I have a dataset called attrition_raw_tbl that looks like this.
library(tidyverse)## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ ggplot2 3.5.1 ✔ tibble 3.2.1
## ✔ lubridate 1.9.3 ✔ tidyr 1.3.1
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorsattrition_raw_tbl <- read_csv("00_data/WA_Fn-UseC_-HR-Employee-Attrition.csv")## Rows: 1470 Columns: 35
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (9): Attrition, BusinessTravel, Department, EducationField, Gender, Job...
## dbl (26): Age, DailyRate, DistanceFromHome, Education, EmployeeCount, Employ...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.attrition_raw_tbl %>% slice %>% glimpse## Rows: 0
## Columns: 35
## $ Age <dbl>
## $ Attrition <chr>
## $ BusinessTravel <chr>
## $ DailyRate <dbl>
## $ Department <chr>
## $ DistanceFromHome <dbl>
## $ Education <dbl>
## $ EducationField <chr>
## $ EmployeeCount <dbl>
## $ EmployeeNumber <dbl>
## $ EnvironmentSatisfaction <dbl>
## $ Gender <chr>
## $ HourlyRate <dbl>
## $ JobInvolvement <dbl>
## $ JobLevel <dbl>
## $ JobRole <chr>
## $ JobSatisfaction <dbl>
## $ MaritalStatus <chr>
## $ MonthlyIncome <dbl>
## $ MonthlyRate <dbl>
## $ NumCompaniesWorked <dbl>
## $ Over18 <chr>
## $ OverTime <chr>
## $ PercentSalaryHike <dbl>
## $ PerformanceRating <dbl>
## $ RelationshipSatisfaction <dbl>
## $ StandardHours <dbl>
## $ StockOptionLevel <dbl>
## $ TotalWorkingYears <dbl>
## $ TrainingTimesLastYear <dbl>
## $ WorkLifeBalance <dbl>
## $ YearsAtCompany <dbl>
## $ YearsInCurrentRole <dbl>
## $ YearsSinceLastPromotion <dbl>
## $ YearsWithCurrManager <dbl>The goal is to help predict attrition for employees.
Please write R code to create a predictive model that predicts the probability of attrition.
# Load required libraries
library(tidymodels)## ── Attaching packages ────────────────────────────────────── tidymodels 1.2.0 ──## ✔ broom 1.0.7 ✔ rsample 1.2.1
## ✔ dials 1.2.1 ✔ tune 1.2.1
## ✔ infer 1.0.7 ✔ workflows 1.1.4
## ✔ modeldata 1.4.0 ✔ workflowsets 1.1.0
## ✔ parsnip 1.2.1 ✔ yardstick 1.3.1
## ✔ recipes 1.1.0## Warning: package 'broom' was built under R version 4.3.3## Warning: package 'modeldata' was built under R version 4.3.3## Warning: package 'recipes' was built under R version 4.3.3## ── Conflicts ───────────────────────────────────────── tidymodels_conflicts() ──
## ✖ scales::discard() masks purrr::discard()
## ✖ dplyr::filter() masks stats::filter()
## ✖ recipes::fixed() masks stringr::fixed()
## ✖ dplyr::lag() masks stats::lag()
## ✖ yardstick::spec() masks readr::spec()
## ✖ recipes::step() masks stats::step()
## • Search for functions across packages at https://www.tidymodels.org/find/library(h2o)##
## ----------------------------------------------------------------------
##
## Your next step is to start H2O:
## > h2o.init()
##
## For H2O package documentation, ask for help:
## > ??h2o
##
## After starting H2O, you can use the Web UI at http://localhost:54321
## For more information visit https://docs.h2o.ai
##
## ----------------------------------------------------------------------##
## Attaching package: 'h2o'## The following objects are masked from 'package:lubridate':
##
## day, hour, month, week, year## The following objects are masked from 'package:stats':
##
## cor, sd, var## The following objects are masked from 'package:base':
##
## &&, %*%, %in%, ||, apply, as.factor, as.numeric, colnames,
## colnames<-, ifelse, is.character, is.factor, is.numeric, log,
## log10, log1p, log2, round, signif, trunc# Step 1: Data Preparation
set.seed(123)
attrition_split <- initial_split(attrition_raw_tbl, prop = 0.8, strata = Attrition)
train_data <- training(attrition_split)
test_data <- testing(attrition_split)
# Step 2: Define Recipe
attrition_recipe <- recipe(Attrition ~ ., data = train_data) %>%
step_rm(all_nominal_predictors()) %>% # Remove single-level factors
step_dummy(all_nominal_predictors(), -all_outcomes()) %>% # One-hot encode remaining categorical variables
step_normalize(all_numeric_predictors())
# Prepare data for H2O
train_processed <- attrition_recipe %>% prep() %>% juice()## Warning: ! The following columns have zero variance so scaling cannot be used:
## EmployeeCount and StandardHours.
## ℹ Consider using ?step_zv (`?recipes::step_zv()`) to remove those columns
## before normalizing.test_processed <- attrition_recipe %>% prep() %>% bake(test_data)## Warning: ! The following columns have zero variance so scaling cannot be used:
## EmployeeCount and StandardHours.
## ℹ Consider using ?step_zv (`?recipes::step_zv()`) to remove those columns
## before normalizing.# Step 3: Initialize H2O
h2o.init()##
## H2O is not running yet, starting it now...
##
## Note: In case of errors look at the following log files:
## /var/folders/x_/s4jnxcsx0fsd3qx1z_c351r80000gn/T//Rtmp6fy9Dt/file87d07c107a6d/h2o_jordanlanowy_started_from_r.out
## /var/folders/x_/s4jnxcsx0fsd3qx1z_c351r80000gn/T//Rtmp6fy9Dt/file87d01dd5c9bd/h2o_jordanlanowy_started_from_r.err
##
##
## Starting H2O JVM and connecting: .... Connection successful!
##
## R is connected to the H2O cluster:
## H2O cluster uptime: 3 seconds 976 milliseconds
## H2O cluster timezone: America/New_York
## H2O data parsing timezone: UTC
## H2O cluster version: 3.44.0.3
## H2O cluster version age: 11 months and 13 days
## H2O cluster name: H2O_started_from_R_jordanlanowy_rxf816
## H2O cluster total nodes: 1
## H2O cluster total memory: 1.77 GB
## H2O cluster total cores: 8
## H2O cluster allowed cores: 8
## H2O cluster healthy: TRUE
## H2O Connection ip: localhost
## H2O Connection port: 54321
## H2O Connection proxy: NA
## H2O Internal Security: FALSE
## R Version: R version 4.3.2 (2023-10-31)## Warning in h2o.clusterInfo():
## Your H2O cluster version is (11 months and 13 days) old. There may be a newer version available.
## Please download and install the latest version from: https://h2o-release.s3.amazonaws.com/h2o/latest_stable.html# Convert preprocessed data to H2OFrame
train_h2o <- as.h2o(train_processed)##
|
| | 0%
|
|======================================================================| 100%test_h2o <- as.h2o(test_processed)##
|
| | 0%
|
|======================================================================| 100%# Step 4: Train AutoML Model
aml_model <- h2o.automl(
x = setdiff(names(train_h2o), "Attrition"), # Predictors
y = "Attrition", # Target variable
training_frame = train_h2o,
max_runtime_secs = 90, # Extend runtime to 10 minutes
seed = 123,
balance_classes = TRUE # Handle class imbalance
)##
|
| | 0%
|
|= | 2%
## 17:53:20.302: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|=== | 4%
## 17:53:24.574: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|===== | 7%
## 17:53:26.567: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|====== | 9%
|
|======== | 11%
## 17:53:30.397: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
## 17:53:31.24: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|========== | 14%
## 17:53:32.184: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|=========== | 16%
## 17:53:34.570: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|============= | 18%
## 17:53:35.872: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
## 17:53:36.797: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
## 17:53:37.867: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
|
|=============== | 21%
## 17:53:38.281: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
## 17:53:38.621: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
## 17:53:39.552: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|================ | 23%
## 17:53:41.116: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
## 17:53:42.10: _train param, Dropping bad and constant columns: [StandardHours, EmployeeCount]
|
|================== | 26%
## 17:53:44.35: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
|
|==================== | 28%
## 17:53:44.477: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
|
|===================== | 30%
|
|======================= | 33%
|
|======================== | 35%
|
|========================== | 37%
|
|============================ | 40%
|
|============================= | 42%
|
|=============================== | 44%
|
|================================= | 47%
|
|================================== | 49%
|
|==================================== | 51%
|
|====================================== | 54%
|
|======================================= | 56%
|
|========================================= | 58%
|
|=========================================== | 61%
|
|============================================ | 63%
|
|============================================== | 65%
|
|=============================================== | 68%
|
|================================================= | 70%
|
|=================================================== | 72%
|
|==================================================== | 75%
|
|====================================================== | 77%
|
|======================================================== | 79%
|
|========================================================= | 82%
|
|=========================================================== | 84%
|
|============================================================= | 86%
|
|============================================================== | 89%
|
|================================================================ | 91%
|
|================================================================== | 94%
## 17:54:44.49: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
## 17:54:44.364: _train param, Dropping unused columns: [StandardHours, EmployeeCount]
|
|=================================================================== | 96%
|
|===================================================================== | 98%
|
|======================================================================| 100%# Step 5: Get Leader Model and Summarize Results
leader_model <- aml_model@leader
print(leader_model)## Model Details:
## ==============
##
## H2OBinomialModel: stackedensemble
## Model ID: StackedEnsemble_BestOfFamily_4_AutoML_1_20241203_175320
## Model Summary for Stacked Ensemble:
## key value
## 1 Stacking strategy cross_validation
## 2 Number of base models (used / total) 4/6
## 3 # GBM base models (used / total) 1/1
## 4 # XGBoost base models (used / total) 1/1
## 5 # GLM base models (used / total) 1/1
## 6 # DRF base models (used / total) 1/2
## 7 # DeepLearning base models (used / total) 0/1
## 8 Metalearner algorithm GLM
## 9 Metalearner fold assignment scheme Random
## 10 Metalearner nfolds 5
## 11 Metalearner fold_column NA
## 12 Custom metalearner hyperparameters None
##
##
## H2OBinomialMetrics: stackedensemble
## ** Reported on training data. **
##
## MSE: 0.0512914
## RMSE: 0.226476
## LogLoss: 0.2032248
## Mean Per-Class Error: 0.08309186
## AUC: 0.9846904
## AUCPR: 0.9414842
## Gini: 0.9693809
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## No Yes Error Rate
## No 963 23 0.023327 =23/986
## Yes 27 162 0.142857 =27/189
## Totals 990 185 0.042553 =50/1175
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.361525 0.866310 136
## 2 max f2 0.279409 0.878514 171
## 3 max f0point5 0.473075 0.911493 106
## 4 max accuracy 0.399196 0.958298 126
## 5 max precision 0.902397 1.000000 0
## 6 max recall 0.150036 1.000000 244
## 7 max specificity 0.902397 1.000000 0
## 8 max absolute_mcc 0.361525 0.841078 136
## 9 max min_per_class_accuracy 0.268490 0.929006 176
## 10 max mean_per_class_accuracy 0.268490 0.930112 176
## 11 max tns 0.902397 986.000000 0
## 12 max fns 0.902397 188.000000 0
## 13 max fps 0.005819 986.000000 399
## 14 max tps 0.150036 189.000000 244
## 15 max tnr 0.902397 1.000000 0
## 16 max fnr 0.902397 0.994709 0
## 17 max fpr 0.005819 1.000000 399
## 18 max tpr 0.150036 1.000000 244
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
##
## H2OBinomialMetrics: stackedensemble
## ** Reported on cross-validation data. **
## ** 5-fold cross-validation on training data (Metrics computed for combined holdout predictions) **
##
## MSE: 0.1143952
## RMSE: 0.3382236
## LogLoss: 0.3780821
## Mean Per-Class Error: 0.3051665
## AUC: 0.7521733
## AUCPR: 0.4402968
## Gini: 0.5043466
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## No Yes Error Rate
## No 812 174 0.176471 =174/986
## Yes 82 107 0.433862 =82/189
## Totals 894 281 0.217872 =256/1175
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.216966 0.455319 162
## 2 max f2 0.158981 0.567498 208
## 3 max f0point5 0.377343 0.467980 80
## 4 max accuracy 0.535384 0.856170 32
## 5 max precision 0.854803 1.000000 0
## 6 max recall 0.012075 1.000000 396
## 7 max specificity 0.854803 1.000000 0
## 8 max absolute_mcc 0.237916 0.336953 147
## 9 max min_per_class_accuracy 0.158981 0.698413 208
## 10 max mean_per_class_accuracy 0.158981 0.709754 208
## 11 max tns 0.854803 986.000000 0
## 12 max fns 0.854803 188.000000 0
## 13 max fps 0.008062 986.000000 399
## 14 max tps 0.012075 189.000000 396
## 15 max tnr 0.854803 1.000000 0
## 16 max fnr 0.854803 0.994709 0
## 17 max fpr 0.008062 1.000000 399
## 18 max tpr 0.012075 1.000000 396
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## Cross-Validation Metrics Summary:
## mean sd cv_1_valid cv_2_valid cv_3_valid cv_4_valid
## accuracy 0.824865 0.043497 0.806723 0.838565 0.795918 0.894737
## auc 0.754676 0.034352 0.813168 0.753417 0.744000 0.724341
## err 0.175135 0.043497 0.193277 0.161435 0.204082 0.105263
## err_count 41.400000 11.392981 46.000000 36.000000 50.000000 24.000000
## f0point5 0.485235 0.038597 0.472727 0.510204 0.476190 0.533981
## cv_5_valid
## accuracy 0.788382
## auc 0.738454
## err 0.211618
## err_count 51.000000
## f0point5 0.433071
##
## ---
## mean sd cv_1_valid cv_2_valid cv_3_valid
## precision 0.478172 0.064230 0.440678 0.500000 0.456140
## r2 0.146401 0.032966 0.169370 0.170207 0.162054
## recall 0.546243 0.094023 0.666667 0.555556 0.577778
## residual_deviance 177.698600 22.106504 173.156460 166.093600 204.036830
## rmse 0.337397 0.018619 0.337354 0.335160 0.354457
## specificity 0.875328 0.052675 0.834171 0.893048 0.845000
## cv_4_valid cv_5_valid
## precision 0.578947 0.415094
## r2 0.092174 0.138197
## recall 0.407407 0.523810
## residual_deviance 149.648180 195.557900
## rmse 0.307855 0.352158
## specificity 0.960199 0.844221# Step 6: Evaluate Performance on Test Data
performance <- h2o.performance(leader_model, newdata = test_h2o)
# Print AUC
auc <- h2o.auc(performance)
print(paste("AUC Score:", auc))## [1] "AUC Score: 0.808367071524966"# Shutdown H2O when done
h2o.shutdown(prompt = FALSE)