To perform advanced data analysis with risk identification, you can use various statistical and machine learning techniques in R. Here's an example case study with R code that demonstrates risk identification using logistic regression:
Case Study: Loan Default Prediction
1. Data Preparation:
- Obtain a dataset containing information about loan applicants, including various features such as credit score, income, employment status, loan amount, etc.
- Split the dataset into a training set and a test set.
2. Data Exploration:
- Load the necessary packages:
```R
library(dplyr)
library(ggplot2)
library(corrplot)
```
- Explore the dataset by examining its structure and summary statistics:
```R
# Load the dataset
loan_data <- read.csv("loan_data.csv")
# Overview of the dataset
str(loan_data)
summary(loan_data)
```
- Visualize the relationships between variables and identify potential risk factors:
```R
# Create a correlation matrix
cor_matrix <- cor(loan_data[, c("CreditScore", "Income", "LoanAmount", "Default")])
# Plot a correlation heatmap
corrplot(cor_matrix, method = "color", type = "upper")
```
3. Data Preprocessing:
- Handle missing values and outliers:
```R
# Replace missing values with appropriate imputation techniques
loan_data$CreditScore[is.na(loan_data$CreditScore)] <- mean(loan_data$CreditScore, na.rm = TRUE)
# Identify and handle outliers
outlier_threshold <- quantile(loan_data$LoanAmount, c(0.01, 0.99))
loan_data$LoanAmount[loan_data$LoanAmount < outlier_threshold[1]] <- outlier_threshold[1]
loan_data$LoanAmount[loan_data$LoanAmount > outlier_threshold[2]] <- outlier_threshold[2]
```
- Encode categorical variables:
```R
# Convert categorical variables into factors
loan_data$EmploymentStatus <- as.factor(loan_data$EmploymentStatus)
```
4. Model Development - Logistic Regression:
- Split the data into a training set and a test set:
```R
set.seed(123)
train_indices <- sample(1:nrow(loan_data), 0.7 * nrow(loan_data))
train_data <- loan_data[train_indices, ]
test_data <- loan_data[-train_indices, ]
```
- Train a logistic regression model:
```R
# Build the logistic regression model
model <- glm(Default ~ ., data = train_data, family = "binomial")
# View the model summary
summary(model)
```
5. Model Evaluation:
- Predict on the test set and evaluate the model performance:
```R
# Make predictions on the test set
test_data$predicted_prob <- predict(model, newdata = test_data, type = "response")
# Create a binary prediction based on a probability threshold
threshold <- 0.5
test_data$predicted_default <- ifelse(test_data$predicted_prob >= threshold, 1, 0)
# Evaluate the model performance
confusion_matrix <- table(test_data$Default, test_data$predicted_default)
accuracy <- sum(diag(confusion_matrix)) / sum(confusion_matrix)
precision <- confusion_matrix[2, 2] / sum(confusion_matrix[, 2])
recall <- confusion_matrix[2, 2] / sum(confusion_matrix[2, ])
f1_score <- 2 * precision * recall / (precision + recall
No comments:
Post a Comment