Loan Default Prediction using Logistic Regression

# File Name: loan_default_prediction.R

# Author: Gerard King - www.gerardking.dev

# Title: Loan Default Prediction using Logistic Regression

# Description: This program uses logistic regression to predict the likelihood of a customer defaulting on a loan.

# It utilizes features such as credit score, income, loan amount, and loan term to classify customers into default or non-default categories.

# Use Cases:

# - Predicting loan default risk for customers

# - Identifying high-risk customers for proactive measures

# - Assessing the creditworthiness of applicants based on various factors

# Audience:

# - Data scientists working in banking and financial services

# - Loan officers and credit analysts in financial institutions

# - Students and researchers learning about machine learning in finance

# Blue Team Uses:

# - Predicting and identifying potential fraudulent loan applications

# - Monitoring the likelihood of loan defaults to reduce credit risk

# - Assessing customer risk profiles for business decisions

# Red Team Uses:

# - Simulating loan defaults to identify weaknesses in the risk assessment models

# - Analyzing various attack strategies on banking systems based on loan performance metrics

# - Detecting patterns in customer behavior that may signal fraudulent loan activity

# Current Date: 2025-03-06

# Load necessary library

library(ggplot2)

# Set the seed for reproducibility

set.seed(505)

# Generate synthetic dataset for loan applicants (features: credit score, income, loan amount, loan term)

n <- 500  # number of applicants

credit_score <- rnorm(n, mean = 650, sd = 50)

income <- rnorm(n, mean = 50000, sd = 10000)

loan_amount <- rnorm(n, mean = 15000, sd = 5000)

loan_term <- sample(c(15, 30), n, replace = TRUE)  # loan term in years

default <- ifelse(credit_score < 600 & income < 40000 & loan_amount > 20000, 1, 0)  # 1 means default, 0 means no default

# Create a data frame

loan_data <- data.frame(credit_score, income, loan_amount, loan_term, default = as.factor(default))

# Fit a logistic regression model

logit_model <- glm(default ~ credit_score + income + loan_amount + loan_term, data = loan_data, family = binomial)

# Make predictions using the logistic regression model

loan_data$predicted_prob <- predict(logit_model, type = "response")

loan_data$predicted_class <- ifelse(loan_data$predicted_prob > 0.5, 1, 0)

# Visualizing the predictions based on credit score and income

ggplot(loan_data, aes(x = credit_score, y = income, color = as.factor(predicted_class))) +

  geom_point(size = 3) +

  labs(title = "Loan Default Prediction: Credit Score vs Income", 

       x = "Credit Score", y = "Income", color = "Predicted Class") +

  scale_color_manual(values = c("red", "green")) +

  theme_minimal()

# Display the logistic regression model summary

cat("Logistic Regression Model Summary:\n")

summary(logit_model)

# Print the current date for reference

cat("Date of execution:", Sys.Date(), "\n")