Credit Risk Analysis Mini Project

Overview

This project aims to analyze credit risk using a dataset containing various customer attributes. The goal is to predict whether a person has good or bad credit risk based on their characteristics, employing logistic regression for classification.

Results and Conclusion

This model serves as a valuable tool for banks, enabling them to efficiently assess potential credit risks based on readily available customer information. By identifying high-risk customers early in the process, banks can save time and resources before conducting more detailed evaluations. The model achieved an accuracy of approximately 78% on the test set, with a confusion matrix revealing the performance across different classes of credit risk. This project provides a foundational approach to credit risk analysis using logistic regression. Future improvements could include experimenting with more advanced models and feature engineering techniques.

Table of Contents

• Data Description
• Data Visualization
• Data Management
• Model Training
• Installation
• Usage

Data Description

Two datasets are utilized in this project:

1. credit_s.csv: Contains customer attributes but lacks a target variable for credit risk.

   • Features include:
     • Age (numeric)
     • Sex (text): male, female
     • Job (numeric): 0 - unskilled and non-resident, 1 - unskilled and resident, 2 - skilled, 3 - highly skilled
     • Housing (text): own, rent, or free
     • Saving (text): little, moderate, quite rich, rich
     • Checking (text): little, moderate, rich
     • Credit_amount (numeric, in DM)
     • Duration (numeric, in months)
     • Purpose (text): various categories

2. credit_g.csv: Contains customer attributes along with the target variable for credit risk.

   • Key columns after renaming:
     • status: Credit risk status (1 for good, 2 for bad)
     • Duration: Duration of the credit
     • Credit_amount: Amount of credit
     • Credit_risk: Target variable

Data Visualization

The following visualizations are created to understand the dataset better:

• Pie chart for Credit_risk
• Histogram for Age
• Boxplot comparing Age distributions by Credit_risk
• Bar plot comparing Credit_risk across different Sex groups
• Scatterplot showing the relationship between Age and Credit_amount colored by Credit_risk

Data Management

• Summary statistics for key features are calculated.
• Missing values are handled by encoding them as a new category: "Unknown."
• Categorical variables are converted to dummy variables.

Model Training

A logistic regression model is trained using the processed dataset:

• The data is split into training and testing sets, reserving 20% for testing.

X_train, X_test, y_train, y_test = train_test_split(features_standardized, target, random_state=99,test_size=0.2 )

• Cross-validation is used to find the optimal model parameters.

logistic_regression = LogisticRegressionCV(cv=5,solver='lbfgs', 
                                    multi_class='multinomial', penalty='l2', 
                                           Cs=20, random_state=100, n_jobs=-1)

Installation

To run this project, you will need Python and the following libraries:

• pandas
• numpy
• matplotlib
• seaborn
• plotly
• scikit-learn

Usage

1. Clone this repository to your local machine.
2. Update the file paths in the code to point to your datasets (credit_s.csv and credit_g.csv).
3. Run the Python script to execute the analysis.