Employee Attrition Prediction

Q1. Employee Attrition Prediction Using Regression

a. Suitable Model Selection

Logistic Regression is suitable because attrition is a binary outcome (“Yes”/“No”).
It estimates the probability of an employee leaving based on predictors like Age, Salary, and YearsAtJob.

b. Model Implementation in R

library(caret)

## Warning: package 'caret' was built under R version 4.5.2

## Loading required package: ggplot2

## Loading required package: lattice

library(pROC)

## Warning: package 'pROC' was built under R version 4.5.2

## Type 'citation("pROC")' for a citation.

## 
## Attaching package: 'pROC'

## The following objects are masked from 'package:stats':
## 
##     cov, smooth, var

set.seed(1)

data <- data.frame(
  Age = c(25, 30, 45, 28, 35, 40, 50, 29, 31, 38),
  Salary = c(30000, 40000, 60000, 35000, 45000, 55000, 65000, 37000, 42000, 50000),
  YearsAtJob = c(1, 3, 10, 2, 5, 8, 12, 3, 4, 7),
  Attrition = c("Yes", "No", "Yes", "No", "No", "Yes", "No", "Yes", "No", "Yes")
)

# Convert target variable
data$Attrition <- as.factor(ifelse(data$Attrition == "Yes", 1, 0))

# Fit logistic regression model
model <- glm(Attrition ~ Age + Salary + YearsAtJob, data = data, family = binomial)

## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

summary(model)

## 
## Call:
## glm(formula = Attrition ~ Age + Salary + YearsAtJob, family = binomial, 
##     data = data)
## 
## Coefficients:
##               Estimate Std. Error z value Pr(>|z|)
## (Intercept)  1.718e+03  4.613e+05   0.004    0.997
## Age         -5.430e+01  1.457e+04  -0.004    0.997
## Salary      -1.801e-02  4.856e+00  -0.004    0.997
## YearsAtJob   1.806e+02  4.856e+04   0.004    0.997
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 13.8629  on 9  degrees of freedom
## Residual deviance:  6.5064  on 6  degrees of freedom
## AIC: 14.506
## 
## Number of Fisher Scoring iterations: 20

# Predictions
data$Predicted_Prob <- predict(model, type="response")
data$Predicted_Class <- ifelse(data$Predicted_Prob > 0.5, 1, 0)

# Confusion Matrix
confusionMatrix(as.factor(data$Predicted_Class), data$Attrition)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction 0 1
##          0 4 0
##          1 1 5
##                                          
##                Accuracy : 0.9            
##                  95% CI : (0.555, 0.9975)
##     No Information Rate : 0.5            
##     P-Value [Acc > NIR] : 0.01074        
##                                          
##                   Kappa : 0.8            
##                                          
##  Mcnemar's Test P-Value : 1.00000        
##                                          
##             Sensitivity : 0.8000         
##             Specificity : 1.0000         
##          Pos Pred Value : 1.0000         
##          Neg Pred Value : 0.8333         
##              Prevalence : 0.5000         
##          Detection Rate : 0.4000         
##    Detection Prevalence : 0.4000         
##       Balanced Accuracy : 0.9000         
##                                          
##        'Positive' Class : 0              
## 

# ROC Curve and AUC
roc_obj <- roc(data$Attrition, data$Predicted_Prob)

## Setting levels: control = 0, case = 1

## Setting direction: controls < cases

plot(roc_obj, col="blue", main="ROC Curve for Employee Attrition Model")

auc(roc_obj)

## Area under the curve: 0.92

# 10-fold Cross Validation
train_control <- trainControl(method="cv", number=10)
cv_model <- train(Attrition ~ Age + Salary + YearsAtJob,
                  data=data,
                  method="glm",
                  family="binomial",
                  trControl=train_control)

## Warning: glm.fit: algorithm did not converge

## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

## Warning in nominalTrainWorkflow(x = x, y = y, wts = weights, info = trainInfo,
## : There were missing values in resampled performance measures.

## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

cv_model

## Generalized Linear Model 
## 
## 10 samples
##  3 predictor
##  2 classes: '0', '1' 
## 
## No pre-processing
## Resampling: Cross-Validated (10 fold) 
## Summary of sample sizes: 9, 9, 9, 9, 9, 9, ... 
## Resampling results:
## 
##   Accuracy  Kappa
##   0.5       0