Credit Risk Prediction
Carolin
2024-12-31
1 Latar Belakang
Membangun model yang dapat memprediksi credit risk menggunakan dataset yang disediakan oleh company yang terdiri dari data pinjaman yang diterima dan yang ditolak. Dilakukan dengan pengerjaan end-to-end solution.
Tujuan utama adalah untuk melakukan prediksi terhadap suatu individu akan kemampuan mereka untuk melakukan pembayaran terhadap pinjaman/kredit yang diberikan.
setwd("D:/Internship/IDX Partner/Final Project")
# df <- read.csv("loan_data_2007_2014.csv")library(feather)## Warning: package 'feather' was built under R version 4.3.2# feather_filename <- "loan_data_2007_2014.feather"
# if (!file.exists(feather_filename)) {
# # Convert to feather
# write_feather(df, "loan_data_2007_2014.feather")
# }
df <- read_feather("loan_data_2007_2014.feather")head(df,5)## # A tibble: 5 × 75
## X id member_id loan_amnt funded_amnt funded_amnt_inv term int_rate
## <int> <int> <int> <int> <int> <dbl> <chr> <dbl>
## 1 0 1077501 1296599 5000 5000 4975 " 36 m… 10.6
## 2 1 1077430 1314167 2500 2500 2500 " 60 m… 15.3
## 3 2 1077175 1313524 2400 2400 2400 " 36 m… 16.0
## 4 3 1076863 1277178 10000 10000 10000 " 36 m… 13.5
## 5 4 1075358 1311748 3000 3000 3000 " 60 m… 12.7
## # ℹ 67 more variables: installment <dbl>, grade <chr>, sub_grade <chr>,
## # emp_title <chr>, emp_length <chr>, home_ownership <chr>, annual_inc <dbl>,
## # verification_status <chr>, issue_d <chr>, loan_status <chr>,
## # pymnt_plan <chr>, url <chr>, desc <chr>, purpose <chr>, title <chr>,
## # zip_code <chr>, addr_state <chr>, dti <dbl>, delinq_2yrs <dbl>,
## # earliest_cr_line <chr>, inq_last_6mths <dbl>, mths_since_last_delinq <dbl>,
## # mths_since_last_record <dbl>, open_acc <dbl>, pub_rec <dbl>, …colnames(df)## [1] "X" "id"
## [3] "member_id" "loan_amnt"
## [5] "funded_amnt" "funded_amnt_inv"
## [7] "term" "int_rate"
## [9] "installment" "grade"
## [11] "sub_grade" "emp_title"
## [13] "emp_length" "home_ownership"
## [15] "annual_inc" "verification_status"
## [17] "issue_d" "loan_status"
## [19] "pymnt_plan" "url"
## [21] "desc" "purpose"
## [23] "title" "zip_code"
## [25] "addr_state" "dti"
## [27] "delinq_2yrs" "earliest_cr_line"
## [29] "inq_last_6mths" "mths_since_last_delinq"
## [31] "mths_since_last_record" "open_acc"
## [33] "pub_rec" "revol_bal"
## [35] "revol_util" "total_acc"
## [37] "initial_list_status" "out_prncp"
## [39] "out_prncp_inv" "total_pymnt"
## [41] "total_pymnt_inv" "total_rec_prncp"
## [43] "total_rec_int" "total_rec_late_fee"
## [45] "recoveries" "collection_recovery_fee"
## [47] "last_pymnt_d" "last_pymnt_amnt"
## [49] "next_pymnt_d" "last_credit_pull_d"
## [51] "collections_12_mths_ex_med" "mths_since_last_major_derog"
## [53] "policy_code" "application_type"
## [55] "annual_inc_joint" "dti_joint"
## [57] "verification_status_joint" "acc_now_delinq"
## [59] "tot_coll_amt" "tot_cur_bal"
## [61] "open_acc_6m" "open_il_6m"
## [63] "open_il_12m" "open_il_24m"
## [65] "mths_since_rcnt_il" "total_bal_il"
## [67] "il_util" "open_rv_12m"
## [69] "open_rv_24m" "max_bal_bc"
## [71] "all_util" "total_rev_hi_lim"
## [73] "inq_fi" "total_cu_tl"
## [75] "inq_last_12m"Deskripsi data
library(readxl)
desc <- read_xlsx("LCDataDictionary.xlsx", sheet = "LoanStats")## New names:
## • `` -> `...3`desc## # A tibble: 98 × 3
## Feature Description ...3
## <chr> <chr> <chr>
## 1 _rec The total amount committed by investors for… Juml…
## 2 acc_now_delinq The number of accounts on which the borrowe… Juml…
## 3 addr_state The state provided by the borrower in the l… Nega…
## 4 all_util Balance to credit limit on all trades Sald…
## 5 annual_inc The self-reported annual income provided by… v
## 6 annual_inc_joint The combined self-reported annual income pr… Peng…
## 7 application_type Indicates whether the loan is an individual… <NA>
## 8 collection_recovery_fee post charge off collection fee Biay…
## 9 collections_12_mths_ex_med Number of collections in 12 months excludin… Juml…
## 10 delinq_2yrs The number of 30+ days past-due incidences … Juml…
## # ℹ 88 more rowslibrary(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionglimpse(df)## Rows: 466,285
## Columns: 75
## $ X <int> 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, …
## $ id <int> 1077501, 1077430, 1077175, 1076863, 107535…
## $ member_id <int> 1296599, 1314167, 1313524, 1277178, 131174…
## $ loan_amnt <int> 5000, 2500, 2400, 10000, 3000, 5000, 7000,…
## $ funded_amnt <int> 5000, 2500, 2400, 10000, 3000, 5000, 7000,…
## $ funded_amnt_inv <dbl> 4975.00, 2500.00, 2400.00, 10000.00, 3000.…
## $ term <chr> " 36 months", " 60 months", " 36 months", …
## $ int_rate <dbl> 10.65, 15.27, 15.96, 13.49, 12.69, 7.90, 1…
## $ installment <dbl> 162.87, 59.83, 84.33, 339.31, 67.79, 156.4…
## $ grade <chr> "B", "C", "C", "C", "B", "A", "C", "E", "F…
## $ sub_grade <chr> "B2", "C4", "C5", "C1", "B5", "A4", "C5", …
## $ emp_title <chr> "", "Ryder", "", "AIR RESOURCES BOARD", "U…
## $ emp_length <chr> "10+ years", "< 1 year", "10+ years", "10+…
## $ home_ownership <chr> "RENT", "RENT", "RENT", "RENT", "RENT", "R…
## $ annual_inc <dbl> 24000.00, 30000.00, 12252.00, 49200.00, 80…
## $ verification_status <chr> "Verified", "Source Verified", "Not Verifi…
## $ issue_d <chr> "Dec-11", "Dec-11", "Dec-11", "Dec-11", "D…
## $ loan_status <chr> "Fully Paid", "Charged Off", "Fully Paid",…
## $ pymnt_plan <chr> "n", "n", "n", "n", "n", "n", "n", "n", "n…
## $ url <chr> "https://www.lendingclub.com/browse/loanDe…
## $ desc <chr> " Borrower added on 12/22/11 > I need to …
## $ purpose <chr> "credit_card", "car", "small_business", "o…
## $ title <chr> "Computer", "bike", "real estate business"…
## $ zip_code <chr> "860xx", "309xx", "606xx", "917xx", "972xx…
## $ addr_state <chr> "AZ", "GA", "IL", "CA", "OR", "AZ", "NC", …
## $ dti <dbl> 27.65, 1.00, 8.72, 20.00, 17.94, 11.20, 23…
## $ delinq_2yrs <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ earliest_cr_line <chr> "Jan-85", "Apr-99", "Nov-01", "Feb-96", "J…
## $ inq_last_6mths <dbl> 1, 5, 2, 1, 0, 3, 1, 2, 2, 0, 2, 0, 1, 2, …
## $ mths_since_last_delinq <dbl> NA, NA, NA, 35, 38, NA, NA, NA, NA, NA, NA…
## $ mths_since_last_record <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_acc <dbl> 3, 3, 2, 10, 15, 9, 7, 4, 11, 2, 14, 12, 4…
## $ pub_rec <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ revol_bal <int> 13648, 1687, 2956, 5598, 27783, 7963, 1772…
## $ revol_util <dbl> 83.70, 9.40, 98.50, 21.00, 53.90, 28.30, 8…
## $ total_acc <dbl> 9, 4, 10, 37, 38, 12, 11, 4, 13, 3, 23, 34…
## $ initial_list_status <chr> "f", "f", "f", "f", "f", "f", "f", "f", "f…
## $ out_prncp <dbl> 0.00, 0.00, 0.00, 0.00, 766.90, 0.00, 1889…
## $ out_prncp_inv <dbl> 0.00, 0.00, 0.00, 0.00, 766.90, 0.00, 1889…
## $ total_pymnt <dbl> 5861.071, 1008.710, 3003.654, 12226.302, 3…
## $ total_pymnt_inv <dbl> 5831.78, 1008.71, 3003.65, 12226.30, 3242.…
## $ total_rec_prncp <dbl> 5000.00, 456.46, 2400.00, 10000.00, 2233.1…
## $ total_rec_int <dbl> 861.07, 435.17, 603.65, 2209.33, 1009.07, …
## $ total_rec_late_fee <dbl> 0.00, 0.00, 0.00, 16.97, 0.00, 0.00, 0.00,…
## $ recoveries <dbl> 0.00, 117.08, 0.00, 0.00, 0.00, 0.00, 0.00…
## $ collection_recovery_fee <dbl> 0.0000, 1.1100, 0.0000, 0.0000, 0.0000, 0.…
## $ last_pymnt_d <chr> "Jan-15", "Apr-13", "Jun-14", "Jan-15", "J…
## $ last_pymnt_amnt <dbl> 171.62, 119.66, 649.91, 357.48, 67.79, 161…
## $ next_pymnt_d <chr> "", "", "", "", "Feb-16", "", "Feb-16", ""…
## $ last_credit_pull_d <chr> "Jan-16", "Sep-13", "Jan-16", "Jan-15", "J…
## $ collections_12_mths_ex_med <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ mths_since_last_major_derog <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ policy_code <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
## $ application_type <chr> "INDIVIDUAL", "INDIVIDUAL", "INDIVIDUAL", …
## $ annual_inc_joint <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ dti_joint <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ verification_status_joint <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ acc_now_delinq <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ tot_coll_amt <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ tot_cur_bal <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_acc_6m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_il_6m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_il_12m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_il_24m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ mths_since_rcnt_il <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ total_bal_il <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ il_util <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_rv_12m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ open_rv_24m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ max_bal_bc <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ all_util <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ total_rev_hi_lim <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ inq_fi <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ total_cu_tl <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
## $ inq_last_12m <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…Akan dilakukan pemisahan kolom yang bertipe kategorik dan numerik
categorical <- df[,sapply(df, is.character)]
categorical <- lapply(categorical, is.factor)
library(Hmisc)## Warning: package 'Hmisc' was built under R version 4.3.2##
## Attaching package: 'Hmisc'## The following objects are masked from 'package:dplyr':
##
## src, summarize## The following objects are masked from 'package:base':
##
## format.pval, unitsdescribe(df[,names(categorical)])## df[, names(categorical)]
##
## 22 Variables 466285 Observations
## --------------------------------------------------------------------------------
## term
## n missing distinct
## 466285 0 2
##
## Value 36 months 60 months
## Frequency 337953 128332
## Proportion 0.725 0.275
## --------------------------------------------------------------------------------
## grade
## n missing distinct
## 466285 0 7
##
## Value A B C D E F G
## Frequency 74867 136929 125293 76888 35757 13229 3322
## Proportion 0.161 0.294 0.269 0.165 0.077 0.028 0.007
## --------------------------------------------------------------------------------
## sub_grade
## n missing distinct
## 466285 0 35
##
## lowest : A1 A2 A3 A4 A5, highest: G1 G2 G3 G4 G5
## --------------------------------------------------------------------------------
## emp_title
## n missing distinct
## 438697 27588 205475
##
## lowest : 'roduction manager NSA Industries llc MOTHER THERESA HOME Quitman Street Community School Bus Operator and Instructor
## highest: Zynga Inc. Zynga, Inc Zynx Health Zyquest Inc Zytron Control Products, Inc.
## --------------------------------------------------------------------------------
## emp_length
## n missing distinct
## 445277 21008 11
##
## Value < 1 year 1 year 10+ years 2 years 3 years 4 years
## Frequency 36265 29622 150049 41373 36596 28023
## Proportion 0.081 0.067 0.337 0.093 0.082 0.063
##
## Value 5 years 6 years 7 years 8 years 9 years
## Frequency 30774 26112 26180 22395 17888
## Proportion 0.069 0.059 0.059 0.050 0.040
## --------------------------------------------------------------------------------
## home_ownership
## n missing distinct
## 466285 0 6
##
## Value ANY MORTGAGE NONE OTHER OWN RENT
## Frequency 1 235875 50 182 41704 188473
## Proportion 0.000 0.506 0.000 0.000 0.089 0.404
## --------------------------------------------------------------------------------
## verification_status
## n missing distinct
## 466285 0 3
##
## Value Not Verified Source Verified Verified
## Frequency 148237 149993 168055
## Proportion 0.318 0.322 0.360
## --------------------------------------------------------------------------------
## issue_d
## n missing distinct
## 466285 0 91
##
## lowest : Apr-08 Apr-09 Apr-10 Apr-11 Apr-12, highest: Sep-10 Sep-11 Sep-12 Sep-13 Sep-14
## --------------------------------------------------------------------------------
## loan_status
## n missing distinct
## 466285 0 9
##
## lowest : Charged Off Current Default Does not meet the credit policy. Status:Charged Off Does not meet the credit policy. Status:Fully Paid
## highest: Does not meet the credit policy. Status:Fully Paid Fully Paid In Grace Period Late (16-30 days) Late (31-120 days)
## --------------------------------------------------------------------------------
## pymnt_plan
## n missing distinct
## 466285 0 2
##
## Value n y
## Frequency 466276 9
## Proportion 1 0
## --------------------------------------------------------------------------------
## url
## n missing distinct
## 466285 0 466285
##
## lowest : https://www.lendingclub.com/browse/loanDetail.action?loan_id=1000007 https://www.lendingclub.com/browse/loanDetail.action?loan_id=1000030 https://www.lendingclub.com/browse/loanDetail.action?loan_id=1000033 https://www.lendingclub.com/browse/loanDetail.action?loan_id=1000045 https://www.lendingclub.com/browse/loanDetail.action?loan_id=1000067
## highest: https://www.lendingclub.com/browse/loanDetail.action?loan_id=999951 https://www.lendingclub.com/browse/loanDetail.action?loan_id=999955 https://www.lendingclub.com/browse/loanDetail.action?loan_id=999981 https://www.lendingclub.com/browse/loanDetail.action?loan_id=999983 https://www.lendingclub.com/browse/loanDetail.action?loan_id=999989
## --------------------------------------------------------------------------------
## desc
## n missing distinct
## 125749 340536 124435
##
## lowest : - Pay off Dell Financial: $ 1300.00 - Pay off IRS for 2005: $ 1400.00 - Pay off Mac Comp : $ 1700.00 - Pay off Bill Me Later : $ 1000.00 - Pay off WAMU card : $ 1000.00 - Pay off Becu Visa : $ 1000.00 -To pay off loans and Credit Card at 17-25% interest -Daughter College tuition and books -Bring VW BUS 1970 and House Good storage in NM. Storage not covered by US Army when assign to Oversee duty. -Replace Matters for better sleeps Looking to expand seafood buisness. Purchasing a new boat with gear. Market to sell my catch is very good and outlook is good. I also work for sea tow so I always have steady income coming in. I would like to increase my cash availability mainly to potentially be used for emergencies. While my income to expese flow is solvent and I have some spare cash, I would like to protect myself against uncertainty and am willing to take a small loan out for it.
## highest: www.music-housecalls.com--The Amadeus Home Music School will be using the loan to expand our Home Music Lesson business which currently has 250 instructors in 17 states. The expansion will be to add Ontario and other states within the United States of America. This will assist us in paying our first few months of Google charges. With Google we have generated a 5 to 1 return. Our web presence is at the top of Google, Yahoo and MSN in the sponsored links. Please take a look at our work and consider funding our expansion. Yes I have a line of credit that the interest rate just went up to 20.99 so I am looking to refi for a lower rate. Young couple just married. Own home but have two car loans and wife has school loans. Husband never had credit card debt. Was lucky enough that parents paid school college tuition. Husband is Senior Finance Officer at large denominational healthcare facility chain. Wife is an Electronic Medical Record Intermediate Analyst and BSN RN by training. Wife accumulated $15000 in credit card debt during school and throughout her twenties. Attempting to consolidate this debt into a more manageable rate and not just open up another credit card line. Young professional who needs to pay down debt acquired from college. Zensah Performance Apparel is a start-up, and we are now seeking to expand our product line as we enter the midwya point in our 3rd year! We currently sell through our website at www.zensah.com and directly to professional teams. Type ZENSAH in YOU TUBE, and you will see one of the most famous athletes in the world who sports ZENSAH! Zensah has create a proprietary moisture wicking fabric that doesn't carry odor and is not coarse. Unlike Under Armour and Nike you won't chafe with our fabric. We want to try Lending Club even though we are well capitalized. If this works out we thinks it could be a great way to allow our fans helps us grow even more! I encourage you to visit our website to learn more. Thanks for your interest. TEAM ZENSAH
## --------------------------------------------------------------------------------
## purpose
## n missing distinct
## 466285 0 14
##
## lowest : car credit_card debt_consolidation educational home_improvement
## highest: other renewable_energy small_business vacation wedding
## --------------------------------------------------------------------------------
## title
## n missing distinct
## 466265 20 63099
##
## lowest : '08 & '09 Roth IRA Investments '08 Rehab '56 Dream Trailer '71 Bobbed Duece 'Cut the Cards'
## highest: ZMH zonball Loan Zx Zx6r zxcvb
## --------------------------------------------------------------------------------
## zip_code
## n missing distinct
## 466285 0 888
##
## lowest : 007xx 008xx 010xx 011xx 012xx, highest: 995xx 996xx 997xx 998xx 999xx
## --------------------------------------------------------------------------------
## addr_state
## n missing distinct
## 466285 0 50
##
## lowest : AK AL AR AZ CA, highest: VT WA WI WV WY
## --------------------------------------------------------------------------------
## earliest_cr_line
## n missing distinct
## 466256 29 664
##
## lowest : Apr-00 Apr-01 Apr-02 Apr-03 Apr-04, highest: Sep-95 Sep-96 Sep-97 Sep-98 Sep-99
## --------------------------------------------------------------------------------
## initial_list_status
## n missing distinct
## 466285 0 2
##
## Value f w
## Frequency 303005 163280
## Proportion 0.65 0.35
## --------------------------------------------------------------------------------
## last_pymnt_d
## n missing distinct
## 465909 376 98
##
## lowest : Apr-08 Apr-09 Apr-10 Apr-11 Apr-12, highest: Sep-11 Sep-12 Sep-13 Sep-14 Sep-15
## --------------------------------------------------------------------------------
## next_pymnt_d
## n missing distinct
## 239071 227214 100
##
## lowest : Apr-08 Apr-09 Apr-10 Apr-11 Apr-12, highest: Sep-11 Sep-12 Sep-13 Sep-14 Sep-15
## --------------------------------------------------------------------------------
## last_credit_pull_d
## n missing distinct
## 466243 42 103
##
## lowest : Apr-09 Apr-10 Apr-11 Apr-12 Apr-13, highest: Sep-11 Sep-12 Sep-13 Sep-14 Sep-15
## --------------------------------------------------------------------------------
## application_type
## n missing distinct value
## 466285 0 1 INDIVIDUAL
##
## Value INDIVIDUAL
## Frequency 466285
## Proportion 1
## --------------------------------------------------------------------------------numerical <- df[,sapply(df, is.numeric)]
summary(df[,names(numerical)])## X id member_id loan_amnt
## Min. : 0 Min. : 54734 Min. : 70473 Min. : 500
## 1st Qu.:116571 1st Qu.: 3639987 1st Qu.: 4379705 1st Qu.: 8000
## Median :233142 Median :10107897 Median :11941075 Median :12000
## Mean :233142 Mean :13079730 Mean :14597661 Mean :14317
## 3rd Qu.:349713 3rd Qu.:20731209 3rd Qu.:23001541 3rd Qu.:20000
## Max. :466284 Max. :38098114 Max. :40860827 Max. :35000
##
## funded_amnt funded_amnt_inv int_rate installment
## Min. : 500 Min. : 0 Min. : 5.42 Min. : 15.67
## 1st Qu.: 8000 1st Qu.: 8000 1st Qu.:10.99 1st Qu.: 256.69
## Median :12000 Median :12000 Median :13.66 Median : 379.89
## Mean :14292 Mean :14222 Mean :13.83 Mean : 432.06
## 3rd Qu.:20000 3rd Qu.:19950 3rd Qu.:16.49 3rd Qu.: 566.58
## Max. :35000 Max. :35000 Max. :26.06 Max. :1409.99
##
## annual_inc dti delinq_2yrs inq_last_6mths
## Min. : 1896 Min. : 0.00 Min. : 0.0000 Min. : 0.0000
## 1st Qu.: 45000 1st Qu.:11.36 1st Qu.: 0.0000 1st Qu.: 0.0000
## Median : 63000 Median :16.87 Median : 0.0000 Median : 0.0000
## Mean : 73277 Mean :17.22 Mean : 0.2847 Mean : 0.8047
## 3rd Qu.: 88960 3rd Qu.:22.78 3rd Qu.: 0.0000 3rd Qu.: 1.0000
## Max. :7500000 Max. :39.99 Max. :29.0000 Max. :33.0000
## NA's :4 NA's :29 NA's :29
## mths_since_last_delinq mths_since_last_record open_acc
## Min. : 0.0 Min. : 0.0 Min. : 0.00
## 1st Qu.: 16.0 1st Qu.: 53.0 1st Qu.: 8.00
## Median : 31.0 Median : 76.0 Median :10.00
## Mean : 34.1 Mean : 74.3 Mean :11.19
## 3rd Qu.: 49.0 3rd Qu.:102.0 3rd Qu.:14.00
## Max. :188.0 Max. :129.0 Max. :84.00
## NA's :250351 NA's :403647 NA's :29
## pub_rec revol_bal revol_util total_acc
## Min. : 0.0000 Min. : 0 Min. : 0.00 Min. : 1.00
## 1st Qu.: 0.0000 1st Qu.: 6413 1st Qu.: 39.20 1st Qu.: 17.00
## Median : 0.0000 Median : 11764 Median : 57.60 Median : 23.00
## Mean : 0.1606 Mean : 16230 Mean : 56.18 Mean : 25.06
## 3rd Qu.: 0.0000 3rd Qu.: 20333 3rd Qu.: 74.70 3rd Qu.: 32.00
## Max. :63.0000 Max. :2568995 Max. :892.30 Max. :156.00
## NA's :29 NA's :340 NA's :29
## out_prncp out_prncp_inv total_pymnt total_pymnt_inv
## Min. : 0.0 Min. : 0.0 Min. : 0 Min. : 0
## 1st Qu.: 0.0 1st Qu.: 0.0 1st Qu.: 5552 1st Qu.: 5499
## Median : 441.5 Median : 441.4 Median : 9419 Median : 9355
## Mean : 4410.1 Mean : 4408.4 Mean :11541 Mean :11470
## 3rd Qu.: 7341.6 3rd Qu.: 7338.4 3rd Qu.:15308 3rd Qu.:15231
## Max. :32160.4 Max. :32160.4 Max. :57778 Max. :57778
##
## total_rec_prncp total_rec_int total_rec_late_fee recoveries
## Min. : 0 Min. : 0.0 Min. : 0.0000 Min. : 0.00
## 1st Qu.: 3709 1st Qu.: 957.3 1st Qu.: 0.0000 1st Qu.: 0.00
## Median : 6818 Median : 1818.9 Median : 0.0000 Median : 0.00
## Mean : 8866 Mean : 2588.7 Mean : 0.6501 Mean : 85.34
## 3rd Qu.:12000 3rd Qu.: 3304.5 3rd Qu.: 0.0000 3rd Qu.: 0.00
## Max. :35000 Max. :24205.6 Max. :358.6800 Max. :33520.27
##
## collection_recovery_fee last_pymnt_amnt collections_12_mths_ex_med
## Min. : 0.000 Min. : 0.0 Min. : 0.00000
## 1st Qu.: 0.000 1st Qu.: 312.6 1st Qu.: 0.00000
## Median : 0.000 Median : 546.0 Median : 0.00000
## Mean : 8.962 Mean : 3123.9 Mean : 0.00909
## 3rd Qu.: 0.000 3rd Qu.: 3187.5 3rd Qu.: 0.00000
## Max. :7002.190 Max. :36234.4 Max. :20.00000
## NA's :145
## mths_since_last_major_derog policy_code acc_now_delinq tot_coll_amt
## Min. : 0.0 Min. :1 Min. :0.000000 Min. : 0
## 1st Qu.: 26.0 1st Qu.:1 1st Qu.:0.000000 1st Qu.: 0
## Median : 42.0 Median :1 Median :0.000000 Median : 0
## Mean : 42.9 Mean :1 Mean :0.004002 Mean : 192
## 3rd Qu.: 59.0 3rd Qu.:1 3rd Qu.:0.000000 3rd Qu.: 0
## Max. :188.0 Max. :1 Max. :5.000000 Max. :9152545
## NA's :367311 NA's :29 NA's :70276
## tot_cur_bal total_rev_hi_lim
## Min. : 0 Min. : 0
## 1st Qu.: 28618 1st Qu.: 13500
## Median : 81539 Median : 22800
## Mean : 138802 Mean : 30379
## 3rd Qu.: 208953 3rd Qu.: 37900
## Max. :8000078 Max. :9999999
## NA's :70276 NA's :70276library(dplyr)
library(tidyr)
df_null <- df %>%
summarise_all(~ mean(is.na(.))*100) %>%
gather(key = "column", value = "df_null") %>%
arrange(desc(df_null))
df_null## # A tibble: 75 × 2
## column df_null
## <chr> <dbl>
## 1 annual_inc_joint 100
## 2 dti_joint 100
## 3 verification_status_joint 100
## 4 open_acc_6m 100
## 5 open_il_6m 100
## 6 open_il_12m 100
## 7 open_il_24m 100
## 8 mths_since_rcnt_il 100
## 9 total_bal_il 100
## 10 il_util 100
## # ℹ 65 more rows2 Data Cleaning
Menghapus kolom yang tidak diperlukan:
df <- df %>%
select((-one_of(c('X', 'id', 'url', 'desc','member_id', 'url', 'title', 'addr_state', 'zip_code', 'policy_code', 'application_type', 'emp_title'))))Pengecekan Data Duplicated
df[duplicated(df),]## # A tibble: 0 × 64
## # ℹ 64 variables: loan_amnt <int>, funded_amnt <int>, funded_amnt_inv <dbl>,
## # term <chr>, int_rate <dbl>, installment <dbl>, grade <chr>,
## # sub_grade <chr>, emp_length <chr>, home_ownership <chr>, annual_inc <dbl>,
## # verification_status <chr>, issue_d <chr>, loan_status <chr>,
## # pymnt_plan <chr>, purpose <chr>, dti <dbl>, delinq_2yrs <dbl>,
## # earliest_cr_line <chr>, inq_last_6mths <dbl>, mths_since_last_delinq <dbl>,
## # mths_since_last_record <dbl>, open_acc <dbl>, pub_rec <dbl>, …df %>%
distinct() %>%
anti_join(df, by = names(df))## # A tibble: 0 × 64
## # ℹ 64 variables: loan_amnt <int>, funded_amnt <int>, funded_amnt_inv <dbl>,
## # term <chr>, int_rate <dbl>, installment <dbl>, grade <chr>,
## # sub_grade <chr>, emp_length <chr>, home_ownership <chr>, annual_inc <dbl>,
## # verification_status <chr>, issue_d <chr>, loan_status <chr>,
## # pymnt_plan <chr>, purpose <chr>, dti <dbl>, delinq_2yrs <dbl>,
## # earliest_cr_line <chr>, inq_last_6mths <dbl>, mths_since_last_delinq <dbl>,
## # mths_since_last_record <dbl>, open_acc <dbl>, pub_rec <dbl>, …Tidak ada data yang duplicated. Hal ini mengindikasikan setiap kolom telah mewakili satu individu unik.
df %>%
summarise_all(n_distinct, na.rm = TRUE)## # A tibble: 1 × 64
## loan_amnt funded_amnt funded_amnt_inv term int_rate installment grade
## <int> <int> <int> <int> <int> <int> <int>
## 1 1352 1354 9854 2 506 55622 7
## # ℹ 57 more variables: sub_grade <int>, emp_length <int>, home_ownership <int>,
## # annual_inc <int>, verification_status <int>, issue_d <int>,
## # loan_status <int>, pymnt_plan <int>, purpose <int>, dti <int>,
## # delinq_2yrs <int>, earliest_cr_line <int>, inq_last_6mths <int>,
## # mths_since_last_delinq <int>, mths_since_last_record <int>, open_acc <int>,
## # pub_rec <int>, revol_bal <int>, revol_util <int>, total_acc <int>,
## # initial_list_status <int>, out_prncp <int>, out_prncp_inv <int>, …Missing Values
Menghapus kolom yang memiliki percentage missing values > 50%.
df <- df %>%
select(-which(colMeans(is.na(.)) > 0.5))Pada kolom tot_coll_amt, tot_cur_bal, total_rev_hi_lim, terdapat 15% missing values. Penggisian missing values pada kolom ini dilakukan dengan nilai 0. Hal ini dilakukan dengan asumsi bahwa customer tersebut hanya meminjam sekali.
| tot_coll_amt | Total collection amounts ever owed |
| tot_cur_bal | Total current balance of all accounts |
| total_rev_hi_lim | Total revolving high credit/credit limit |
df <- df %>%
mutate(across(c('tot_coll_amt', 'tot_cur_bal', 'total_rev_hi_lim'), ~ifelse(is.na(.), 0, .)))Selanjutnya, pada kolom numerik dilakukan pengisian missing values dengan nilai rata-rata.
df <- df %>%
mutate(across(where(is.numeric), ~ifelse(is.na(.), median(., na.rm = TRUE), .)))Pada kolom kategorik, penggisian missing values dengan nilai modusnya.
# membangun fungsi modus terlebih dahulu
mode <- function(x, na.rm = FALSE) {
if (na.rm) {
x <- x[!is.na(x)]
}
ux <- unique(x)
ux[which.max(tabulate(match(x, ux)))]
}# Mendapatkan nama-nama kolom kategorikal
categorical_columns <- colnames(df[, sapply(df, function(x) is.factor(x) | is.character(x))])
# Mengisi nilai yang hilang dengan modus pada kolom kategorikal
df <- df %>%
mutate(across(all_of(categorical_columns), ~ifelse(is.na(.), mode(., na.rm = TRUE), .)))# df %>%
# summarise(across(everything(), ~sum(is.na(.))))
colSums(is.na(df))## loan_amnt funded_amnt
## 0 0
## funded_amnt_inv term
## 0 0
## int_rate installment
## 0 0
## grade sub_grade
## 0 0
## emp_length home_ownership
## 0 0
## annual_inc verification_status
## 0 0
## issue_d loan_status
## 0 0
## pymnt_plan purpose
## 0 0
## dti delinq_2yrs
## 0 0
## earliest_cr_line inq_last_6mths
## 0 0
## open_acc pub_rec
## 0 0
## revol_bal revol_util
## 0 0
## total_acc initial_list_status
## 0 0
## out_prncp out_prncp_inv
## 0 0
## total_pymnt total_pymnt_inv
## 0 0
## total_rec_prncp total_rec_int
## 0 0
## total_rec_late_fee recoveries
## 0 0
## collection_recovery_fee last_pymnt_d
## 0 0
## last_pymnt_amnt next_pymnt_d
## 0 0
## last_credit_pull_d collections_12_mths_ex_med
## 0 0
## acc_now_delinq tot_coll_amt
## 0 0
## tot_cur_bal total_rev_hi_lim
## 0 0Semua kolom tidak lagi memiliki missing values.
3 Labelling
Dalam project credit risk modeling, tujuan utama adalah untuk melakukan prediksi terhadap suatu individu akan kemampuan mereka untuk melakukan pembayaran terhadap pinjaman/kredit yang diberikan. Oleh karena itu, variabel target yang digunakan harus mencerminkan kemampuan individu dalam hal tersebut.
Dalam dataset ini, variabel loan_status adalah variabel
yang dapat dijadikan variabel target karena mencerminkan performa
masing-masing individu dalam melakukan pembayaran terhadap
pinjaman/kredit selama ini.
library(ggplot2)
df %>%
ggplot(aes(y = reorder(loan_status, -table(loan_status)[loan_status]))) +
geom_bar() +
geom_text(stat = "count",aes(label = scales::percent(
..count../sum(after_stat(count))),
x = ..count..),
position = position_dodge(width = 0.9),
hjust = -0.1)+
labs(y = "Loan Status",x = "Count")+
theme_bw() +
theme(
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank()
)## Warning: The dot-dot notation (`..count..`) was deprecated in ggplot2 3.4.0.
## ℹ Please use `after_stat(count)` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
prop.table(table(df$loan_status))*100##
## Charged Off
## 9.1092358
## Current
## 48.0877575
## Default
## 0.1784316
## Does not meet the credit policy. Status:Charged Off
## 0.1632049
## Does not meet the credit policy. Status:Fully Paid
## 0.4263487
## Fully Paid
## 39.6193315
## In Grace Period
## 0.6746947
## Late (16-30 days)
## 0.2612136
## Late (31-120 days)
## 1.4797817Pada kolom loan_status memiliki beberapa nilai:
Current artinya pembayaran lancar;
Charged Off artinya pembayaran macet sehingga
dihapusbukukan; Late artinya pembayaran telat dilakukan;
In Grace Period artinya dalam masa tenggang;
Fully Paid artinya pembayaran lunas; Default
artinya pembayaran macet
Dari definisi-definisi tersebut, masing-masing individu dapat
ditandai apakah mereka merupakan bad loan (peminjam yang
buruk) atau good loan (peminjam yang baik)
Definisi bad dan good loan terkadang bisa berbeda tergantung dari kebutuhan bisnis. Dalam project ini akan menggunakan keterlambatan pembayaran di atas 30 hari dan yang lebih buruk dari itu sebagai penanda bad loan.
library(dplyr)
dfa <- df # untuk backup data
df <- df %>%
mutate(bad_flag = ifelse(loan_status %in% c('Charged Off', 'Default', 'Does not meet the credit policy. Status:Charged Off', 'Late (31-120 days)'), 1, 0)) %>%
select(-loan_status)prop.table(table(df$bad_flag))*100##
## 0 1
## 89.06935 10.93065library(ggplot2)
df %>%
ggplot(aes(x = factor(bad_flag))) +
geom_bar(aes(y = ..count.., fill = factor(bad_flag)),
position = "dodge") +
geom_text(stat = "count",aes(label = scales::percent(
..count../sum(..count..)),
y = ..count.., group = factor(bad_flag)),
position = position_dodge(width = 0.9),
vjust = -0.5)+
scale_y_continuous(breaks = seq(0, 300000, by = 50000))+
labs(title = "Proportion of Clients Loan Status",
x = "Loan Status", y = "Count") +
scale_fill_manual(values = c("skyblue", "pink"),
name = "Loan Status",
labels = c("Good", "Bad")) +
scale_x_discrete(labels = c("0" = "Good", "1" = "Bad")) +
theme_bw()
Diperoleh imbalanced dataset problem setelah melakukan labelling terhadap bad/good loan.
4 Feature Enggineering
4.1 emp_length & term
table(unique(df$emp_length))##
## < 1 year 1 year 10+ years 2 years 3 years 4 years 5 years
## 1 1 1 1 1 1 1 1
## 6 years 7 years 8 years 9 years
## 1 1 1 1unique(df$term)## [1] " 36 months" " 60 months"df <- df %>%
mutate(emp_length = as.numeric(gsub("\\+ years| years| year", "", gsub("< 1 year", "0", emp_length)))) %>%
mutate(term = as.numeric(gsub("months", "", term)))sum(is.na(df$emp_length))## [1] 21008df <- df %>%
mutate(across(where(is.numeric), ~ifelse(is.na(.), median(., na.rm = TRUE), .)))sum(is.na(df$emp_length))## [1] 04.2 variabel tanggal
| earliest_cr_line | The month the borrower’s earliest reported credit line was opened | Bulan jalur kredit yang paling awal yang dilaporkan peminjam dibuka |
| issue_d | The month which the loan was funded | Bulan yang didanai pinjaman |
| last_pymnt_d | Last month payment was received | Bulan lalu pembayaran diterima |
| next_pymnt_d | Next scheduled payment date | Tanggal Pembayaran Terjadwal Berikutnya |
library(lubridate)##
## Attaching package: 'lubridate'## The following objects are masked from 'package:base':
##
## date, intersect, setdiff, uniondate_columns <- c('earliest_cr_line', 'last_credit_pull_d', 'last_pymnt_d', 'issue_d', 'next_pymnt_d')
df<- df %>%
mutate(across(all_of(date_columns),
~as.Date(parse_date_time(., orders = '%b-%y'))))
class(df$next_pymnt_d)## [1] "Date"class(df$issue_d)## [1] "Date"head(df[, c('earliest_cr_line', 'last_credit_pull_d', 'last_pymnt_d', 'issue_d', 'next_pymnt_d')])## # A tibble: 6 × 5
## earliest_cr_line last_credit_pull_d last_pymnt_d issue_d next_pymnt_d
## <date> <date> <date> <date> <date>
## 1 1985-01-01 2016-01-01 2015-01-01 2011-12-01 NA
## 2 1999-04-01 2013-09-01 2013-04-01 2011-12-01 NA
## 3 2001-11-01 2016-01-01 2014-06-01 2011-12-01 NA
## 4 1996-02-01 2015-01-01 2015-01-01 2011-12-01 NA
## 5 1996-01-01 2016-01-01 2016-01-01 2011-12-01 2016-02-01
## 6 2004-11-01 2015-09-01 2015-01-01 2011-12-01 NA# mengisi nilai kosong pada kolom tangga dengan modus
df <- df %>%
mutate(across(all_of(date_columns), ~replace_na(as.Date(., format = "%Y-%m-%d"), mode(as.Date(.[!is.na(.)], format = "%Y-%m-%d")))))
class(df$issue_d)## [1] "Date"df[, c('earliest_cr_line', 'last_credit_pull_d', 'last_pymnt_d', 'issue_d', 'next_pymnt_d')]## # A tibble: 466,285 × 5
## earliest_cr_line last_credit_pull_d last_pymnt_d issue_d next_pymnt_d
## <date> <date> <date> <date> <date>
## 1 1985-01-01 2016-01-01 2015-01-01 2011-12-01 2016-02-01
## 2 1999-04-01 2013-09-01 2013-04-01 2011-12-01 2016-02-01
## 3 2001-11-01 2016-01-01 2014-06-01 2011-12-01 2016-02-01
## 4 1996-02-01 2015-01-01 2015-01-01 2011-12-01 2016-02-01
## 5 1996-01-01 2016-01-01 2016-01-01 2011-12-01 2016-02-01
## 6 2004-11-01 2015-09-01 2015-01-01 2011-12-01 2016-02-01
## 7 2005-07-01 2016-01-01 2016-01-01 2011-12-01 2016-02-01
## 8 2007-01-01 2014-12-01 2015-01-01 2011-12-01 2016-02-01
## 9 2004-04-01 2012-08-01 2012-04-01 2011-12-01 2016-02-01
## 10 2004-09-01 2013-03-01 2012-11-01 2011-12-01 2016-02-01
## # ℹ 466,275 more rowsSelanjutnya ditambahkan variabel payment_time yang berisi jangka waktu pembayaran loan dan variabel credit_time sebagai jangka waktu credit telah berlangsung.
df<- df %>%
mutate(payment_time = mapply(function(d1, d2) (year(d1) - year(d2)) * 12 + (month(d1) - month(d2)), next_pymnt_d, last_pymnt_d),
credit_time = mapply(function(d1, d2) year(d1) - year(d2), last_credit_pull_d, earliest_cr_line),
loan_time = mapply(function(d1, d2) (year(d1) - year(d2)) * 12 + (month(d1) - month(d2)), next_pymnt_d, issue_d))
head(df[,c('last_pymnt_d', 'next_pymnt_d', 'payment_time')])## # A tibble: 6 × 3
## last_pymnt_d next_pymnt_d payment_time
## <date> <date> <dbl>
## 1 2015-01-01 2016-02-01 13
## 2 2013-04-01 2016-02-01 34
## 3 2014-06-01 2016-02-01 20
## 4 2015-01-01 2016-02-01 13
## 5 2016-01-01 2016-02-01 1
## 6 2015-01-01 2016-02-01 13summary(df$payment_time)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## -85.000 1.000 2.000 8.909 13.000 97.000head(df[,c('last_credit_pull_d', 'earliest_cr_line',
'credit_time')])## # A tibble: 6 × 3
## last_credit_pull_d earliest_cr_line credit_time
## <date> <date> <dbl>
## 1 2016-01-01 1985-01-01 31
## 2 2013-09-01 1999-04-01 14
## 3 2016-01-01 2001-11-01 15
## 4 2015-01-01 1996-02-01 19
## 5 2016-01-01 1996-01-01 20
## 6 2015-09-01 2004-11-01 11Terdapat nilai yang negatif pada kolom payment_time
dengan nilai ini akan diganti menjadi 0. Penyebab diperoleh nilai yang
negatif adalah adanya clients yang tidak dapat melakukan pembayaran.
df[df$payment_time<0,c('last_pymnt_d', 'next_pymnt_d', 'payment_time')]## # A tibble: 12 × 3
## last_pymnt_d next_pymnt_d payment_time
## <date> <date> <dbl>
## 1 2016-01-01 2011-02-01 -59
## 2 2016-01-01 2010-10-01 -63
## 3 2016-01-01 2010-10-01 -63
## 4 2016-01-01 2010-06-01 -67
## 5 2016-01-01 2010-06-01 -67
## 6 2016-01-01 2010-06-01 -67
## 7 2016-01-01 2010-05-01 -68
## 8 2016-01-01 2010-04-01 -69
## 9 2016-01-01 2010-04-01 -69
## 10 2016-01-01 2009-11-01 -74
## 11 2016-01-01 2008-12-01 -85
## 12 2016-01-01 2009-01-01 -84summary(df$credit_time)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## -57.00 13.00 17.00 18.03 22.00 47.00df[df$credit_time<0,c('last_credit_pull_d',
'earliest_cr_line','credit_time')]## # A tibble: 1,169 × 3
## last_credit_pull_d earliest_cr_line credit_time
## <date> <date> <dbl>
## 1 2013-09-01 2062-09-01 -49
## 2 2015-09-01 2068-09-01 -53
## 3 2016-01-01 2064-09-01 -48
## 4 2015-05-01 2067-09-01 -52
## 5 2014-11-01 2065-02-01 -51
## 6 2013-04-01 2067-06-01 -54
## 7 2014-04-01 2067-08-01 -53
## 8 2014-11-01 2063-12-01 -49
## 9 2016-01-01 2068-09-01 -52
## 10 2015-10-01 2068-09-01 -53
## # ℹ 1,159 more rowssummary(df$loan_time)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.00 19.00 25.00 28.88 35.00 104.00head(df[,c('issue_d', 'next_pymnt_d',
'loan_time')])## # A tibble: 6 × 3
## issue_d next_pymnt_d loan_time
## <date> <date> <dbl>
## 1 2011-12-01 2016-02-01 50
## 2 2011-12-01 2016-02-01 50
## 3 2011-12-01 2016-02-01 50
## 4 2011-12-01 2016-02-01 50
## 5 2011-12-01 2016-02-01 50
## 6 2011-12-01 2016-02-01 50Juga terdapat nilai yang negatif pada kolom credit_time
dengan nilai ini akan diganti menjadi maksimum. Penyebab diperoleh nilai
yang negatif adalah adanya kesalahan pengubahan data tanggal yaitu
misalnya sept-62 diubah menjadi 2062-09-01. Data ini harusnya diubah
menjadi 1962-09-01.
df <- df %>%
mutate(
payment_time = ifelse(payment_time<0, 0, payment_time),
credit_time = ifelse(credit_time<0, max(credit_time, na.rm = TRUE), credit_time)
)Perubahan summary pada kolom payment_time dan
credit_time.
summary(df$payment_time)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.000 1.000 2.000 8.911 13.000 97.000summary(df$credit_time)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.00 13.00 17.00 18.27 22.00 47.00Selanjutnya, kolom tanggal akan dihapus dari dataset
df <- df[,!names(df) %in% date_columns]5 EDA
Exploratory Data Analysis
5.1 Correlation Check
library(corrplot)## corrplot 0.92 loadedcorrplot(cor(df[,sapply(df, is.numeric)]), method="number",
type="upper", diag=FALSE, tl.cex=0.7)
library(ggcorrplot)## Warning: package 'ggcorrplot' was built under R version 4.3.2pmat <- df %>%
select_if(is.numeric) %>%
cor_pmat()
df %>%
select_if(is.numeric) %>%
cor() %>%
ggcorrplot(colors = c("firebrick", "white", "green"),
hc.order = TRUE, p.mat = pmat,
lab = TRUE, lab_size = 4, digits = 1,
ggtheme = theme_bw())
data.frame(abs(cor(df[,sapply(df, is.numeric)])))## loan_amnt funded_amnt funded_amnt_inv term
## loan_amnt 1.000000000 0.998547547 0.994347021 0.4128702547
## funded_amnt 0.998547547 1.000000000 0.996125407 0.4108621600
## funded_amnt_inv 0.994347021 0.996125407 1.000000000 0.4120047058
## term 0.412870255 0.410862160 0.412004706 1.0000000000
## int_rate 0.167182789 0.167920761 0.169353404 0.4433946928
## installment 0.949665620 0.951786604 0.947387015 0.1596305798
## emp_length 0.117569597 0.118105464 0.121100727 0.0795046945
## annual_inc 0.370892683 0.370310580 0.367805276 0.0699496280
## dti 0.057318779 0.058672721 0.062606252 0.0919121413
## delinq_2yrs 0.006868767 0.007413858 0.008177758 0.0041488967
## inq_last_6mths 0.020268657 0.020910467 0.027503013 0.0100029233
## open_acc 0.204189445 0.204990936 0.205813634 0.0777218963
## pub_rec 0.081107663 0.080564033 0.078855470 0.0227776455
## revol_bal 0.331422648 0.331039167 0.327436495 0.0925771092
## revol_util 0.117675554 0.118803283 0.120684027 0.0841125934
## total_acc 0.235263056 0.235136164 0.235454597 0.1013299180
## out_prncp 0.518699588 0.521450679 0.525143168 0.4343065006
## out_prncp_inv 0.518659672 0.521412227 0.525115350 0.4342730598
## total_pymnt 0.743840619 0.743362844 0.738036877 0.1749232411
## total_pymnt_inv 0.743580430 0.743440906 0.745459871 0.1773277404
## total_rec_prncp 0.612776695 0.612317698 0.607027005 0.0097980128
## total_rec_int 0.715875859 0.715664457 0.713460135 0.5343329157
## total_rec_late_fee 0.043649948 0.043350435 0.038984991 0.0119152816
## recoveries 0.110691145 0.110338153 0.107944429 0.0900185914
## collection_recovery_fee 0.077336200 0.077189551 0.073781226 0.0559001087
## last_pymnt_amnt 0.304243272 0.304178424 0.303104762 0.1115968183
## collections_12_mths_ex_med 0.008039230 0.007793597 0.007131168 0.0003727981
## acc_now_delinq 0.006287670 0.006476418 0.006806056 0.0077262582
## tot_coll_amt 0.002023684 0.001989434 0.001890075 0.0014347596
## tot_cur_bal 0.330107186 0.333024613 0.337985206 0.1187821441
## total_rev_hi_lim 0.292122180 0.294750624 0.299241516 0.0672086473
## bad_flag 0.011819203 0.011003248 0.007191411 0.0635648402
## payment_time 0.121654654 0.127028170 0.147383752 0.1007395526
## credit_time 0.184766244 0.184309567 0.184472644 0.0637074295
## loan_time 0.120368906 0.127753649 0.148743161 0.0805462217
## int_rate installment emp_length annual_inc
## loan_amnt 0.167182789 0.949665620 0.1175695974 0.3708926827
## funded_amnt 0.167920761 0.951786604 0.1181054637 0.3703105805
## funded_amnt_inv 0.169353404 0.947387015 0.1211007268 0.3678052764
## term 0.443394693 0.159630580 0.0795046945 0.0699496280
## int_rate 1.000000000 0.148790096 0.0297775049 0.0460689629
## installment 0.148790096 1.000000000 0.1024024650 0.3681450476
## emp_length 0.029777505 0.102402465 1.0000000000 0.0729817218
## annual_inc 0.046068963 0.368145048 0.0729817218 1.0000000000
## dti 0.159657407 0.049918721 0.0486368488 0.1885421251
## delinq_2yrs 0.079209110 0.017027938 0.0297582840 0.0588947748
## inq_last_6mths 0.205676570 0.002304978 0.0160189611 0.0566743497
## open_acc 0.012321458 0.196847258 0.0416596265 0.1577782383
## pub_rec 0.066712130 0.070064477 0.0380229074 0.0154371450
## revol_bal 0.004588639 0.320239374 0.0903617294 0.3323159370
## revol_util 0.323249068 0.136759333 0.0434388712 0.0375630079
## total_acc 0.032999652 0.215908000 0.1114628944 0.2246824781
## out_prncp 0.137427089 0.408837770 0.0856896406 0.1683168076
## out_prncp_inv 0.137451034 0.408810796 0.0856914062 0.1683170848
## total_pymnt 0.127035449 0.764360214 0.0782186118 0.2997844369
## total_pymnt_inv 0.129165807 0.763894699 0.0818547608 0.2987280965
## total_rec_prncp 0.032130464 0.663398905 0.0574376624 0.2758236095
## total_rec_int 0.485153918 0.641947759 0.0961843664 0.2129518016
## total_rec_late_fee 0.057875379 0.052340847 0.0116547697 0.0202729256
## recoveries 0.127892527 0.105666476 0.0068811325 0.0169540700
## collection_recovery_fee 0.082197687 0.075344082 0.0046930307 0.0136803950
## last_pymnt_amnt 0.075608886 0.299867139 0.0373457042 0.1421560231
## collections_12_mths_ex_med 0.020319236 0.006027600 0.0002173682 0.0004448236
## acc_now_delinq 0.030340591 0.008927168 0.0096787831 0.0171326649
## tot_coll_amt 0.001415883 0.001522130 0.0010212463 0.0019074386
## tot_cur_bal 0.044509453 0.304368976 0.1133953970 0.4309300561
## total_rev_hi_lim 0.093074431 0.276346626 0.0867805332 0.2651156981
## bad_flag 0.172360667 0.013935632 0.0161638181 0.0498630960
## payment_time 0.050143572 0.102778148 0.0843159591 0.0274456304
## credit_time 0.084212604 0.164275000 0.2210533559 0.1712608663
## loan_time 0.076691358 0.109865608 0.0836697587 0.0351246765
## dti delinq_2yrs inq_last_6mths
## loan_amnt 0.0573187794 0.0068687670 0.020268657
## funded_amnt 0.0586727211 0.0074138584 0.020910467
## funded_amnt_inv 0.0626062525 0.0081777577 0.027503013
## term 0.0919121413 0.0041488967 0.010002923
## int_rate 0.1596574074 0.0792091100 0.205676570
## installment 0.0499187207 0.0170279380 0.002304978
## emp_length 0.0486368488 0.0297582840 0.016018961
## annual_inc 0.1885421251 0.0588947748 0.056674350
## dti 1.0000000000 0.0036671069 0.012485003
## delinq_2yrs 0.0036671069 1.0000000000 0.018000976
## inq_last_6mths 0.0124850030 0.0180009761 1.000000000
## open_acc 0.3039144471 0.0591390343 0.092797307
## pub_rec 0.0461659793 0.0108032590 0.038341369
## revol_bal 0.1433802655 0.0311032392 0.015678670
## revol_util 0.2002844038 0.0131672201 0.094671264
## total_acc 0.2282215407 0.1307964968 0.124077802
## out_prncp 0.1199841574 0.0438714904 0.070368170
## out_prncp_inv 0.1199721397 0.0438670711 0.070357459
## total_pymnt 0.0260071365 0.0196764626 0.020683323
## total_pymnt_inv 0.0217879468 0.0189634781 0.014826606
## total_rec_prncp 0.0638953484 0.0320802184 0.005339648
## total_rec_int 0.0895823585 0.0242924213 0.044105180
## total_rec_late_fee 0.0057469635 0.0243884608 0.029925981
## recoveries 0.0214618789 0.0044794288 0.042931465
## collection_recovery_fee 0.0180835458 0.0053636240 0.032670847
## last_pymnt_amnt 0.0430676308 0.0137276775 0.041252185
## collections_12_mths_ex_med 0.0003350904 0.0391319901 0.001824517
## acc_now_delinq 0.0094944460 0.1265332731 0.006914717
## tot_coll_amt 0.0025062606 0.0004267052 0.001323279
## tot_cur_bal 0.0354641236 0.0894899630 0.028149856
## total_rev_hi_lim 0.1002964574 0.0110366785 0.012436673
## bad_flag 0.0481018913 0.0012276433 0.073099143
## payment_time 0.1491317191 0.0597273145 0.065690449
## credit_time 0.0446264520 0.0944119875 0.006707277
## loan_time 0.1673186105 0.0819043421 0.042821424
## open_acc pub_rec revol_bal revol_util
## loan_amnt 0.2041894452 0.081107663 0.331422648 0.117675554
## funded_amnt 0.2049909360 0.080564033 0.331039167 0.118803283
## funded_amnt_inv 0.2058136340 0.078855470 0.327436495 0.120684027
## term 0.0777218963 0.022777646 0.092577109 0.084112593
## int_rate 0.0123214582 0.066712130 0.004588639 0.323249068
## installment 0.1968472576 0.070064477 0.320239374 0.136759333
## emp_length 0.0416596265 0.038022907 0.090361729 0.043438871
## annual_inc 0.1577782383 0.015437145 0.332315937 0.037563008
## dti 0.3039144471 0.046165979 0.143380265 0.200284404
## delinq_2yrs 0.0591390343 0.010803259 0.031103239 0.013167220
## inq_last_6mths 0.0927973074 0.038341369 0.015678670 0.094671264
## open_acc 1.0000000000 0.030473338 0.222908137 0.124231591
## pub_rec 0.0304733383 1.000000000 0.097960889 0.062494937
## revol_bal 0.2229081366 0.097960889 1.000000000 0.206277024
## revol_util 0.1242315913 0.062494937 0.206277024 1.000000000
## total_acc 0.6821143279 0.007262237 0.198631658 0.094287508
## out_prncp 0.1445640375 0.002379727 0.181300649 0.096677529
## out_prncp_inv 0.1445665986 0.002370931 0.181284730 0.096678224
## total_pymnt 0.1202069882 0.091161390 0.242185883 0.085910770
## total_pymnt_inv 0.1219014957 0.089689836 0.240251030 0.088216421
## total_rec_prncp 0.0972274820 0.088756684 0.209921839 0.024331798
## total_rec_int 0.1218005672 0.049249874 0.206982234 0.209598626
## total_rec_late_fee 0.0064886760 0.011815211 0.005888523 0.024641687
## recoveries 0.0134323572 0.012689274 0.020966149 0.033102850
## collection_recovery_fee 0.0120704976 0.007016956 0.015156119 0.021145916
## last_pymnt_amnt 0.0549004486 0.027225452 0.087131869 0.010453890
## collections_12_mths_ex_med 0.0118596988 0.021527715 0.016146221 0.027665449
## acc_now_delinq 0.0181945005 0.002349158 0.001002554 0.022747353
## tot_coll_amt 0.0001830299 0.004372794 0.003635221 0.005136353
## tot_cur_bal 0.2537101368 0.038368880 0.388440351 0.081904808
## total_rev_hi_lim 0.2940651915 0.061555463 0.732382216 0.085171079
## bad_flag 0.0061736872 0.009110424 0.018348468 0.050187844
## payment_time 0.1028287321 0.072875486 0.054071746 0.084595528
## credit_time 0.1479151986 0.071914752 0.208978888 0.007199994
## loan_time 0.1332687216 0.116584698 0.043303384 0.047833406
## total_acc out_prncp out_prncp_inv total_pymnt
## loan_amnt 0.235263056 0.5186995880 0.5186596717 0.7438406193
## funded_amnt 0.235136164 0.5214506792 0.5214122270 0.7433628439
## funded_amnt_inv 0.235454597 0.5251431685 0.5251153496 0.7380368769
## term 0.101329918 0.4343065006 0.4342730598 0.1749232411
## int_rate 0.032999652 0.1374270895 0.1374510340 0.1270354495
## installment 0.215908000 0.4088377704 0.4088107958 0.7643602142
## emp_length 0.111462894 0.0856896406 0.0856914062 0.0782186118
## annual_inc 0.224682478 0.1683168076 0.1683170848 0.2997844369
## dti 0.228221541 0.1199841574 0.1199721397 0.0260071365
## delinq_2yrs 0.130796497 0.0438714904 0.0438670711 0.0196764626
## inq_last_6mths 0.124077802 0.0703681701 0.0703574590 0.0206833230
## open_acc 0.682114328 0.1445640375 0.1445665986 0.1202069882
## pub_rec 0.007262237 0.0023797267 0.0023709310 0.0911613902
## revol_bal 0.198631658 0.1813006490 0.1812847304 0.2421858834
## revol_util 0.094287508 0.0966775294 0.0966782244 0.0859107697
## total_acc 1.000000000 0.1234039130 0.1233937266 0.1686782810
## out_prncp 0.123403913 1.0000000000 0.9999982477 0.0220724107
## out_prncp_inv 0.123393727 0.9999982477 1.0000000000 0.0221411096
## total_pymnt 0.168678281 0.0220724107 0.0221411096 1.0000000000
## total_pymnt_inv 0.169945815 0.0166246723 0.0166813859 0.9958617153
## total_rec_prncp 0.152217528 0.1899325238 0.1899831211 0.9566581894
## total_rec_int 0.126851435 0.4881063840 0.4880193852 0.6244233556
## total_rec_late_fee 0.005511287 0.0066997764 0.0067116512 0.0261583352
## recoveries 0.016007882 0.1072481328 0.1072407148 0.0224211213
## collection_recovery_fee 0.016140045 0.0727418693 0.0727368380 0.0008899632
## last_pymnt_amnt 0.113015554 0.3173808031 0.3173586917 0.6142763499
## collections_12_mths_ex_med 0.013274555 0.0203460992 0.0203392535 0.0247290523
## acc_now_delinq 0.027718622 0.0164777057 0.0164736264 0.0042887059
## tot_coll_amt 0.005127203 0.0006327966 0.0006323573 0.0025907744
## tot_cur_bal 0.307199521 0.2451596496 0.2451719538 0.1959408990
## total_rev_hi_lim 0.232427774 0.2195087733 0.2195095329 0.1612131873
## bad_flag 0.023068054 0.1586758643 0.1586622653 0.1944376532
## payment_time 0.054888526 0.4361718206 0.4361407161 0.0489334273
## credit_time 0.298908651 0.1381579071 0.1381279611 0.1262558938
## loan_time 0.103985776 0.4052968919 0.4053500877 0.1786651253
## total_pymnt_inv total_rec_prncp total_rec_int
## loan_amnt 0.743580430 0.612776695 0.715875859
## funded_amnt 0.743440906 0.612317698 0.715664457
## funded_amnt_inv 0.745459871 0.607027005 0.713460135
## term 0.177327740 0.009798013 0.534332916
## int_rate 0.129165807 0.032130464 0.485153918
## installment 0.763894699 0.663398905 0.641947759
## emp_length 0.081854761 0.057437662 0.096184366
## annual_inc 0.298728097 0.275823610 0.212951802
## dti 0.021787947 0.063895348 0.089582358
## delinq_2yrs 0.018963478 0.032080218 0.024292421
## inq_last_6mths 0.014826606 0.005339648 0.044105180
## open_acc 0.121901496 0.097227482 0.121800567
## pub_rec 0.089689836 0.088756684 0.049249874
## revol_bal 0.240251030 0.209921839 0.206982234
## revol_util 0.088216421 0.024331798 0.209598626
## total_acc 0.169945815 0.152217528 0.126851435
## out_prncp 0.016624672 0.189932524 0.488106384
## out_prncp_inv 0.016681386 0.189983121 0.488019385
## total_pymnt 0.995861715 0.956658189 0.624423356
## total_pymnt_inv 1.000000000 0.952157884 0.623730606
## total_rec_prncp 0.952157884 1.000000000 0.378303926
## total_rec_int 0.623730606 0.378303926 1.000000000
## total_rec_late_fee 0.022136760 0.001048272 0.071733703
## recoveries 0.023904134 0.115783587 0.030690965
## collection_recovery_fee 0.003727803 0.073920417 0.028152277
## last_pymnt_amnt 0.614216281 0.709675976 0.050888129
## collections_12_mths_ex_med 0.024087677 0.026160178 0.008005301
## acc_now_delinq 0.003949829 0.008614144 0.009493475
## tot_coll_amt 0.002496250 0.002335063 0.001864433
## tot_cur_bal 0.202371552 0.174675988 0.157870720
## total_rev_hi_lim 0.167005460 0.151075701 0.109082261
## bad_flag 0.195169168 0.253874586 0.026725785
## payment_time 0.028562726 0.124622260 0.222452198
## credit_time 0.127163874 0.104464559 0.128424173
## loan_time 0.155873010 0.190346367 0.046253350
## total_rec_late_fee recoveries
## loan_amnt 0.0436499484 0.1106911453
## funded_amnt 0.0433504353 0.1103381525
## funded_amnt_inv 0.0389849914 0.1079444294
## term 0.0119152816 0.0900185914
## int_rate 0.0578753794 0.1278925270
## installment 0.0523408470 0.1056664759
## emp_length 0.0116547697 0.0068811325
## annual_inc 0.0202729256 0.0169540700
## dti 0.0057469635 0.0214618789
## delinq_2yrs 0.0243884608 0.0044794288
## inq_last_6mths 0.0299259811 0.0429314651
## open_acc 0.0064886760 0.0134323572
## pub_rec 0.0118152108 0.0126892737
## revol_bal 0.0058885234 0.0209661493
## revol_util 0.0246416873 0.0331028500
## total_acc 0.0055112867 0.0160078819
## out_prncp 0.0066997764 0.1072481328
## out_prncp_inv 0.0067116512 0.1072407148
## total_pymnt 0.0261583352 0.0224211213
## total_pymnt_inv 0.0221367603 0.0239041337
## total_rec_prncp 0.0010482724 0.1157835873
## total_rec_int 0.0717337027 0.0306909648
## total_rec_late_fee 1.0000000000 0.0727026464
## recoveries 0.0727026464 1.0000000000
## collection_recovery_fee 0.0690150543 0.8006664413
## last_pymnt_amnt 0.0348203742 0.0707220589
## collections_12_mths_ex_med 0.0005565259 0.0010219600
## acc_now_delinq 0.0034496846 0.0027615654
## tot_coll_amt 0.0006081156 0.0006534373
## tot_cur_bal 0.0067132556 0.0014090735
## total_rev_hi_lim 0.0176665653 0.0011510894
## bad_flag 0.1477504480 0.4411705780
## payment_time 0.0290911992 0.1458420790
## credit_time 0.0077855875 0.0179608157
## loan_time 0.0557540533 0.0419128259
## collection_recovery_fee last_pymnt_amnt
## loan_amnt 0.0773362002 0.3042432724
## funded_amnt 0.0771895506 0.3041784240
## funded_amnt_inv 0.0737812260 0.3031047618
## term 0.0559001087 0.1115968183
## int_rate 0.0821976869 0.0756088858
## installment 0.0753440819 0.2998671391
## emp_length 0.0046930307 0.0373457042
## annual_inc 0.0136803950 0.1421560231
## dti 0.0180835458 0.0430676308
## delinq_2yrs 0.0053636240 0.0137276775
## inq_last_6mths 0.0326708466 0.0412521854
## open_acc 0.0120704976 0.0549004486
## pub_rec 0.0070169564 0.0272254517
## revol_bal 0.0151561190 0.0871318688
## revol_util 0.0211459160 0.0104538899
## total_acc 0.0161400455 0.1130155536
## out_prncp 0.0727418693 0.3173808031
## out_prncp_inv 0.0727368380 0.3173586917
## total_pymnt 0.0008899632 0.6142763499
## total_pymnt_inv 0.0037278034 0.6142162807
## total_rec_prncp 0.0739204166 0.7096759757
## total_rec_int 0.0281522772 0.0508881289
## total_rec_late_fee 0.0690150543 0.0348203742
## recoveries 0.8006664413 0.0707220589
## collection_recovery_fee 1.0000000000 0.0477920275
## last_pymnt_amnt 0.0477920275 1.0000000000
## collections_12_mths_ex_med 0.0003710405 0.0097890508
## acc_now_delinq 0.0013778417 0.0009956164
## tot_coll_amt 0.0003216863 0.0014050910
## tot_cur_bal 0.0017849011 0.1238012055
## total_rev_hi_lim 0.0008177373 0.0842468716
## bad_flag 0.2992273310 0.1680295309
## payment_time 0.0860448933 0.2866262033
## credit_time 0.0026421432 0.0039942499
## loan_time 0.0222862492 0.0213078869
## collections_12_mths_ex_med acc_now_delinq
## loan_amnt 0.0080392300 0.0062876700
## funded_amnt 0.0077935972 0.0064764177
## funded_amnt_inv 0.0071311677 0.0068060557
## term 0.0003727981 0.0077262582
## int_rate 0.0203192362 0.0303405912
## installment 0.0060276003 0.0089271681
## emp_length 0.0002173682 0.0096787831
## annual_inc 0.0004448236 0.0171326649
## dti 0.0003350904 0.0094944460
## delinq_2yrs 0.0391319901 0.1265332731
## inq_last_6mths 0.0018245171 0.0069147168
## open_acc 0.0118596988 0.0181945005
## pub_rec 0.0215277146 0.0023491577
## revol_bal 0.0161462205 0.0010025544
## revol_util 0.0276654494 0.0227473535
## total_acc 0.0132745554 0.0277186225
## out_prncp 0.0203460992 0.0164777057
## out_prncp_inv 0.0203392535 0.0164736264
## total_pymnt 0.0247290523 0.0042887059
## total_pymnt_inv 0.0240876771 0.0039498292
## total_rec_prncp 0.0261601779 0.0086141441
## total_rec_int 0.0080053005 0.0094934745
## total_rec_late_fee 0.0005565259 0.0034496846
## recoveries 0.0010219600 0.0027615654
## collection_recovery_fee 0.0003710405 0.0013778417
## last_pymnt_amnt 0.0097890508 0.0009956164
## collections_12_mths_ex_med 1.0000000000 0.0192869119
## acc_now_delinq 0.0192869119 1.0000000000
## tot_coll_amt 0.0068958753 0.0001059103
## tot_cur_bal 0.0075629776 0.0304194293
## total_rev_hi_lim 0.0004373544 0.0161110094
## bad_flag 0.0051194698 0.0001968804
## payment_time 0.0330503497 0.0189002678
## credit_time 0.0077897286 0.0252586599
## loan_time 0.0514619288 0.0289237634
## tot_coll_amt tot_cur_bal total_rev_hi_lim
## loan_amnt 0.0020236840 0.330107186 0.2921221804
## funded_amnt 0.0019894338 0.333024613 0.2947506243
## funded_amnt_inv 0.0018900751 0.337985206 0.2992415155
## term 0.0014347596 0.118782144 0.0672086473
## int_rate 0.0014158831 0.044509453 0.0930744308
## installment 0.0015221297 0.304368976 0.2763466260
## emp_length 0.0010212463 0.113395397 0.0867805332
## annual_inc 0.0019074386 0.430930056 0.2651156981
## dti 0.0025062606 0.035464124 0.1002964574
## delinq_2yrs 0.0004267052 0.089489963 0.0110366785
## inq_last_6mths 0.0013232787 0.028149856 0.0124366735
## open_acc 0.0001830299 0.253710137 0.2940651915
## pub_rec 0.0043727941 0.038368880 0.0615554629
## revol_bal 0.0036352206 0.388440351 0.7323822157
## revol_util 0.0051363534 0.081904808 0.0851710785
## total_acc 0.0051272028 0.307199521 0.2324277741
## out_prncp 0.0006327966 0.245159650 0.2195087733
## out_prncp_inv 0.0006323573 0.245171954 0.2195095329
## total_pymnt 0.0025907744 0.195940899 0.1612131873
## total_pymnt_inv 0.0024962498 0.202371552 0.1670054599
## total_rec_prncp 0.0023350633 0.174675988 0.1510757012
## total_rec_int 0.0018644332 0.157870720 0.1090822605
## total_rec_late_fee 0.0006081156 0.006713256 0.0176665653
## recoveries 0.0006534373 0.001409074 0.0011510894
## collection_recovery_fee 0.0003216863 0.001784901 0.0008177373
## last_pymnt_amnt 0.0014050910 0.123801206 0.0842468716
## collections_12_mths_ex_med 0.0068958753 0.007562978 0.0004373544
## acc_now_delinq 0.0001059103 0.030419429 0.0161110094
## tot_coll_amt 1.0000000000 0.002397711 0.0015358893
## tot_cur_bal 0.0023977108 1.000000000 0.4210577115
## total_rev_hi_lim 0.0015358893 0.421057712 1.0000000000
## bad_flag 0.0013937395 0.063741222 0.0503257477
## payment_time 0.0045787456 0.207107599 0.1951475629
## credit_time 0.0032554656 0.172766847 0.1929114148
## loan_time 0.0066111235 0.266781870 0.2457667167
## bad_flag payment_time credit_time loan_time
## loan_amnt 0.0118192031 0.121654654 0.184766244 0.120368906
## funded_amnt 0.0110032476 0.127028170 0.184309567 0.127753649
## funded_amnt_inv 0.0071914109 0.147383752 0.184472644 0.148743161
## term 0.0635648402 0.100739553 0.063707430 0.080546222
## int_rate 0.1723606670 0.050143572 0.084212604 0.076691358
## installment 0.0139356323 0.102778148 0.164275000 0.109865608
## emp_length 0.0161638181 0.084315959 0.221053356 0.083669759
## annual_inc 0.0498630960 0.027445630 0.171260866 0.035124677
## dti 0.0481018913 0.149131719 0.044626452 0.167318610
## delinq_2yrs 0.0012276433 0.059727315 0.094411987 0.081904342
## inq_last_6mths 0.0730991426 0.065690449 0.006707277 0.042821424
## open_acc 0.0061736872 0.102828732 0.147915199 0.133268722
## pub_rec 0.0091104237 0.072875486 0.071914752 0.116584698
## revol_bal 0.0183484678 0.054071746 0.208978888 0.043303384
## revol_util 0.0501878437 0.084595528 0.007199994 0.047833406
## total_acc 0.0230680539 0.054888526 0.298908651 0.103985776
## out_prncp 0.1586758643 0.436171821 0.138157907 0.405296892
## out_prncp_inv 0.1586622653 0.436140716 0.138127961 0.405350088
## total_pymnt 0.1944376532 0.048933427 0.126255894 0.178665125
## total_pymnt_inv 0.1951691683 0.028562726 0.127163874 0.155873010
## total_rec_prncp 0.2538745860 0.124622260 0.104464559 0.190346367
## total_rec_int 0.0267257852 0.222452198 0.128424173 0.046253350
## total_rec_late_fee 0.1477504480 0.029091199 0.007785588 0.055754053
## recoveries 0.4411705780 0.145842079 0.017960816 0.041912826
## collection_recovery_fee 0.2992273310 0.086044893 0.002642143 0.022286249
## last_pymnt_amnt 0.1680295309 0.286626203 0.003994250 0.021307887
## collections_12_mths_ex_med 0.0051194698 0.033050350 0.007789729 0.051461929
## acc_now_delinq 0.0001968804 0.018900268 0.025258660 0.028923763
## tot_coll_amt 0.0013937395 0.004578746 0.003255466 0.006611124
## tot_cur_bal 0.0637412217 0.207107599 0.172766847 0.266781870
## total_rev_hi_lim 0.0503257477 0.195147563 0.192911415 0.245766717
## bad_flag 1.0000000000 0.237647430 0.053208439 0.076956923
## payment_time 0.2376474302 1.000000000 0.098204034 0.736315466
## credit_time 0.0532084388 0.098204034 1.000000000 0.037916352
## loan_time 0.0769569226 0.736315466 0.037916352 1.000000000Terdapat pasangan variabel yang memiliki korelai tinggi > 0.8, akan dipilih salah satu variabel saja.
dff <- df # untuk backup data# Calculate the correlation matrix
corr_matrix <- abs(cor(dff[,sapply(dff, is.numeric)]))
# Create and apply the mask
mask <- upper.tri(corr_matrix, diag = TRUE)
corr_matrix[mask] <- NA
# Find columns with correlation greater than 0.7
high_col <- colnames(corr_matrix)[apply(corr_matrix, 1, function(x) any(x > 0.8, na.rm = TRUE))]
high_col## [1] "funded_amnt" "funded_amnt_inv"
## [3] "installment" "out_prncp_inv"
## [5] "total_pymnt_inv" "total_rec_prncp"
## [7] "collection_recovery_fee"Variabel dengan correlation greater than 0.7 dihapus dari dataset.
dff <- dff[,!names(dff) %in% high_col]5.2 Categorical Check
colnames(dff[,sapply(dff, is.character)])## [1] "grade" "sub_grade" "home_ownership"
## [4] "verification_status" "pymnt_plan" "purpose"
## [7] "initial_list_status"summarise_all(dff[,sapply(dff,is.character)], list(Unique_Count = ~length(unique(.)))) %>%
gather(Column, Unique_Count) %>%
filter(!is.na(Unique_Count))## # A tibble: 7 × 2
## Column Unique_Count
## <chr> <int>
## 1 grade_Unique_Count 7
## 2 sub_grade_Unique_Count 35
## 3 home_ownership_Unique_Count 6
## 4 verification_status_Unique_Count 3
## 5 pymnt_plan_Unique_Count 2
## 6 purpose_Unique_Count 14
## 7 initial_list_status_Unique_Count 2Kolom sub_grade akan dihapus karena memiliki nilai unik
yang sangat banyak.
dff <- dff %>%
select(-sub_grade)sapply(dff %>% select_if(is.character), function(x) table(x) / length(x) * 100)## $grade
## x
## A B C D E F G
## 16.0560601 29.3659457 26.8704762 16.4894860 7.6684860 2.8371061 0.7124398
##
## $home_ownership
## x
## ANY MORTGAGE NONE OTHER OWN RENT
## 2.144611e-04 5.058601e+01 1.072306e-02 3.903192e-02 8.943886e+00 4.042013e+01
##
## $verification_status
## x
## Not Verified Source Verified Verified
## 31.79107 32.16767 36.04126
##
## $pymnt_plan
## x
## n y
## 99.99806985 0.00193015
##
## $purpose
## x
## car credit_card debt_consolidation educational
## 1.15744663 22.33762613 58.80416483 0.09050259
## home_improvement house major_purchase medical
## 5.69115455 0.48661227 2.10772382 0.98695004
## moving other renewable_energy small_business
## 0.64209657 5.08058376 0.07527585 1.50401578
## vacation wedding
## 0.53336479 0.50248239
##
## $initial_list_status
## x
## f w
## 64.98279 35.01721Pada kolom pymnt_plan terdiri dari 2 data unik yaitu
n dan y. Namun, perbandingan jumlah nilai
n dan y sangat jauh berbeda (kolom ini sangat
didominasi oleh nilai n saja). Akibatnya, variabel
pymnt_plan akan dihapus dari dataset.
dff <- dff %>%
select(-pymnt_plan)colnames(dff)## [1] "loan_amnt" "term"
## [3] "int_rate" "grade"
## [5] "emp_length" "home_ownership"
## [7] "annual_inc" "verification_status"
## [9] "purpose" "dti"
## [11] "delinq_2yrs" "inq_last_6mths"
## [13] "open_acc" "pub_rec"
## [15] "revol_bal" "revol_util"
## [17] "total_acc" "initial_list_status"
## [19] "out_prncp" "total_pymnt"
## [21] "total_rec_int" "total_rec_late_fee"
## [23] "recoveries" "last_pymnt_amnt"
## [25] "collections_12_mths_ex_med" "acc_now_delinq"
## [27] "tot_coll_amt" "tot_cur_bal"
## [29] "total_rev_hi_lim" "bad_flag"
## [31] "payment_time" "credit_time"
## [33] "loan_time"colnames(dff[,sapply(dff, is.character)])## [1] "grade" "home_ownership" "verification_status"
## [4] "purpose" "initial_list_status"Prediksi yang dilakukan menggunakan regresi logistik. Namun dalam pemodelan regresi logistik, seluruh variabel haruslah bertipe numerik. Oleh sebab itu, untuk variabel kategorik akan dihitung WoE dan IV.
6 WoE & IV
crs <- dff %>%
group_by(grade, bad_flag) %>%
summarise(count = n(), .groups = "drop") %>%
group_by(grade) %>%
mutate(probability = count / sum(count)) %>%
# filter(bad_flag == 1) %>%
select(grade, bad_flag, probability) %>%
pivot_wider(names_from = bad_flag, values_from = probability) %>%
arrange(grade)
crs## # A tibble: 7 × 3
## # Groups: grade [7]
## grade `0` `1`
## <chr> <dbl> <dbl>
## 1 A 0.961 0.0387
## 2 B 0.921 0.0786
## 3 C 0.885 0.115
## 4 D 0.846 0.154
## 5 E 0.805 0.195
## 6 F 0.759 0.241
## 7 G 0.722 0.278dff <- dff %>%
mutate_if(is.character,as.factor)IV <- Information::create_infotables(dff, y = "bad_flag")
format(IV$Summary, scientific = FALSE)## Variable IV
## 30 payment_time 1.672809480038
## 24 last_pymnt_amnt 1.465967583644
## 19 out_prncp 0.763527398792
## 20 total_pymnt 0.609577273280
## 3 int_rate 0.325231670615
## 23 recoveries 0.310996853162
## 4 grade 0.290782139541
## 22 total_rec_late_fee 0.169374635760
## 32 loan_time 0.097538855862
## 29 total_rev_hi_lim 0.059994509496
## 28 tot_cur_bal 0.055526858355
## 7 annual_inc 0.053941606223
## 12 inq_last_6mths 0.046715671352
## 2 term 0.038859834388
## 9 purpose 0.036980194720
## 31 credit_time 0.035722840355
## 16 revol_util 0.026880760157
## 10 dti 0.025982966556
## 18 initial_list_status 0.024801321308
## 21 total_rec_int 0.023441581240
## 6 home_ownership 0.021661906308
## 8 verification_status 0.020831404235
## 17 total_acc 0.007103883123
## 5 emp_length 0.005983625175
## 15 revol_bal 0.005743621809
## 1 loan_amnt 0.003629138665
## 27 tot_coll_amt 0.001404716364
## 13 open_acc 0.000676399516
## 14 pub_rec 0.000581987123
## 25 collections_12_mths_ex_med 0.000238378880
## 11 delinq_2yrs 0.000005419885
## 26 acc_now_delinq 0.000002470560IV_drop <- format(IV$Summary, scientific = FALSE) %>%
filter(IV < 0.02) %>%
pull(Variable)akan didrop variabel yang memiliki nilai Information Value < 0.02
colnames(dff)## [1] "loan_amnt" "term"
## [3] "int_rate" "grade"
## [5] "emp_length" "home_ownership"
## [7] "annual_inc" "verification_status"
## [9] "purpose" "dti"
## [11] "delinq_2yrs" "inq_last_6mths"
## [13] "open_acc" "pub_rec"
## [15] "revol_bal" "revol_util"
## [17] "total_acc" "initial_list_status"
## [19] "out_prncp" "total_pymnt"
## [21] "total_rec_int" "total_rec_late_fee"
## [23] "recoveries" "last_pymnt_amnt"
## [25] "collections_12_mths_ex_med" "acc_now_delinq"
## [27] "tot_coll_amt" "tot_cur_bal"
## [29] "total_rev_hi_lim" "bad_flag"
## [31] "payment_time" "credit_time"
## [33] "loan_time"dffs <- dff %>%
select (-one_of(IV_drop))
head(dffs)## # A tibble: 6 × 23
## term int_rate grade home_ownership annual_inc verification_status purpose
## <dbl> <dbl> <fct> <fct> <dbl> <fct> <fct>
## 1 36 10.6 B RENT 24000 Verified credit_card
## 2 60 15.3 C RENT 30000 Source Verified car
## 3 36 16.0 C RENT 12252 Not Verified small_busi…
## 4 36 13.5 C RENT 49200 Source Verified other
## 5 60 12.7 B RENT 80000 Source Verified other
## 6 36 7.9 A RENT 36000 Source Verified wedding
## # ℹ 16 more variables: dti <dbl>, inq_last_6mths <dbl>, revol_util <dbl>,
## # initial_list_status <fct>, out_prncp <dbl>, total_pymnt <dbl>,
## # total_rec_int <dbl>, total_rec_late_fee <dbl>, recoveries <dbl>,
## # last_pymnt_amnt <dbl>, tot_cur_bal <dbl>, total_rev_hi_lim <dbl>,
## # bad_flag <dbl>, payment_time <dbl>, credit_time <dbl>, loan_time <dbl>Terdapat kolom yang bertipe kategorik. Dilakukan transformasi feature encoding untuk kolom kategorik ini.
7 Feature Encoding
sapply(dffs %>% select_if(is.factor), function(x) table(x))## $grade
## x
## A B C D E F G
## 74867 136929 125293 76888 35757 13229 3322
##
## $home_ownership
## x
## ANY MORTGAGE NONE OTHER OWN RENT
## 1 235875 50 182 41704 188473
##
## $verification_status
## x
## Not Verified Source Verified Verified
## 148237 149993 168055
##
## $purpose
## x
## car credit_card debt_consolidation educational
## 5397 104157 274195 422
## home_improvement house major_purchase medical
## 26537 2269 9828 4602
## moving other renewable_energy small_business
## 2994 23690 351 7013
## vacation wedding
## 2487 2343
##
## $initial_list_status
## x
## f w
## 303005 163280pada kolom home_ownership terdapat nilai
any, none, other yang jumlahnya cukup kecil. Oleh sebab
itu, ketiga nilai ini akan disatukan.
dffs <- dffs %>%
mutate(home_ownership = factor(case_when(
home_ownership %in% c("ANY", "NONE") ~ "OTHER",
TRUE ~ as.character(home_ownership)
)))
head(dffs)## # A tibble: 6 × 23
## term int_rate grade home_ownership annual_inc verification_status purpose
## <dbl> <dbl> <fct> <fct> <dbl> <fct> <fct>
## 1 36 10.6 B RENT 24000 Verified credit_card
## 2 60 15.3 C RENT 30000 Source Verified car
## 3 36 16.0 C RENT 12252 Not Verified small_busi…
## 4 36 13.5 C RENT 49200 Source Verified other
## 5 60 12.7 B RENT 80000 Source Verified other
## 6 36 7.9 A RENT 36000 Source Verified wedding
## # ℹ 16 more variables: dti <dbl>, inq_last_6mths <dbl>, revol_util <dbl>,
## # initial_list_status <fct>, out_prncp <dbl>, total_pymnt <dbl>,
## # total_rec_int <dbl>, total_rec_late_fee <dbl>, recoveries <dbl>,
## # last_pymnt_amnt <dbl>, tot_cur_bal <dbl>, total_rev_hi_lim <dbl>,
## # bad_flag <dbl>, payment_time <dbl>, credit_time <dbl>, loan_time <dbl>sapply(dffs %>% select_if(is.factor), function(x) table(x))## $grade
## x
## A B C D E F G
## 74867 136929 125293 76888 35757 13229 3322
##
## $home_ownership
## x
## MORTGAGE OTHER OWN RENT
## 235875 233 41704 188473
##
## $verification_status
## x
## Not Verified Source Verified Verified
## 148237 149993 168055
##
## $purpose
## x
## car credit_card debt_consolidation educational
## 5397 104157 274195 422
## home_improvement house major_purchase medical
## 26537 2269 9828 4602
## moving other renewable_energy small_business
## 2994 23690 351 7013
## vacation wedding
## 2487 2343
##
## $initial_list_status
## x
## f w
## 303005 1632807.1 Numerical Features Encode
Pada kolom numerik dilakukan proses standarisasi (penskalaan)
dffs_encoded <- dffs %>%
select(-starts_with("bad_flag")) %>%
mutate(across(where(is.numeric), scale)) %>%
bind_cols(select(df, starts_with("bad_flag")), .)
head(dffs_encoded)## # A tibble: 6 × 23
## bad_flag term[,1] int_rate[,1] grade home_ownership annual_inc[,1]
## <dbl> <dbl> <dbl> <fct> <fct> <dbl>
## 1 0 -0.616 -0.730 B RENT -0.897
## 2 1 1.62 0.331 C RENT -0.787
## 3 0 -0.616 0.489 C RENT -1.11
## 4 0 -0.616 -0.0778 C RENT -0.438
## 5 0 1.62 -0.261 B RENT 0.122
## 6 0 -0.616 -1.36 A RENT -0.678
## # ℹ 17 more variables: verification_status <fct>, purpose <fct>, dti <dbl[,1]>,
## # inq_last_6mths <dbl[,1]>, revol_util <dbl[,1]>, initial_list_status <fct>,
## # out_prncp <dbl[,1]>, total_pymnt <dbl[,1]>, total_rec_int <dbl[,1]>,
## # total_rec_late_fee <dbl[,1]>, recoveries <dbl[,1]>,
## # last_pymnt_amnt <dbl[,1]>, tot_cur_bal <dbl[,1]>,
## # total_rev_hi_lim <dbl[,1]>, payment_time <dbl[,1]>, credit_time <dbl[,1]>,
## # loan_time <dbl[,1]>7.2 Label Encoding
kolom grade menunjukkan nilai pinjaman, berisi nilai
A,B,C,D,E,F,G. Akan diberikan nilai numerik sesuai dengan
urutan performa dari A (paling baik) hingga G
paling buruk.
dffs_encoded <- dffs_encoded %>%
mutate(grade_encoded = case_when(
grade == "A" ~ 7,
grade == "B" ~ 6,
grade == "C" ~ 5,
grade == "D" ~ 4,
grade == "E" ~ 3,
grade == "F" ~ 2,
grade == "G" ~ 1,
TRUE ~ NA_real_
)) %>%
select(-grade)
head(dffs_encoded)## # A tibble: 6 × 23
## bad_flag term[,1] int_rate[,1] home_ownership annual_inc[,1]
## <dbl> <dbl> <dbl> <fct> <dbl>
## 1 0 -0.616 -0.730 RENT -0.897
## 2 1 1.62 0.331 RENT -0.787
## 3 0 -0.616 0.489 RENT -1.11
## 4 0 -0.616 -0.0778 RENT -0.438
## 5 0 1.62 -0.261 RENT 0.122
## 6 0 -0.616 -1.36 RENT -0.678
## # ℹ 18 more variables: verification_status <fct>, purpose <fct>, dti <dbl[,1]>,
## # inq_last_6mths <dbl[,1]>, revol_util <dbl[,1]>, initial_list_status <fct>,
## # out_prncp <dbl[,1]>, total_pymnt <dbl[,1]>, total_rec_int <dbl[,1]>,
## # total_rec_late_fee <dbl[,1]>, recoveries <dbl[,1]>,
## # last_pymnt_amnt <dbl[,1]>, tot_cur_bal <dbl[,1]>,
## # total_rev_hi_lim <dbl[,1]>, payment_time <dbl[,1]>, credit_time <dbl[,1]>,
## # loan_time <dbl[,1]>, grade_encoded <dbl>7.3 One-Hot Encoding
Digunakan untuk variabel kategorik nominal (tidak memiliki tingkatan/order).
dffs_encoded <- dffs_encoded %>%
select_if(is.factor) %>%
model.matrix(~ . - 1, data = .) %>%
bind_cols(dffs_encoded, .) %>%
select(-one_of(colnames(dffs_encoded[,sapply(dffs_encoded, is.factor)]))) %>%
rename(verification_status_Source_Verified = `verification_statusSource Verified`)
head(dffs_encoded)## # A tibble: 6 × 39
## bad_flag term[,1] int_rate[,1] annual_inc[,1] dti[,1] inq_last_6mths[,1]
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0 -0.616 -0.730 -0.897 1.33 0.179
## 2 1 1.62 0.331 -0.787 -2.07 3.84
## 3 0 -0.616 0.489 -1.11 -1.08 1.10
## 4 0 -0.616 -0.0778 -0.438 0.354 0.179
## 5 0 1.62 -0.261 0.122 0.0919 -0.737
## 6 0 -0.616 -1.36 -0.678 -0.767 2.01
## # ℹ 33 more variables: revol_util <dbl[,1]>, out_prncp <dbl[,1]>,
## # total_pymnt <dbl[,1]>, total_rec_int <dbl[,1]>,
## # total_rec_late_fee <dbl[,1]>, recoveries <dbl[,1]>,
## # last_pymnt_amnt <dbl[,1]>, tot_cur_bal <dbl[,1]>,
## # total_rev_hi_lim <dbl[,1]>, payment_time <dbl[,1]>, credit_time <dbl[,1]>,
## # loan_time <dbl[,1]>, grade_encoded <dbl>, home_ownershipMORTGAGE <dbl>,
## # home_ownershipOTHER <dbl>, home_ownershipOWN <dbl>, …8 Modeling
Dataset masih memiliki masalah imbalanced data yang belum diatasi. Problem ini akan diatasi dengan SMOTE (synthetic minority oversampling technique)
library(rsample)## Warning: package 'rsample' was built under R version 4.3.2set.seed(100)
df_split <- dffs_encoded %>%
initial_split(prop = 0.8, strata = bad_flag)
df_split## <Training/Testing/Total>
## <373027/93258/466285>df_train <- df_split %>% training()
head(df_train)## # A tibble: 6 × 39
## bad_flag term[,1] int_rate[,1] annual_inc[,1] dti[,1] inq_last_6mths[,1]
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0 -0.616 -0.730 -0.897 1.33 0.179
## 2 0 -0.616 0.489 -1.11 -1.08 1.10
## 3 0 -0.616 -0.0778 -0.438 0.354 0.179
## 4 0 1.62 -0.261 0.122 0.0919 -0.737
## 5 0 -0.616 -1.36 -0.678 -0.767 2.01
## 6 0 1.62 0.489 -0.478 0.801 0.179
## # ℹ 33 more variables: revol_util <dbl[,1]>, out_prncp <dbl[,1]>,
## # total_pymnt <dbl[,1]>, total_rec_int <dbl[,1]>,
## # total_rec_late_fee <dbl[,1]>, recoveries <dbl[,1]>,
## # last_pymnt_amnt <dbl[,1]>, tot_cur_bal <dbl[,1]>,
## # total_rev_hi_lim <dbl[,1]>, payment_time <dbl[,1]>, credit_time <dbl[,1]>,
## # loan_time <dbl[,1]>, grade_encoded <dbl>, home_ownershipMORTGAGE <dbl>,
## # home_ownershipOTHER <dbl>, home_ownershipOWN <dbl>, …df_test <- df_split %>% testing()
head(df_test)## # A tibble: 6 × 39
## bad_flag term[,1] int_rate[,1] annual_inc[,1] dti[,1] inq_last_6mths[,1]
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 1 1.62 1.71 -0.605 -1.49 1.10
## 2 0 1.62 0.188 -0.0232 -0.140 1.10
## 3 1 -0.616 -0.0778 -0.787 -0.909 0.179
## 4 0 -0.616 -1.79 0.668 -0.853 -0.737
## 5 1 -0.616 -0.486 0.0495 -1.89 0.179
## 6 1 1.62 0.675 -0.423 -0.414 -0.737
## # ℹ 33 more variables: revol_util <dbl[,1]>, out_prncp <dbl[,1]>,
## # total_pymnt <dbl[,1]>, total_rec_int <dbl[,1]>,
## # total_rec_late_fee <dbl[,1]>, recoveries <dbl[,1]>,
## # last_pymnt_amnt <dbl[,1]>, tot_cur_bal <dbl[,1]>,
## # total_rev_hi_lim <dbl[,1]>, payment_time <dbl[,1]>, credit_time <dbl[,1]>,
## # loan_time <dbl[,1]>, grade_encoded <dbl>, home_ownershipMORTGAGE <dbl>,
## # home_ownershipOTHER <dbl>, home_ownershipOWN <dbl>, …Melakukan SMOTE
# # Mengonversi kolom menjadi numerik dengan penanganan nilai-nilai yang hilang
# df_train <- df_train %>%
# mutate(across(everything(), ~as.numeric(coalesce(as.character(.), "NA"))))
#
# library(DMwR)
#
# df_train[] <- lapply(df_train, as.numeric)
# df_train <- as.data.frame(df_train) %>%
# rename(verification_status_Source_Verified = `verification_statusSource Verified`)
#
#
# # # Apply ROSE to address imbalanced data
# # df_train_balanced <- SMOTE(bad_flag ~ ., data = df_train, perc.over = 100, k = 5)
library(ROSE)## Warning: package 'ROSE' was built under R version 4.3.2## Loaded ROSE 0.0-4# Specify the target variable and predictors
# Specify the target variable and predictors
target_var <- "bad_flag"
predictors <- setdiff(colnames(df_train), target_var)
# Create the formula
formula_str <- paste(target_var, "~", paste(predictors, collapse = " + "))
# Use ovun.sample for oversampling
df_train_balanced <- ovun.sample(as.formula(formula_str), data = df_train, method = "over")$dataprop.table(table(df_train_balanced$bad_flag))##
## 0 1
## 0.499401 0.5005998.1 Binary Logistic Regression
X_train <- df_train_balanced[,-1]
y_train <- df_train_balanced$bad_flaglibrary(pROC)## Type 'citation("pROC")' for a citation.##
## Attaching package: 'pROC'## The following objects are masked from 'package:stats':
##
## cov, smooth, varhead(X_train)## term int_rate annual_inc dti inq_last_6mths revol_util
## 1 -0.6162244 -0.7295864 -0.8965484 1.32863088 0.1789195 1.16009143
## 2 -0.6162244 0.4889780 -1.1102908 -1.08248971 1.0950206 1.78393214
## 3 -0.6162244 -0.0778495 -0.4380609 0.35424776 0.1789195 -1.48280129
## 4 1.6227821 -0.2614374 0.1223126 0.09186485 -0.7371815 -0.09602026
## 5 -0.6162244 -1.3606697 -0.6782210 -0.76661126 2.0111216 -1.17509607
## 6 1.6227821 0.4889780 -0.4780149 0.80131766 0.1789195 1.24017909
## out_prncp total_pymnt total_rec_int total_rec_late_fee recoveries
## 1 -0.6939430 -0.68713659 -0.6955473 -0.1234642 -0.1545486
## 2 -0.6939430 -1.03283544 -0.7991865 -0.1234642 -0.1545486
## 3 -0.6939430 0.08294785 -0.1527280 3.0992610 -0.1545486
## 4 -0.5732679 -1.00397902 -0.6359615 -0.1234642 -0.1545486
## 5 -0.6939430 -0.71492561 -0.7880222 -0.1234642 -0.1545486
## 6 -0.3966768 -0.41180738 0.1760653 -0.1234642 -0.1545486
## last_pymnt_amnt tot_cur_bal total_rev_hi_lim payment_time credit_time
## 1 -0.5314912 -0.7926475 -0.7165747 0.3336908 1.7490531
## 2 -0.4453863 -0.7926475 -0.7165747 0.9049419 -0.4488870
## 3 -0.4980314 -0.7926475 -0.7165747 0.3336908 0.1005980
## 4 -0.5501833 -0.7926475 -0.7165747 -0.6455968 0.2379693
## 5 -0.5333976 -0.7926475 -0.7165747 0.3336908 -0.9983720
## 6 -0.5317684 -0.7926475 -0.7165747 -0.6455968 -0.9983720
## loan_time grade_encoded home_ownershipMORTGAGE home_ownershipOTHER
## 1 1.532957 6 0 0
## 2 1.532957 5 0 0
## 3 1.532957 5 0 0
## 4 1.532957 6 0 0
## 5 1.532957 7 0 0
## 6 1.532957 5 0 0
## home_ownershipOWN home_ownershipRENT verification_status_Source_Verified
## 1 0 1 0
## 2 0 1 0
## 3 0 1 1
## 4 0 1 1
## 5 0 1 1
## 6 0 1 0
## verification_statusVerified purposecredit_card purposedebt_consolidation
## 1 1 1 0
## 2 0 0 0
## 3 0 0 0
## 4 0 0 0
## 5 0 0 0
## 6 0 0 1
## purposeeducational purposehome_improvement purposehouse purposemajor_purchase
## 1 0 0 0 0
## 2 0 0 0 0
## 3 0 0 0 0
## 4 0 0 0 0
## 5 0 0 0 0
## 6 0 0 0 0
## purposemedical purposemoving purposeother purposerenewable_energy
## 1 0 0 0 0
## 2 0 0 0 0
## 3 0 0 1 0
## 4 0 0 1 0
## 5 0 0 0 0
## 6 0 0 0 0
## purposesmall_business purposevacation purposewedding initial_list_statusw
## 1 0 0 0 0
## 2 1 0 0 0
## 3 0 0 0 0
## 4 0 0 0 0
## 5 0 0 1 0
## 6 0 0 0 0# Create a logistic regression model
logreg <- glm(bad_flag ~ ., data = df_train_balanced, family = "binomial")## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurredlogreg %>%
summary##
## Call:
## glm(formula = bad_flag ~ ., family = "binomial", data = df_train_balanced)
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 5.680e+02 4.750e+03 0.120 0.904816
## term -7.307e-02 6.448e-03 -11.333 < 2e-16 ***
## int_rate 8.197e-01 1.596e-02 51.371 < 2e-16 ***
## annual_inc 3.768e-02 6.779e-03 5.559 2.72e-08 ***
## dti 9.148e-02 4.589e-03 19.935 < 2e-16 ***
## inq_last_6mths 6.863e-02 4.348e-03 15.787 < 2e-16 ***
## revol_util 6.289e-02 4.991e-03 12.600 < 2e-16 ***
## out_prncp -1.244e+00 7.251e-03 -171.565 < 2e-16 ***
## total_pymnt -1.225e+00 1.354e-02 -90.507 < 2e-16 ***
## total_rec_int 1.856e+00 1.238e-02 150.004 < 2e-16 ***
## total_rec_late_fee 4.625e-01 4.514e-03 102.471 < 2e-16 ***
## recoveries 3.726e+03 3.074e+04 0.121 0.903506
## last_pymnt_amnt -6.736e+00 4.837e-02 -139.274 < 2e-16 ***
## tot_cur_bal 2.516e-03 6.801e-03 0.370 0.711401
## total_rev_hi_lim 1.547e-01 7.408e-03 20.883 < 2e-16 ***
## payment_time 2.665e+00 1.180e-02 225.911 < 2e-16 ***
## credit_time 1.515e-02 4.572e-03 3.313 0.000924 ***
## loan_time -2.215e+00 1.104e-02 -200.503 < 2e-16 ***
## grade_encoded 5.845e-01 1.147e-02 50.947 < 2e-16 ***
## home_ownershipMORTGAGE -6.477e-02 1.089e-02 -5.946 2.75e-09 ***
## home_ownershipOTHER -3.247e-01 2.299e-01 -1.413 0.157756
## home_ownershipOWN -9.575e-02 1.568e-02 -6.107 1.02e-09 ***
## home_ownershipRENT NA NA NA NA
## verification_status_Source_Verified 2.386e-01 1.135e-02 21.029 < 2e-16 ***
## verification_statusVerified 3.917e-01 1.172e-02 33.424 < 2e-16 ***
## purposecredit_card 7.412e-01 4.749e-02 15.608 < 2e-16 ***
## purposedebt_consolidation 7.686e-01 4.685e-02 16.406 < 2e-16 ***
## purposeeducational -5.147e-01 1.654e-01 -3.112 0.001857 **
## purposehome_improvement 5.099e-01 5.026e-02 10.146 < 2e-16 ***
## purposehouse 6.527e-01 8.082e-02 8.075 6.73e-16 ***
## purposemajor_purchase 1.758e-01 5.629e-02 3.124 0.001787 **
## purposemedical 1.420e-01 6.212e-02 2.285 0.022301 *
## purposemoving 2.239e-01 6.751e-02 3.317 0.000911 ***
## purposeother 2.319e-01 4.987e-02 4.649 3.34e-06 ***
## purposerenewable_energy 2.543e-01 1.698e-01 1.497 0.134330
## purposesmall_business 7.910e-01 5.871e-02 13.474 < 2e-16 ***
## purposevacation -1.485e-01 7.285e-02 -2.038 0.041536 *
## purposewedding 1.239e-01 9.351e-02 1.325 0.185071
## initial_list_statusw 1.469e-03 9.390e-03 0.156 0.875649
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 922305 on 665302 degrees of freedom
## Residual deviance: 335807 on 665265 degrees of freedom
## AIC: 335883
##
## Number of Fisher Scoring iterations: 20# Make predictions on the training set
y_pred_proba_train <- predict(logreg, newdata = df_train_balanced, type = "response")## Warning in predict.lm(object, newdata, se.fit, scale = 1, type = if (type == :
## prediction from rank-deficient fit; attr(*, "non-estim") has doubtful cases# Make predictions on the test set
data_test <- model.matrix(as.formula(formula_str), data = df_test)
X_test <- data_test[, -1]
y_pred_proba_test <- predict(logreg, newdata = df_test, type = "response")## Warning in predict.lm(object, newdata, se.fit, scale = 1, type = if (type == :
## prediction from rank-deficient fit; attr(*, "non-estim") has doubtful caseslibrary(pROC)
# Calculate AUC-ROC for training set
auc_train <- roc(df_train_balanced$bad_flag, y_pred_proba_train)## Setting levels: control = 0, case = 1## Setting direction: controls < casesprint(paste('AUC train probability:', auc_train$auc))## [1] "AUC train probability: 0.957483917229818"# Calculate AUC-ROC for test set
auc_test <- roc(df_test$bad_flag, y_pred_proba_test)## Setting levels: control = 0, case = 1
## Setting direction: controls < casesprint(paste('AUC test probability:', auc_test$auc))## [1] "AUC test probability: 0.956787299471255"Diperoleh Nilai AUC yang sangat baik untuk data training dan data testing. Hal ini menunjukkan bahwa model memiliki kemampuan yang baik dalam melakukan klasifikasi antara dua kelas.
library(caret)## Loading required package: lattice# Function to calculate precision, recall, and F1 score
calculate_metrics <- function(actual, predicted) {
confusion_matrix <- confusionMatrix(factor(predicted, levels = c(0, 1)), factor(actual, levels = c(0, 1)))
precision <- confusion_matrix$byClass["Pos Pred Value"]
recall <- confusion_matrix$byClass["Sensitivity"]
f1_score <- 2 * (precision * recall) / (precision + recall)
metrics <- c(precision = precision, recall = recall, f1_score = f1_score)
return(metrics)
}
# Apply the function to training set
metrics_train <- calculate_metrics(df_train_balanced$bad_flag, ifelse(y_pred_proba_train > 0.5, 1, 0))
print("Metrics on Training Set:")## [1] "Metrics on Training Set:"print(metrics_train)## precision.Pos Pred Value recall.Sensitivity f1_score.Pos Pred Value
## 0.8694850 0.9409065 0.9037869# Apply the function to test set
metrics_test <- calculate_metrics(df_test$bad_flag, ifelse(y_pred_proba_test > 0.5, 1, 0))
print("Metrics on Test Set:")## [1] "Metrics on Test Set:"print(metrics_test)## precision.Pos Pred Value recall.Sensitivity f1_score.Pos Pred Value
## 0.9815343 0.9406843 0.9606752# Convert predicted probabilities to class labels using a threshold of 0.5
predicted_labels_train <- ifelse(y_pred_proba_train > 0.5, 1, 0)
predicted_labels_test <- ifelse(y_pred_proba_test > 0.5, 1, 0)
# Confusion matrices
conf_matrix_train <- confusionMatrix(factor(predicted_labels_train, levels = c(0, 1)),
factor(df_train_balanced$bad_flag, levels = c(0, 1)))
conf_matrix_test <- confusionMatrix(factor(predicted_labels_test, levels = c(0, 1)),
factor(df_test$bad_flag, levels = c(0, 1)))
# Function to create a classification report
classification_report <- function(conf_matrix) {
precision <- conf_matrix$byClass["Pos Pred Value"]
recall <- conf_matrix$byClass["Sensitivity"]
f1_score <- conf_matrix$byClass["F1"]
support <- table(factor(df_test$bad_flag, levels = c(0, 1)))
report <- data.frame(Precision = precision,
Recall = recall,
F1_Score = f1_score,
Support = support)
return(report)
}
# Create classification reports for training and test sets
report_train <- classification_report(conf_matrix_train)## Warning in data.frame(Precision = precision, Recall = recall, F1_Score =
## f1_score, : row names were found from a short variable and have been discardedreport_test <- classification_report(conf_matrix_test)## Warning in data.frame(Precision = precision, Recall = recall, F1_Score =
## f1_score, : row names were found from a short variable and have been discarded# Print the reports
print("Classification Report for Training Set:")## [1] "Classification Report for Training Set:"print(report_train)## Precision Recall F1_Score Support.Var1 Support.Freq
## 1 0.869485 0.9409065 0.9037869 0 83064
## 2 0.869485 0.9409065 0.9037869 1 10194print("Classification Report for Test Set:")## [1] "Classification Report for Test Set:"print(report_test)## Precision Recall F1_Score Support.Var1 Support.Freq
## 1 0.9815343 0.9406843 0.9606752 0 83064
## 2 0.9815343 0.9406843 0.9606752 1 10194# Classification Table for training set
table_train <- table(Reference = df_train_balanced$bad_flag, Prediction = predicted_labels_train)
print("Classification Table - Training Set:")## [1] "Classification Table - Training Set:"print(table_train)## Prediction
## Reference 0 1
## 0 312619 19634
## 1 46926 286124# Calculate accuracy for training set using table_train
accuracy_train <- sum(diag(table_train)) / sum(table_train)
print(paste("Accuracy - Training Set:", accuracy_train))## [1] "Accuracy - Training Set: 0.899955358686193"# Classification Table for test set
table_test <- table(Reference = df_test$bad_flag, Prediction = predicted_labels_test)
print("Classification Table - Test Set:")## [1] "Classification Table - Test Set:"print(table_test)## Prediction
## Reference 0 1
## 0 78137 4927
## 1 1470 8724# Calculate accuracy for training set using table_train
accuracy_test <- sum(diag(table_test)) / sum(table_test)
print(paste("Accuracy - Testing Set:", accuracy_test))## [1] "Accuracy - Testing Set: 0.931405348602801"Diperoleh Model Binary ogistic Regression dengan tingkat akurasi 93% pada data testing.