Credit Risk Analysis using ISLR’s Default Dataset

Step1. Load libraries

library(ISLR)
library(ggplot2)
library(dplyr)
library(pROC)

Step2: Data Preprocessing

# Load the Default dataset from ISLR
data("Default")

# Preview the dataset
head(Default)

# Summary of the dataset
summary(Default)

 default    student       balance           income     
 No :9667   No :7056   Min.   :   0.0   Min.   :  772  
 Yes: 333   Yes:2944   1st Qu.: 481.7   1st Qu.:21340  
                       Median : 823.6   Median :34553  
                       Mean   : 835.4   Mean   :33517  
                       3rd Qu.:1166.3   3rd Qu.:43808  
                       Max.   :2654.3   Max.   :73554  

# Structure of the dataset
str(Default)

'data.frame':   10000 obs. of  4 variables:
 $ default: Factor w/ 2 levels "No","Yes": 1 1 1 1 1 1 1 1 1 1 ...
 $ student: Factor w/ 2 levels "No","Yes": 1 2 1 1 1 2 1 2 1 1 ...
 $ balance: num  730 817 1074 529 786 ...
 $ income : num  44362 12106 31767 35704 38463 ...

# Convert target variable 'default' to a binary numeric variable (1 for 'Yes', 0 for 'No')
Default <- Default %>%
  mutate(default = ifelse(default == "Yes", 1, 0))

# Check the conversion
table(Default$default)

Step3:Train-Test split

# Set seed for reproducibility
set.seed(123)

# Split the dataset
train_idx <- sample(1:nrow(Default), 0.8 * nrow(Default))
train_data <- Default[train_idx, ]
test_data <- Default[-train_idx, ]

Step 4: Logistic Regression Model

# Train the logistic regression model
logit_model <- glm(
  default ~ balance + income + student,
  data = train_data,
  family = binomial(link = "logit")
)

# Summary of the model
summary(logit_model)

Step 5: Model Evaluation

# Make predictions on the test data
test_data$predicted_prob <- predict(logit_model, newdata = test_data, type = "response")

# Convert probabilities to binary predictions
test_data$predicted_class <- ifelse(test_data$predicted_prob > 0.5, 1, 0)

# Confusion Matrix
conf_matrix <- table(Predicted = test_data$predicted_class, Actual = test_data$default)
print(conf_matrix)

         Actual
Predicted    0    1
        0 1928   46
        1    6   20

# Calculate Accuracy
accuracy <- sum(diag(conf_matrix)) / sum(conf_matrix)
print(paste("Accuracy:", round(accuracy, 4)))

[1] "Accuracy: 0.974"

# Calculate AUC
roc_curve <- roc(test_data$default, test_data$predicted_prob)

Setting levels: control = 0, case = 1
Setting direction: controls < cases

auc_value <- auc(roc_curve)
print(paste("AUC:", round(auc_value, 4)))

[1] "AUC: 0.9458"

# Plot the ROC curve
plot(roc_curve, main = "ROC Curve", col = "blue", lwd = 2)

Step 6: Visualizing Results

# Visualize the relationship between balance and default
ggplot(Default, aes(x = balance, fill = factor(default))) +
  geom_histogram(binwidth = 500, position = "dodge") +
  labs(title = "Balance vs Default", x = "Balance", y = "Count", fill = "Default") +
  theme_minimal()

# Visualize the relationship between income and default
ggplot(Default, aes(x = income, fill = factor(default))) +
  geom_histogram(binwidth = 5000, position = "dodge") +
  labs(title = "Income vs Default", x = "Income", y = "Count", fill = "Default") +
  theme_minimal()

Observations: The model performs well, with high accuracy and AUC scores. Balance is the strongest predictor of default, with a clear separation between the two classes. Income shows a weaker relationship with default.

Conclusion:

This analysis demonstrates the application of logistic regression for credit risk prediction. The results indicate that the customer’s balance is the most significant factor in predicting default.