff_fct=4: DisAsDiff <= 19822                                        -0.687
## DisAsDiff_fct=5: DisAsDiff > 19822                                        Dropped
## PRI_DISBURSED_AMOUNT_fct=1: PRI_DISBURSED_AMOUNT <= 218581                 -6.609
## PRI_DISBURSED_AMOUNT_fct=2: PRI_DISBURSED_AMOUNT > 218581                 Dropped
## PRI_OVERDUE_ACCTS_fct=1: PRI_OVERDUE_ACCTS <= 0                             5.037
## PRI_OVERDUE_ACCTS_fct=2: PRI_OVERDUE_ACCTS > 0                            Dropped
## manufacturer_id_fct=1: manufacturer_id <= 210                               5.159
## manufacturer_id_fct=2: manufacturer_id <= 221                              -0.127
## manufacturer_id_fct=3: manufacturer_id <= 228                               3.751
## manufacturer_id_fct=4: manufacturer_id > 228                              Dropped
## VoterID_flag_fct=1: VoterID_flag = 0                                        4.800
## VoterID_flag_fct=2: VoterID_flag = 1                                      Dropped
## ShareOverdue_fct=1: ShareOverdue <= -2                                      2.900
## ShareOverdue_fct=2: ShareOverdue <= -1                                      5.542
## ShareOverdue_fct=3: ShareOverdue > -1                                     Dropped
## PRI_ACTIVE_ACCTS_fct=1: PRI_ACTIVE_ACCTS <= 0                              -5.131
## PRI_ACTIVE_ACCTS_fct=2: PRI_ACTIVE_ACCTS <= 1                              -7.267
## PRI_ACTIVE_ACCTS_fct=3: PRI_ACTIVE_ACCTS <= 3                              -7.273
## PRI_ACTIVE_ACCTS_fct=4: PRI_ACTIVE_ACCTS > 3                              Dropped
## Day_fct=1: Day <= 28                                                       11.705
## Day_fct=2: Day <= 30                                                        4.930
## Day_fct=3: Day > 30                                                       Dropped
## NO_OF_INQUIRIES_fct=1: NO_OF_INQUIRIES <= 0                                16.055
## NO_OF_INQUIRIES_fct=2: NO_OF_INQUIRIES > 0                                Dropped
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=1: DELINQUENT_ACCTS_IN_LAST_SIX_M <= 0  11.536
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=2: DELINQUENT_ACCTS_IN_LAST_SIX_M > 0  Dropped
## Qrt_fct=1: Qrt = 3                                                         18.577
## Qrt_fct=2: Qrt = 4                                                        Dropped
## YearsOnLoan_fct=1: YearsOnLoan <= 22.8918                                 -12.545
## YearsOnLoan_fct=2: YearsOnLoan <= 28.8496                                 -10.468
## YearsOnLoan_fct=3: YearsOnLoan <= 38.8321                                  -6.953
## YearsOnLoan_fct=4: YearsOnLoan <= 51.8208                                  -3.513
## YearsOnLoan_fct=5: YearsOnLoan > 51.8208                                  Dropped
## Employment_Type_fct=1: Employment_Type = 203                               12.443
## Employment_Type_fct=2: Employment_Type = 215                                6.247
## Employment_Type_fct=3: Employment_Type = 227                              Dropped
## PRIMARY_INSTAL_AMT_fct=1: PRIMARY_INSTAL_AMT <= 1564                       -0.552
## PRIMARY_INSTAL_AMT_fct=2: PRIMARY_INSTAL_AMT <= 2832                       -6.820
## PRIMARY_INSTAL_AMT_fct=3: PRIMARY_INSTAL_AMT <= 5033                       -3.723
## PRIMARY_INSTAL_AMT_fct=4: PRIMARY_INSTAL_AMT <= 25326                      -4.416
## PRIMARY_INSTAL_AMT_fct=5: PRIMARY_INSTAL_AMT > 25326                      Dropped
## asset_cost_fct=1: asset_cost <= 60098                                       1.587
## asset_cost_fct=2: asset_cost <= 70561                                       5.047
## asset_cost_fct=3: asset_cost <= 85738                                       7.920
## asset_cost_fct=4: asset_cost > 85738                                      Dropped
## SEC_OverdueShareSec_fct=1: SEC_OverdueShareSec <= 0                        -0.642
## SEC_OverdueShareSec_fct=11: SEC_OverdueShareSec Is Null                    -0.734
## SEC_OverdueShareSec_fct=8: SEC_OverdueShareSec <= 0.2                       1.167
## SEC_OverdueShareSec_fct=9: SEC_OverdueShareSec <= 1                       Dropped
## SEC_CURRENT_BALANCE_fct=1: SEC_CURRENT_BALANCE <= 0                        -0.307
## SEC_CURRENT_BALANCE_fct=10: SEC_CURRENT_BALANCE > 0                       Dropped
## Passport_flag_fct=1: Passport_flag = 0                                     -1.371
## Passport_flag_fct=2: Passport_flag = 1                                    Dropped
## Driving_flag_fct=1: Driving_flag = 0                                        0.414
## Driving_flag_fct=2: Driving_flag = 1                                      Dropped
## SEC_INSTAL_AMT_fct=1: SEC_INSTAL_AMT <= 0                                   0.783
## SEC_INSTAL_AMT_fct=10: SEC_INSTAL_AMT > 0                                 Dropped
## PAN_flag_fct=1: PAN_flag = 0                                               -1.998
## PAN_flag_fct=2: PAN_flag = 1                                              Dropped
## SEC_OVERDUE_ACCTS_fct=1: SEC_OVERDUE_ACCTS <= 0                           Dropped
## SEC_OVERDUE_ACCTS_fct=10: SEC_OVERDUE_ACCTS > 0                           Dropped
##                                                                                       Pr(>|z|)
## (Intercept)                                                                           0.000219
## Employee_code_ID_fct=1: Employee_code_ID <= 134                           0.000000000000000222
## Employee_code_ID_fct=10: Employee_code_ID <= 242                                      0.011127
## Employee_code_ID_fct=11: Employee_code_ID <= 254                                      0.000605
## Employee_code_ID_fct=12: Employee_code_ID <= 270                          0.000001373400840832
## Employee_code_ID_fct=13: Employee_code_ID <= 289                          0.000000000000000222
## Employee_code_ID_fct=14: Employee_code_ID <= 319                          0.000000000000000222
## Employee_code_ID_fct=15: Employee_code_ID > 319                           0.000000000000000222
## Employee_code_ID_fct=2: Employee_code_ID <= 153                           0.000000000000000222
## Employee_code_ID_fct=3: Employee_code_ID <= 170                           0.000000000000000222
## Employee_code_ID_fct=4: Employee_code_ID <= 179                           0.000000000000000222
## Employee_code_ID_fct=5: Employee_code_ID <= 188                           0.000000000000000222
## Employee_code_ID_fct=6: Employee_code_ID <= 199                           0.000000069185031482
## Employee_code_ID_fct=7: Employee_code_ID <= 211                                       0.001279
## Employee_code_ID_fct=8: Employee_code_ID <= 221                                       0.340780
## Employee_code_ID_fct=9: Employee_code_ID <= 233                                        Dropped
## Current_pincode_ID_fct=1: Current_pincode_ID <= 143                       0.000000000000000222
## Current_pincode_ID_fct=10: Current_pincode_ID <= 257                      0.000000000000000222
## Current_pincode_ID_fct=11: Current_pincode_ID <= 291                      0.000000000000000222
## Current_pincode_ID_fct=12: Current_pincode_ID > 291                       0.000000000000000222
## Current_pincode_ID_fct=2: Current_pincode_ID <= 158                       0.000000000000000222
## Current_pincode_ID_fct=3: Current_pincode_ID <= 174                       0.000000000000000222
## Current_pincode_ID_fct=4: Current_pincode_ID <= 188                       0.000000000000000222
## Current_pincode_ID_fct=5: Current_pincode_ID <= 201                       0.000000000000000222
## Current_pincode_ID_fct=6: Current_pincode_ID <= 212                       0.000000000000000222
## Current_pincode_ID_fct=7: Current_pincode_ID <= 219                       0.000000040128663947
## Current_pincode_ID_fct=8: Current_pincode_ID <= 225                                   0.000285
## Current_pincode_ID_fct=9: Current_pincode_ID <= 238                                    Dropped
## supplier_id_fct=1: supplier_id <= 133                                     0.000000000000000222
## supplier_id_fct=10: supplier_id <= 253                                                0.003385
## supplier_id_fct=11: supplier_id <= 275                                    0.000086946671635557
## supplier_id_fct=12: supplier_id <= 314                                    0.000000299739931098
## supplier_id_fct=13: supplier_id > 314                                     0.000000000000000222
## supplier_id_fct=2: supplier_id <= 149                                     0.000000000000000222
## supplier_id_fct=3: supplier_id <= 165                                     0.000000000000000222
## supplier_id_fct=4: supplier_id <= 178                                     0.000000000123338895
## supplier_id_fct=5: supplier_id <= 196                                     0.000000003559513795
## supplier_id_fct=6: supplier_id <= 206                                     0.000006464945373930
## supplier_id_fct=7: supplier_id <= 214                                                 0.000633
## supplier_id_fct=8: supplier_id <= 225                                                 0.012114
## supplier_id_fct=9: supplier_id <= 240                                                  Dropped
## branch_id_fct=1: branch_id <= 153                                         0.000000000000000222
## branch_id_fct=10: branch_id <= 284                                                    0.070464
## branch_id_fct=11: branch_id > 284                                                     0.203439
## branch_id_fct=2: branch_id <= 174                                         0.000000000000000222
## branch_id_fct=3: branch_id <= 184                                         0.000000000000000222
## branch_id_fct=4: branch_id <= 198                                         0.000000000000000222
## branch_id_fct=5: branch_id <= 214                                         0.000000000059036109
## branch_id_fct=6: branch_id <= 222                                         0.000000009467683748
## branch_id_fct=7: branch_id <= 233                                                     0.002310
## branch_id_fct=8: branch_id <= 261                                         0.000000000000000222
## branch_id_fct=9: branch_id <= 276                                                      Dropped
## ltv_fct=1: ltv <= 55.63                                                   0.000000000000000222
## ltv_fct=10: ltv <= 84.57                                                  0.000056194942485766
## ltv_fct=11: ltv <= 85                                                     0.000000000000000222
## ltv_fct=12: ltv <= 87.8                                                               0.000728
## ltv_fct=13: ltv <= 89.3                                                   0.000000000000000222
## ltv_fct=14: ltv > 89.3                                                    0.000000000000000222
## ltv_fct=2: ltv <= 62.22                                                   0.000000000000000222
## ltv_fct=3: ltv <= 68.34                                                   0.000000000000000222
## ltv_fct=4: ltv <= 72.9301                                                 0.000000000000000222
## ltv_fct=5: ltv <= 74.31                                                   0.000056050266953989
## ltv_fct=6: ltv <= 75                                                                  0.007169
## ltv_fct=7: ltv <= 77.39                                                   0.000000000176378467
## ltv_fct=8: ltv <= 78.92                                                               0.027181
## ltv_fct=9: ltv <= 83.34                                                                Dropped
## PERFORM_CNS_SCORE_fct=1: PERFORM_CNS_SCORE <= 0                                       0.027551
## PERFORM_CNS_SCORE_fct=2: PERFORM_CNS_SCORE <= 18                                      0.300902
## PERFORM_CNS_SCORE_fct=3: PERFORM_CNS_SCORE <= 441                                     0.000481
## PERFORM_CNS_SCORE_fct=4: PERFORM_CNS_SCORE <= 643                                     0.024504
## PERFORM_CNS_SCORE_fct=5: PERFORM_CNS_SCORE <= 738                                     0.053931
## PERFORM_CNS_SCORE_fct=6: PERFORM_CNS_SCORE <= 824                         0.000096707880559599
## PERFORM_CNS_SCORE_fct=7: PERFORM_CNS_SCORE > 824                                       Dropped
## disbursed_amount_fct=1: disbursed_amount <= 39134                                     0.410269
## disbursed_amount_fct=2: disbursed_amount <= 43615                                     0.046391
## disbursed_amount_fct=3: disbursed_amount <= 48555                                     0.005176
## disbursed_amount_fct=4: disbursed_amount <= 51908                                     0.000107
## disbursed_amount_fct=5: disbursed_amount <= 55400                                     0.006200
## disbursed_amount_fct=6: disbursed_amount > 55400                                       Dropped
## OutstandingNow_fct=1: OutstandingNow <= 44402                                         0.306128
## OutstandingNow_fct=2: OutstandingNow <= 50314                                         0.024810
## OutstandingNow_fct=3: OutstandingNow <= 171384                                        0.000155
## OutstandingNow_fct=4: OutstandingNow <= 324324                                        0.045628
## OutstandingNow_fct=5: OutstandingNow <= 746271                                        0.003547
## OutstandingNow_fct=6: OutstandingNow > 746271                                          Dropped
## PERFORM_CNS_SCORE_DESCRIPTION_fct=1: PERFORM_CNS_SCORE_DESCRIPTION <= 150 0.000000021892954560
## PERFORM_CNS_SCORE_DESCRIPTION_fct=2: PERFORM_CNS_SCORE_DESCRIPTION <= 172 0.000000075610862016
## PERFORM_CNS_SCORE_DESCRIPTION_fct=3: PERFORM_CNS_SCORE_DESCRIPTION <= 205 0.000000924302139271
## PERFORM_CNS_SCORE_DESCRIPTION_fct=4: PERFORM_CNS_SCORE_DESCRIPTION <= 231             0.000407
## PERFORM_CNS_SCORE_DESCRIPTION_fct=5: PERFORM_CNS_SCORE_DESCRIPTION <= 256             0.891012
## PERFORM_CNS_SCORE_DESCRIPTION_fct=6: PERFORM_CNS_SCORE_DESCRIPTION > 256               Dropped
## State_ID_fct=1: State_ID <= 183                                                       0.227109
## State_ID_fct=2: State_ID <= 188                                                       0.526790
## State_ID_fct=3: State_ID <= 206                                                       0.960185
## State_ID_fct=4: State_ID <= 214                                                       0.061214
## State_ID_fct=5: State_ID <= 220                                                       0.042315
## State_ID_fct=6: State_ID <= 229                                                       0.211464
## State_ID_fct=7: State_ID <= 272                                                       0.056786
## State_ID_fct=8: State_ID > 272                                                         Dropped
## DisAsDiff_fct=1: DisAsDiff <= 13554                                                   0.001018
## DisAsDiff_fct=2: DisAsDiff <= 15670                                                   0.016306
## DisAsDiff_fct=3: DisAsDiff <= 16661                                                   0.863261
## DisAsDiff_fct=4: DisAsDiff <= 19822                                                   0.492117
## DisAsDiff_fct=5: DisAsDiff > 19822                                                     Dropped
## PRI_DISBURSED_AMOUNT_fct=1: PRI_DISBURSED_AMOUNT <= 218581                0.000000000038574033
## PRI_DISBURSED_AMOUNT_fct=2: PRI_DISBURSED_AMOUNT > 218581                              Dropped
## PRI_OVERDUE_ACCTS_fct=1: PRI_OVERDUE_ACCTS <= 0                           0.000000472350173641
## PRI_OVERDUE_ACCTS_fct=2: PRI_OVERDUE_ACCTS > 0                                         Dropped
## manufacturer_id_fct=1: manufacturer_id <= 210                             0.000000248386704094
## manufacturer_id_fct=2: manufacturer_id <= 221                                         0.899275
## manufacturer_id_fct=3: manufacturer_id <= 228                                         0.000176
## manufacturer_id_fct=4: manufacturer_id > 228                                           Dropped
## VoterID_flag_fct=1: VoterID_flag = 0                                      0.000001583545196970
## VoterID_flag_fct=2: VoterID_flag = 1                                                   Dropped
## ShareOverdue_fct=1: ShareOverdue <= -2                                                0.003735
## ShareOverdue_fct=2: ShareOverdue <= -1                                    0.000000029958245662
## ShareOverdue_fct=3: ShareOverdue > -1                                                  Dropped
## PRI_ACTIVE_ACCTS_fct=1: PRI_ACTIVE_ACCTS <= 0                             0.000000287604146276
## PRI_ACTIVE_ACCTS_fct=2: PRI_ACTIVE_ACCTS <= 1                             0.000000000000000222
## PRI_ACTIVE_ACCTS_fct=3: PRI_ACTIVE_ACCTS <= 3                             0.000000000000000222
## PRI_ACTIVE_ACCTS_fct=4: PRI_ACTIVE_ACCTS > 3                                           Dropped
## Day_fct=1: Day <= 28                                                      0.000000000000000222
## Day_fct=2: Day <= 30                                                      0.000000822043672688
## Day_fct=3: Day > 30                                                                    Dropped
## NO_OF_INQUIRIES_fct=1: NO_OF_INQUIRIES <= 0                               0.000000000000000222
## NO_OF_INQUIRIES_fct=2: NO_OF_INQUIRIES > 0                                             Dropped
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=1: DELINQUENT_ACCTS_IN_LAST_SIX_M <= 0 0.000000000000000222
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=2: DELINQUENT_ACCTS_IN_LAST_SIX_M > 0               Dropped
## Qrt_fct=1: Qrt = 3                                                        0.000000000000000222
## Qrt_fct=2: Qrt = 4                                                                     Dropped
## YearsOnLoan_fct=1: YearsOnLoan <= 22.8918                                 0.000000000000000222
## YearsOnLoan_fct=2: YearsOnLoan <= 28.8496                                 0.000000000000000222
## YearsOnLoan_fct=3: YearsOnLoan <= 38.8321                                 0.000000000003574030
## YearsOnLoan_fct=4: YearsOnLoan <= 51.8208                                             0.000444
## YearsOnLoan_fct=5: YearsOnLoan > 51.8208                                               Dropped
## Employment_Type_fct=1: Employment_Type = 203                              0.000000000000000222
## Employment_Type_fct=2: Employment_Type = 215                              0.000000000417368806
## Employment_Type_fct=3: Employment_Type = 227                                           Dropped
## PRIMARY_INSTAL_AMT_fct=1: PRIMARY_INSTAL_AMT <= 1564                                  0.581034
## PRIMARY_INSTAL_AMT_fct=2: PRIMARY_INSTAL_AMT <= 2832                      0.000000000009078738
## PRIMARY_INSTAL_AMT_fct=3: PRIMARY_INSTAL_AMT <= 5033                                  0.000197
## PRIMARY_INSTAL_AMT_fct=4: PRIMARY_INSTAL_AMT <= 25326                     0.000010036155548399
## PRIMARY_INSTAL_AMT_fct=5: PRIMARY_INSTAL_AMT > 25326                                   Dropped
## asset_cost_fct=1: asset_cost <= 60098                                                 0.112444
## asset_cost_fct=2: asset_cost <= 70561                                     0.000000449390024304
## asset_cost_fct=3: asset_cost <= 85738                                     0.000000000000000222
## asset_cost_fct=4: asset_cost > 85738                                                   Dropped
## SEC_OverdueShareSec_fct=1: SEC_OverdueShareSec <= 0                                   0.520648
## SEC_OverdueShareSec_fct=11: SEC_OverdueShareSec Is Null                               0.463016
## SEC_OverdueShareSec_fct=8: SEC_OverdueShareSec <= 0.2                                 0.243122
## SEC_OverdueShareSec_fct=9: SEC_OverdueShareSec <= 1                                    Dropped
## SEC_CURRENT_BALANCE_fct=1: SEC_CURRENT_BALANCE <= 0                                   0.758758
## SEC_CURRENT_BALANCE_fct=10: SEC_CURRENT_BALANCE > 0                                    Dropped
## Passport_flag_fct=1: Passport_flag = 0                                                0.170376
## Passport_flag_fct=2: Passport_flag = 1                                                 Dropped
## Driving_flag_fct=1: Driving_flag = 0                                                  0.678510
## Driving_flag_fct=2: Driving_flag = 1                                                   Dropped
## SEC_INSTAL_AMT_fct=1: SEC_INSTAL_AMT <= 0                                             0.433359
## SEC_INSTAL_AMT_fct=10: SEC_INSTAL_AMT > 0                                              Dropped
## PAN_flag_fct=1: PAN_flag = 0                                                          0.045763
## PAN_flag_fct=2: PAN_flag = 1                                                           Dropped
## SEC_OVERDUE_ACCTS_fct=1: SEC_OVERDUE_ACCTS <= 0                                        Dropped
## SEC_OVERDUE_ACCTS_fct=10: SEC_OVERDUE_ACCTS > 0                                        Dropped
##                                                                              
## (Intercept)                                                               ***
## Employee_code_ID_fct=1: Employee_code_ID <= 134                           ***
## Employee_code_ID_fct=10: Employee_code_ID <= 242                          *  
## Employee_code_ID_fct=11: Employee_code_ID <= 254                          ***
## Employee_code_ID_fct=12: Employee_code_ID <= 270                          ***
## Employee_code_ID_fct=13: Employee_code_ID <= 289                          ***
## Employee_code_ID_fct=14: Employee_code_ID <= 319                          ***
## Employee_code_ID_fct=15: Employee_code_ID > 319                           ***
## Employee_code_ID_fct=2: Employee_code_ID <= 153                           ***
## Employee_code_ID_fct=3: Employee_code_ID <= 170                           ***
## Employee_code_ID_fct=4: Employee_code_ID <= 179                           ***
## Employee_code_ID_fct=5: Employee_code_ID <= 188                           ***
## Employee_code_ID_fct=6: Employee_code_ID <= 199                           ***
## Employee_code_ID_fct=7: Employee_code_ID <= 211                           ** 
## Employee_code_ID_fct=8: Employee_code_ID <= 221                              
## Employee_code_ID_fct=9: Employee_code_ID <= 233                              
## Current_pincode_ID_fct=1: Current_pincode_ID <= 143                       ***
## Current_pincode_ID_fct=10: Current_pincode_ID <= 257                      ***
## Current_pincode_ID_fct=11: Current_pincode_ID <= 291                      ***
## Current_pincode_ID_fct=12: Current_pincode_ID > 291                       ***
## Current_pincode_ID_fct=2: Current_pincode_ID <= 158                       ***
## Current_pincode_ID_fct=3: Current_pincode_ID <= 174                       ***
## Current_pincode_ID_fct=4: Current_pincode_ID <= 188                       ***
## Current_pincode_ID_fct=5: Current_pincode_ID <= 201                       ***
## Current_pincode_ID_fct=6: Current_pincode_ID <= 212                       ***
## Current_pincode_ID_fct=7: Current_pincode_ID <= 219                       ***
## Current_pincode_ID_fct=8: Current_pincode_ID <= 225                       ***
## Current_pincode_ID_fct=9: Current_pincode_ID <= 238                          
## supplier_id_fct=1: supplier_id <= 133                                     ***
## supplier_id_fct=10: supplier_id <= 253                                    ** 
## supplier_id_fct=11: supplier_id <= 275                                    ***
## supplier_id_fct=12: supplier_id <= 314                                    ***
## supplier_id_fct=13: supplier_id > 314                                     ***
## supplier_id_fct=2: supplier_id <= 149                                     ***
## supplier_id_fct=3: supplier_id <= 165                                     ***
## supplier_id_fct=4: supplier_id <= 178                                     ***
## supplier_id_fct=5: supplier_id <= 196                                     ***
## supplier_id_fct=6: supplier_id <= 206                                     ***
## supplier_id_fct=7: supplier_id <= 214                                     ***
## supplier_id_fct=8: supplier_id <= 225                                     *  
## supplier_id_fct=9: supplier_id <= 240                                        
## branch_id_fct=1: branch_id <= 153                                         ***
## branch_id_fct=10: branch_id <= 284                                        .  
## branch_id_fct=11: branch_id > 284                                            
## branch_id_fct=2: branch_id <= 174                                         ***
## branch_id_fct=3: branch_id <= 184                                         ***
## branch_id_fct=4: branch_id <= 198                                         ***
## branch_id_fct=5: branch_id <= 214                                         ***
## branch_id_fct=6: branch_id <= 222                                         ***
## branch_id_fct=7: branch_id <= 233                                         ** 
## branch_id_fct=8: branch_id <= 261                                         ***
## branch_id_fct=9: branch_id <= 276                                            
## ltv_fct=1: ltv <= 55.63                                                   ***
## ltv_fct=10: ltv <= 84.57                                                  ***
## ltv_fct=11: ltv <= 85                                                     ***
## ltv_fct=12: ltv <= 87.8                                                   ***
## ltv_fct=13: ltv <= 89.3                                                   ***
## ltv_fct=14: ltv > 89.3                                                    ***
## ltv_fct=2: ltv <= 62.22                                                   ***
## ltv_fct=3: ltv <= 68.34                                                   ***
## ltv_fct=4: ltv <= 72.9301                                                 ***
## ltv_fct=5: ltv <= 74.31                                                   ***
## ltv_fct=6: ltv <= 75                                                      ** 
## ltv_fct=7: ltv <= 77.39                                                   ***
## ltv_fct=8: ltv <= 78.92                                                   *  
## ltv_fct=9: ltv <= 83.34                                                      
## PERFORM_CNS_SCORE_fct=1: PERFORM_CNS_SCORE <= 0                           *  
## PERFORM_CNS_SCORE_fct=2: PERFORM_CNS_SCORE <= 18                             
## PERFORM_CNS_SCORE_fct=3: PERFORM_CNS_SCORE <= 441                         ***
## PERFORM_CNS_SCORE_fct=4: PERFORM_CNS_SCORE <= 643                         *  
## PERFORM_CNS_SCORE_fct=5: PERFORM_CNS_SCORE <= 738                         .  
## PERFORM_CNS_SCORE_fct=6: PERFORM_CNS_SCORE <= 824                         ***
## PERFORM_CNS_SCORE_fct=7: PERFORM_CNS_SCORE > 824                             
## disbursed_amount_fct=1: disbursed_amount <= 39134                            
## disbursed_amount_fct=2: disbursed_amount <= 43615                         *  
## disbursed_amount_fct=3: disbursed_amount <= 48555                         ** 
## disbursed_amount_fct=4: disbursed_amount <= 51908                         ***
## disbursed_amount_fct=5: disbursed_amount <= 55400                         ** 
## disbursed_amount_fct=6: disbursed_amount > 55400                             
## OutstandingNow_fct=1: OutstandingNow <= 44402                                
## OutstandingNow_fct=2: OutstandingNow <= 50314                             *  
## OutstandingNow_fct=3: OutstandingNow <= 171384                            ***
## OutstandingNow_fct=4: OutstandingNow <= 324324                            *  
## OutstandingNow_fct=5: OutstandingNow <= 746271                            ** 
## OutstandingNow_fct=6: OutstandingNow > 746271                                
## PERFORM_CNS_SCORE_DESCRIPTION_fct=1: PERFORM_CNS_SCORE_DESCRIPTION <= 150 ***
## PERFORM_CNS_SCORE_DESCRIPTION_fct=2: PERFORM_CNS_SCORE_DESCRIPTION <= 172 ***
## PERFORM_CNS_SCORE_DESCRIPTION_fct=3: PERFORM_CNS_SCORE_DESCRIPTION <= 205 ***
## PERFORM_CNS_SCORE_DESCRIPTION_fct=4: PERFORM_CNS_SCORE_DESCRIPTION <= 231 ***
## PERFORM_CNS_SCORE_DESCRIPTION_fct=5: PERFORM_CNS_SCORE_DESCRIPTION <= 256    
## PERFORM_CNS_SCORE_DESCRIPTION_fct=6: PERFORM_CNS_SCORE_DESCRIPTION > 256     
## State_ID_fct=1: State_ID <= 183                                              
## State_ID_fct=2: State_ID <= 188                                              
## State_ID_fct=3: State_ID <= 206                                              
## State_ID_fct=4: State_ID <= 214                                           .  
## State_ID_fct=5: State_ID <= 220                                           *  
## State_ID_fct=6: State_ID <= 229                                              
## State_ID_fct=7: State_ID <= 272                                           .  
## State_ID_fct=8: State_ID > 272                                               
## DisAsDiff_fct=1: DisAsDiff <= 13554                                       ** 
## DisAsDiff_fct=2: DisAsDiff <= 15670                                       *  
## DisAsDiff_fct=3: DisAsDiff <= 16661                                          
## DisAsDiff_fct=4: DisAsDiff <= 19822                                          
## DisAsDiff_fct=5: DisAsDiff > 19822                                           
## PRI_DISBURSED_AMOUNT_fct=1: PRI_DISBURSED_AMOUNT <= 218581                ***
## PRI_DISBURSED_AMOUNT_fct=2: PRI_DISBURSED_AMOUNT > 218581                    
## PRI_OVERDUE_ACCTS_fct=1: PRI_OVERDUE_ACCTS <= 0                           ***
## PRI_OVERDUE_ACCTS_fct=2: PRI_OVERDUE_ACCTS > 0                               
## manufacturer_id_fct=1: manufacturer_id <= 210                             ***
## manufacturer_id_fct=2: manufacturer_id <= 221                                
## manufacturer_id_fct=3: manufacturer_id <= 228                             ***
## manufacturer_id_fct=4: manufacturer_id > 228                                 
## VoterID_flag_fct=1: VoterID_flag = 0                                      ***
## VoterID_flag_fct=2: VoterID_flag = 1                                         
## ShareOverdue_fct=1: ShareOverdue <= -2                                    ** 
## ShareOverdue_fct=2: ShareOverdue <= -1                                    ***
## ShareOverdue_fct=3: ShareOverdue > -1                                        
## PRI_ACTIVE_ACCTS_fct=1: PRI_ACTIVE_ACCTS <= 0                             ***
## PRI_ACTIVE_ACCTS_fct=2: PRI_ACTIVE_ACCTS <= 1                             ***
## PRI_ACTIVE_ACCTS_fct=3: PRI_ACTIVE_ACCTS <= 3                             ***
## PRI_ACTIVE_ACCTS_fct=4: PRI_ACTIVE_ACCTS > 3                                 
## Day_fct=1: Day <= 28                                                      ***
## Day_fct=2: Day <= 30                                                      ***
## Day_fct=3: Day > 30                                                          
## NO_OF_INQUIRIES_fct=1: NO_OF_INQUIRIES <= 0                               ***
## NO_OF_INQUIRIES_fct=2: NO_OF_INQUIRIES > 0                                   
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=1: DELINQUENT_ACCTS_IN_LAST_SIX_M <= 0 ***
## DELINQUENT_ACCTS_IN_LAST_SIX_M_fct=2: DELINQUENT_ACCTS_IN_LAST_SIX_M > 0     
## Qrt_fct=1: Qrt = 3                                                        ***
## Qrt_fct=2: Qrt = 4                                                           
## YearsOnLoan_fct=1: YearsOnLoan <= 22.8918                                 ***
## YearsOnLoan_fct=2: YearsOnLoan <= 28.8496                                 ***
## YearsOnLoan_fct=3: YearsOnLoan <= 38.8321                                 ***
## YearsOnLoan_fct=4: YearsOnLoan <= 51.8208                                 ***
## YearsOnLoan_fct=5: YearsOnLoan > 51.8208                                     
## Employment_Type_fct=1: Employment_Type = 203                              ***
## Employment_Type_fct=2: Employment_Type = 215                              ***
## Employment_Type_fct=3: Employment_Type = 227                                 
## PRIMARY_INSTAL_AMT_fct=1: PRIMARY_INSTAL_AMT <= 1564                         
## PRIMARY_INSTAL_AMT_fct=2: PRIMARY_INSTAL_AMT <= 2832                      ***
## PRIMARY_INSTAL_AMT_fct=3: PRIMARY_INSTAL_AMT <= 5033                      ***
## PRIMARY_INSTAL_AMT_fct=4: PRIMARY_INSTAL_AMT <= 25326                     ***
## PRIMARY_INSTAL_AMT_fct=5: PRIMARY_INSTAL_AMT > 25326                         
## asset_cost_fct=1: asset_cost <= 60098                                        
## asset_cost_fct=2: asset_cost <= 70561                                     ***
## asset_cost_fct=3: asset_cost <= 85738                                     ***
## asset_cost_fct=4: asset_cost > 85738                                         
## SEC_OverdueShareSec_fct=1: SEC_OverdueShareSec <= 0                          
## SEC_OverdueShareSec_fct=11: SEC_OverdueShareSec Is Null                      
## SEC_OverdueShareSec_fct=8: SEC_OverdueShareSec <= 0.2                        
## SEC_OverdueShareSec_fct=9: SEC_OverdueShareSec <= 1                          
## SEC_CURRENT_BALANCE_fct=1: SEC_CURRENT_BALANCE <= 0                          
## SEC_CURRENT_BALANCE_fct=10: SEC_CURRENT_BALANCE > 0                          
## Passport_flag_fct=1: Passport_flag = 0                                       
## Passport_flag_fct=2: Passport_flag = 1                                       
## Driving_flag_fct=1: Driving_flag = 0                                         
## Driving_flag_fct=2: Driving_flag = 1                                         
## SEC_INSTAL_AMT_fct=1: SEC_INSTAL_AMT <= 0                                    
## SEC_INSTAL_AMT_fct=10: SEC_INSTAL_AMT > 0                                    
## PAN_flag_fct=1: PAN_flag = 0                                              *  
## PAN_flag_fct=2: PAN_flag = 1                                                 
## SEC_OVERDUE_ACCTS_fct=1: SEC_OVERDUE_ACCTS <= 0                              
## SEC_OVERDUE_ACCTS_fct=10: SEC_OVERDUE_ACCTS > 0                              
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Condition number of final variance-covariance matrix: 1997.764 
## Number of iterations: 6

# Estimate a Classification Result on Testing Set
Pred <- 
  RevoScaleR::rxPredict(modelObject = rxLogitFit
    , data = X[inTest, ] #, sqlFraudDS
    # , outData = sqlServerOutDS3
    , predVarNames = 'logitProbs'
    , type = 'response'
    , writeModelVars = FALSE
    # , extraVarsToWrite = 'SUBS_KEY'
    , overwrite = TRUE )

## Rows Read: 23316, Total Rows Processed: 23316, Total Chunk Time: 0.019 seconds

Probs <- Pred[, 'logitProbs']

# PrClass <- factor(ifelse( RevoScaleR::rxImport(sqlServerOutDS3)[, 'predLogitProbs'] < 0.5, 0, 1))
PrClass <- ifelse( Pred[, 'logitProbs'] < 0.5, 0, 1) %>%
    factor
levels(PrClass) = c('Bad', 'Good')

ObClass <- Y[inTest]

writeLines('\n Estimate a Classification Result on Testing Set \n')

## 
##  Estimate a Classification Result on Testing Set

caret::confusionMatrix(data = PrClass, reference = ObClass, positive = 'Bad', mode = 'everything')

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction   Bad  Good
##       Bad    284   338
##       Good  4770 17924
##                                              
##                Accuracy : 0.7809             
##                  95% CI : (0.7756, 0.7862)   
##     No Information Rate : 0.7832             
##     P-Value [Acc > NIR] : 0.8069             
##                                              
##                   Kappa : 0.0552             
##  Mcnemar's Test P-Value : <0.0000000000000002
##                                              
##             Sensitivity : 0.05619            
##             Specificity : 0.98149            
##          Pos Pred Value : 0.45659            
##          Neg Pred Value : 0.78981            
##               Precision : 0.45659            
##                  Recall : 0.05619            
##                      F1 : 0.10007            
##              Prevalence : 0.21676            
##          Detection Rate : 0.01218            
##    Detection Prevalence : 0.02668            
##       Balanced Accuracy : 0.51884            
##                                              
##        'Positive' Class : Bad                
## 

writeLines(paste0('\n Estimate a Gini Coefficient on Testing Set = ', 
           formattable::percent( hmeasure::HMeasure(true.class = ObClass %>% as.integer() - 1 ,
                                 scores = Probs)[['metrics']] %>% .[1, 'Gini'], 2 )))

## 
##  Estimate a Gini Coefficient on Testing Set = 31.79%

# Compute the ROC data for the default number of thresholds 
rxRocObject <- 
  RevoScaleR::rxRocCurve(actualVarName = 'ObClass'
                       , predVarNames = c('Probs')
                       , numBreaks = 100 # length(Probs)
                       , data = data.frame(ObClass = ObClass %>% as.integer() - 1, Probs)
                       , title = 'ROC Curve for Logit Model')

# Testing a New Data by Generalized Linear Model
# data.frame(..._fct = '4') %>%
#   mutate_if(is.character, as.factor) %>%
#     predict(glmFit, newdata = ., type = 'response')

# openxlsx::addWorksheet(wb0 <- openxlsx::createWorkbook(), sheetName = 'Example', gridLines = FALSE)
# openxlsx::writeData(wb0, sheet = 1, x = data.frame(X[inTest, ], ObClass, PrClass), withFilter = TRUE); openxlsx::openXL(wb0)

remove(varsel, scope)

# Generate an ROC Curve for the Best Model
ROCCurveShow(Preds = Probs,                          # `Good` Class Probabilities - numeric vector
             Obsers = ObClass,                       # Observed Classes (Reference) - logical vector
             NameOfModel = 'Initial GLM')            # Name Of The Model

## 
## Initial GLM - Estimate a Area Under the ROC Curve (AUC) on Testing Set = 65.90%

# Generate a KS Curve for the Best Model
KSCurveShow(Preds = Probs,                           # `Good` Class Probabilities - numeric vector
            Obsers = ObClass,                        # Observed Classes (Reference) - logical vector
            NameOfModel = 'Initial GLM')             # Name Of The Model

## 
## Initial GLM - Estimate a Kolmogorov-Smirnov Statistic on Testing Set = 0.2331

# Generate a Distribution's Curve of Scores by Good & Bad Class for the Best Model
ScoresCurveShow(Preds = Probs,                       # `Good` Class Probabilities - numeric vector
                Obsers = ObClass,                    # Observed Classes (Reference) - logical vector
                NameOfModel = 'Initial GLM')         # Name Of The Model

## 
## Initial GLM - Estimate a Kullback-Leibler’s Divergence Statistic on Testing Set = 0.0842

remove(rxLogitFit, rxRocObject, Pred, Probs, PrClass, ObClass)

Descriptive mAchine Learning EXplanations of Models

Machine Learning Models are widely used and have various applications in classification or regression tasks. Due to increasing computational power, availability of new data sources and new methods, ML models are more and more complex. Models created with techniques like boosting, bagging of neural networks are true black boxes. It is hard to trace the link between input variables and model outcomes. They are use because of high performance, but lack of interpretability is one of their weakest sides.

# See https://github.com/pbiecek/DALEX, https://github.com/MI2DataLab/modelDown
# See www.r-bloggers.com/dalex-and-h2o-machine-learning-model-interpretability-and-feature-explanation/
library('iml')              # Interpretable Machine Learning
library('DALEX')            # Descriptive mAchine Learning EXplanations
library('breakDown')        # Model Agnostic Explainers for Individual Predictions

Prob_fun <- function(object, newdata){
  # predict(object, newdata=newdata, type = 'prob')[, 'Good']
  
  rxPredict(modelObject = object, data = newdata, blocksPerRead = 200000,
        reportProgress = 0, verbose = 0) %>%
    dplyr::select(starts_with('Probability.')) %>%
      pull 
}

loss_gini <- function(observed, predicted) {
    hmeasure::HMeasure(true.class = observed, scores = predicted)[['metrics']] %>% .[1, 'Gini']
}

Predict.Fun <- function(model, newdata){
    rxPredict(modelObject = model, data = newdata, blocksPerRead = 200000,
        reportProgress = 0, verbose = 0) %>%
      dplyr::select(starts_with('Probability.')) %>%
          pull 
}

loss_Gini <- function(actual, predicted) {
    hmeasure::HMeasure(true.class = actual, scores = predicted)[['metrics']] %>% .[1, 'Gini']
}

Scorecard <- 'Scorecard'

# Create Model Explainer from 'DALEX' package
explainer_classif_1 <-
  DALEX::explain(TheBestModel, label = NameOfTheBestModel,
                 data = X[inTest, ], y = Y[inTest] %>% as.integer()-1, predict_function = Prob_fun)

explainer_classif_2 <-
  DALEX::explain(TheBestModel, label = NameOfTheSecondModel, 
                 data = X[inTest, ], y = Y[inTest] %>% as.integer()-1, predict_function = Prob_fun)

explainer_classif_3 <-
  DALEX::explain(get(paste0(NameOfTheThirdModel, 'Fit')),  label = NameOfTheThirdModel, 
                 data = X[inTest, ], y = Y[inTest] %>% as.integer()-1, predict_function = Prob_fun)

if (!exists('wb')) {
  # Create MS Excel File for Output
  openxlsx::addWorksheet(wb <- openxlsx::createWorkbook(), sheetName = 'IV Table', 
                         gridLines = FALSE, tabColour = 'olivedrab')
  openxlsx::addWorksheet(wb, sheetName = 'Scorecard', gridLines = FALSE, tabColour = 'brown')
}

if (ncol(X) <= Max_Vars) {

# [Interpretable Machine Learning: Feature Importance](https://christophm.github.io/interpretable-ml-book)
  system.time({
    ### Setup parallel processing - 4 times faster
    library('doParallel'); cl <- makeCluster(detectCores()); registerDoParallel(cl)

    model1 <- iml::Predictor$new(model = TheBestModel, data= X[inTest, ],
                y = Y[inTest] %>% as.integer() - 1, class = 'Good', predict.fun = Predict.Fun, type = 'prob' )
    # Feature Importance
    imp1 <- iml::FeatureImp$new(model1, loss = loss_Gini, compare = 'difference', parallel = TRUE)
    
    importance.df <- imp1$results %>%
      dplyr::mutate(Variable = gsub('_fct', '', feature) %>% as.factor,
                    Importance = -importance) %>%
        dplyr::select(Variable, Importance)
    
    p1 <- imp1$plot() + ggplot2::ggtitle(paste(NameOfTheBestModel, 'by Gini coefficient'))
    print(p1)    
    
    # model2 <- iml::Predictor$new(model = get(paste0(NameOfTheSecondModel, 'Fit')), data= X[inTest, ],
    #           y = Y[inTest] %>% as.integer()-1, class = 'Good', predict.fun = Predict.Fun, type = 'prob' )
    # # Feature Importance
    # imp2 <- iml::FeatureImp$new(model2, loss = loss_Gini, compare = 'difference', parallel = TRUE)
    # imp2$plot() + ggplot2::ggtitle(paste(NameOfTheSecondModel, 'by Gini coefficient'))
    # 
    # model3 <- iml::Predictor$new(model = get(paste0(NameOfTheThirdModel, 'Fit')), data= X[inTest, ],
    #           y = Y[inTest] %>% as.integer()-1, class = 'Good', predict.fun = Predict.Fun, type = 'prob' )
    # # Feature Importance
    # imp3 <- iml::FeatureImp$new(model3, loss = loss_Gini, compare = 'difference', parallel = TRUE)
    # imp3$plot() + ggplot2::ggtitle(paste(NameOfTheThirdModel, 'by Gini coefficient'))
    
    TheImportancePredictor <- imp1$results[nrow(imp1$results), 'feature']
    
    # Remember to stop the cluster in the end again
    stopCluster(cl)
  })
    
    # # Choice of The Importance Predictor - www.machinelearningplus.com/machine-learning/feature-selection/
    # TheImportancePredictor <-
    #   imp1$results %>% arrange(importance) %>% dplyr::select(feature) %>% pull %>% .[1]
    # 
    # # Effect of features on the model predictions
    # ale = iml::FeatureEffect$new(model, feature = TheImportancePredictor, method = 'ale')
    # ale$plot()
} else {
  system.time({
    # # Model Performance from 'DALEX' package
    # plot(DALEX::model_performance(explainer_classif_1)
    #    , DALEX::model_performance(explainer_classif_2)
    #    , DALEX::model_performance(explainer_classif_3)
    #    , geom = 'boxplot')
    
    # Variable importance
    importance.df <- DALEX::variable_importance(explainer_classif_1, type = 'raw', loss_function = loss_gini) %>%       dplyr::mutate(Importance = -(dropout_loss - .[variable == '_full_model_', 'dropout_loss'])) %>% 
        dplyr::filter(!str_detect(variable, 'full_model|baseline')) %>%
          dplyr::mutate(Variable = gsub('_fct', '', variable)) %>% 
            dplyr::select(Variable, Importance)
    
    p1 <- DALEX::variable_importance(explainer_classif_1, type = 'raw', loss_function = loss_gini) %>% 
      plot() + labs(title = 'Variable Importance', caption = 'by Gini Coefficient') + theme_grey()
    print(p1)
    
    # DALEX::variable_importance(explainer_classif_2, type = 'raw', loss_function = loss_gini) %>% 
    #   plot() + labs(subtitle = 'Variable Importance', caption = 'by Gini Coefficient')
    # DALEX::variable_importance(explainer_classif_3, type = 'raw', loss_function = loss_gini) %>% 
    #   plot() + labs(subtitle = 'Variable Importance', caption = 'by Gini Coefficient')
    
    TheImportancePredictor <-
      DALEX::variable_importance(explainer_classif_1, type = 'raw')[2, 'variable'] %>%
        as.character
  })
  } # End if ncol(X) <= Max_Vars

##    user  system elapsed 
##    0.81    0.39  116.79

# Output Chart: EXplanations of Model by Variable Importance
openxlsx::insertPlot(wb, sheet = 'IV Table', xy = c(1, nrow(binning.df)+5), width = 10 * (1 + sqrt(5)) / 2,
                     height = 10, units = 'cm')

if (TheBestModel$Description != 'LogisticRegression')
  openxlsx::insertPlot(wb, sheet = Scorecard, xy = c(1,2), width = 10*(1 + sqrt(5))/2, height = 10, units = 'cm')

# # compute Partial Dependence Plots for a given variable --> uses the pdp package
# plot(DALEX::variable_response(explainer_classif_1, variable = TheImportancePredictor, type = 'factor')) +
#     ggtitle('Marginal Response for a Single Variable by Gini Coefficient')
# plot(DALEX::variable_response(explainer_classif_2, variable = TheImportancePredictor, type = 'factor')) +
#     ggtitle('Marginal Response for a Single Variable by Gini Coefficient')
# plot(DALEX::variable_response(explainer_classif_3, variable = TheImportancePredictor, type = 'factor')) +
#     ggtitle('Marginal Response for a Single Variable by Gini Coefficient')            

if (ncol(X) <= Max_Vars) {
  # Explanations for a Single Prediction for Observations
  # True Good Observation
   new_obj <- data.frame(Probs = preds[[NameOfTheBestModel]], Obs = Y[inTest],
                Equal = ifelse(preds[[NameOfTheBestModel]] < 0.5, 0, 1) == Y[inTest] %>% as.integer() - 1) %>%
              rowid_to_column %>% dplyr::arrange(-Probs) %>%
                dplyr::filter(Probs <= 1, Obs == 'Good', Equal == TRUE) %>%
                  dplyr::select(rowid) %>% pull %>% .[1]
  
  # Recovery of Name of Selecled Scale Variables
  pdp1 <- DALEX::prediction_breakdown(explainer_classif_1, observation = X[inTest, ][new_obj, ])
  pdp1$variable <- pdp1$variable_value
  pdp1[1, 'variable'] <- '(Intercept)'; pdp1[nrow(pdp1), 'variable'] <- 'final_prognosis'
  
  # pdp2 <- DALEX::prediction_breakdown(explainer_classif_2, observation = X[inTest, ][new_obj, ])
  # pdp2$variable <- pdp2$variable_value
  # pdp2[1, 'variable'] <- '(Intercept)'; pdp2[nrow(pdp2), 'variable'] <- 'final_prognosis'
  # 
  # pdp3 <- DALEX::prediction_breakdown(explainer_classif_3, observation = X[inTest, ][new_obj, ])
  # pdp3$variable <- pdp3$variable_value
  # pdp3[1, 'variable'] <- '(Intercept)'; pdp3[nrow(pdp3), 'variable'] <- 'final_prognosis'
  
  p2 <- plot(pdp1, vcolors = c('-1' = 'tomato3', '0' = '#f5f5f5', '1' = 'palegreen3', 'X' = '#00BFC4')) +
    # theme(strip.background = element_rect(fill = 'gray45')) + 
    # theme(strip.text = element_text(colour = 'white')) +
    theme_grey() + theme(legend.position = 'none', panel.border = element_blank()) +
    labs(title = paste0('Reference (', new_obj, ' observation) = True ', Y[inTest][ new_obj ]))
  # plot(pdp2) +
  #   labs(subtitle = paste0('Reference (', new_obj, ' observation) = True ', Y[inTest][ new_obj ]))
  # plot(pdp3) +
  #   labs(subtitle = paste0('Reference (', new_obj, ' observation) = True ', Y[inTest][ new_obj ]))
  print(p2)

  # Output Chart: EXplanations of Model by Predictors of True Bad Observation
  openxlsx::insertPlot(wb, sheet = 'IV Table', xy = c(7, nrow(binning.df) + 5), width = 10 * (1 + sqrt(5)) / 2,
                       height = 10, units = 'cm')
  
  if (TheBestModel$Description != 'LogisticRegression')
    openxlsx::insertPlot(wb, sheet = Scorecard, xy=c(1, 21), width = 10*(1 + sqrt(5))/2, height=10, units = 'cm')

  # True Bad Observation

  # Choice Another Level of The Importance Predictor for True Bad Observation than True Good Observation
  AnotherLevel4TrueBad <- table(X[inTest, TheImportancePredictor], Y[inTest])[-which.max(table(X[inTest,
    TheImportancePredictor], Y[inTest])[, 'Good']), 'Bad'] %>%
      which.max %>% names
  
  if (is.null(AnotherLevel4TrueBad)) {
    AnotherLevel4TrueBad <- table(X[inTest, TheImportancePredictor], Y[inTest]) %>%
        as.data.frame(stringsAsFactors = FALSE) %>% 
          setNames(c('Levels', 'Classes', 'Freq')) %>%
            dplyr::filter(Classes == 'Bad', Freq > 0) %>%
              dplyr::select(Levels) %>% pull    
  }
  
  new_obj <- data.frame(Probs = preds[[NameOfTheBestModel]], Obs = Y[inTest],
    Equal = ifelse(preds[[NameOfTheBestModel]] < 0.5, 0, 1) == Y[inTest] %>% as.integer() - 1,
    TheImportancePredictor = X[inTest, TheImportancePredictor]) %>%
      rowid_to_column %>% dplyr::arrange(-Probs) %>%
        dplyr::filter(Probs < 0.4, Obs == 'Bad', Equal == TRUE,
                      TheImportancePredictor == AnotherLevel4TrueBad) %>% 
          dplyr::select(rowid) %>% pull %>% .[1]
  
  pdp1 <- DALEX::prediction_breakdown(explainer_classif_1, observation = X[inTest, ][new_obj, ])
  pdp1$variable <- pdp1$variable_value
  pdp1[1, 'variable'] <- '(Intercept)'; pdp1[nrow(pdp1), 'variable'] <- 'final_prognosis'
  
  # pdp2 <- DALEX::prediction_breakdown(explainer_classif_2, observation = X[inTest, ][new_obj, ])
  # pdp2$variable <- pdp2$variable_value
  # pdp2[1, 'variable'] <- '(Intercept)'; pdp2[nrow(pdp2), 'variable'] <- 'final_prognosis'
  # 
  # pdp3 <- DALEX::prediction_breakdown(explainer_classif_3, observation = X[inTest, ][new_obj, ])
  # pdp3$variable <- pdp3$variable_value
  # pdp3[1, 'variable'] <- '(Intercept)'; pdp3[nrow(pdp3), 'variable'] <- 'final_prognosis'
  
  p3 <- plot(pdp1, vcolors = c('-1' = 'tomato3', '0' = '#f5f5f5', '1' = 'palegreen3', 'X' = '#F8766D')) +
    # theme(strip.background = element_rect(fill = 'gray45')) + 
    # theme(strip.text = element_text(colour = 'white')) +
    theme_grey() + theme(legend.position = 'none', panel.border = element_blank()) +
    labs(title = paste0('Reference (', new_obj, ' observation) = True ', Y[inTest][ new_obj ]))
  # plot(pdp2) +
  #   labs(subtitle = paste0('Reference (', new_obj, ' observation) = True ', Y[inTest][ new_obj ]))
  # plot(pdp3) +
  #   labs(subtitle = paste0('Reference (', new_obj, ' observation) = True ', Y[inTest][ new_obj ]))
  print(p3)
  
  # Output Chart: EXplanations of Model by Predictors of True Bad Observation
  openxlsx::insertPlot(wb, sheet = 'IV Table', xy = c(15, nrow(binning.df) + 5), width = 10 * (1+sqrt(5)) / 2,
                       height = 10, units = 'cm')
  
  if (TheBestModel$Description != 'LogisticRegression')
    openxlsx::insertPlot(wb, sheet = Scorecard, xy=c(1, 40), width = 10*(1 + sqrt(5))/2, height=10, units = 'cm')  }