German Credit Risk Analysis

This repository contains an analysis of the German Credit Risk dataset. The objective of this project is to evaluate credit risk and predict whether a customer falls into the "good" or "bad" risk category using machine learning models.

📂 Project Structure

The repository is organized as follows:

.
├── data
│   ├── raw                # Raw data files (original dataset)
│   ├── processed          # Processed data files (encoded)
├── notebooks
│   ├── german_credit_risk_analysis.ipynb  # Jupyter notebook with the analysis
├── reports
│   ├── figures            # Visualizations generated during analysis
├── requirements.txt       # Python dependencies for the project
├── README.md              # Project overview (this file)

📊 Dataset Overview

The German Credit Risk dataset includes information on 1,000 customers with the goal of predicting their credit risk. Key features of the dataset include:

• Target Variable: Risk - indicates whether the customer is a good or bad credit risk.
• Input Variables: The selected attributes are:
  1. Age (numeric)
  2. Sex (text: male, female)
  3. Job (numeric: 0 - unskilled and non-resident, 1 - unskilled and resident, 2 - skilled, 3 - highly skilled)
  4. Housing (text: own, rent, or free)
  5. Saving accounts (text - little, moderate, quite rich, rich)
  6. Checking account (numeric, in DM - Deutsch Mark)
  7. Credit amount (numeric, in DM)
  8. Duration (numeric, in month)
  9. Purpose (text: car, furniture/equipment, radio/TV, domestic appliances, repairs, education, business, vacation/others)

🔍 Analysis Steps

1. Data Preprocessing:

   • Handling missing values in features like Saving accounts and Checking account.

2. Exploratory Data Analysis (EDA):

   • Univariate analysis : plot, treemaps, and creation of categories for Purpose
   • Bivariate analysis
   • Overview : Pairplot

3. Encoding categorical variables for compatibility with machine learning algorithms:

   • Encoding categorical data
   • Correlation heatmap

4. Model Training and Evaluation:

   • Splitting the dataset
   • Standardization
   • Models building:
     • Naive Bayes
     • k-Nearest Neighbors (KNN)
     • XGBoost (XGB)
   • Metrics for evaluation:
     • Accuracy
     • F1-score
     • ROC-AUC

5. Model synthesis and conclusion:

   • ROC Curve
   • The XGBoost model achieved the best performance across all metrics, making it the recommended choice for deployment.

🧰 Installation

1. Clone this repository:

   git clone https://github.com/clemcoste/german_credit_risk.git
   cd german_credit_risk
   

2. Create venv and install the required dependencies:

   python -m venv venv
   pip install -r requirements.txt
   

3. Open the notebook on Visual Studio Code and select venv

📈 Figures

All visualizations and figures generated during the analysis are stored in the reports/figures directory. These include:

• Feature distributions
• Correlation heatmaps
• Model performance comparisons

🛠️ Requirements

For the full list of dependencies, see the requirements.txt file.

Contributions are welcome! Feel free to submit issues or pull requests if you have suggestions for improvement.

📜 License

This project is licensed under the MIT License. See the LICENSE file for more details.