Credit Scoring Model (CIBIL Score Prediction )

This project is a statistical credit scoring model that predicts a customer's likelihood of defaulting on a loan and generates a CIBIL-like score (300-900). The model uses Logistic Regression with features such as credit utilization, on-time payments, and past defaults to assess creditworthiness.

📌 Features

• Preprocesses financial data (categorical & numerical features).
• Handles class imbalance using SMOTE.
• Trains a Logistic Regression model to predict credit risk.
• Calculates feature importance and contribution percentages.
• Normalizes model output to a CIBIL score (300-900).
• Visualizes model performance with an ROC curve and credit score distribution.

---

🚀 Installation

1️⃣ Clone the Repository

git clone https://github.com/your-username/credit-scoring-model.git
cd credit-scoring-model

2️⃣ Install Dependencies

You need Python 3.8+ and the required libraries. Install them using:

pip install -r requirements.txt

3️⃣ Run the Model

python credit_score.py

---

📦 Requirements

The model depends on the following Python libraries:

numpy
pandas
scikit-learn
imblearn
matplotlib
seaborn

You can install them manually with:

pip install numpy pandas scikit-learn imblearn matplotlib seaborn

---

📊 Implementation Steps

1️⃣ Data Preprocessing

• Drops unnecessary columns (e.g., CustomerID).
• Encodes categorical variables.
• Scales numerical data using StandardScaler.

2️⃣ Handling Class Imbalance

• Uses SMOTE (Synthetic Minority Over-sampling Technique) to balance defaulters and non-defaulters.

3️⃣ Train Logistic Regression Model

• Splits data into training/testing sets.
• Trains a Logistic Regression model.
• Evaluates with classification metrics (AUC-ROC, confusion matrix, etc.).

4️⃣ Feature Importance & Contribution

• Extracts feature importance from model coefficients.
• Calculates percentage contribution of each factor to credit score.

5️⃣ Credit Score Normalization

• Converts predicted probabilities into CIBIL score range (300-900).

6️⃣ Model Performance Visualization

• ROC Curve to analyze classification performance.
• Histogram to visualize predicted CIBIL scores.

---

📈 Example Outputs

Feature Contribution (%)

Feature	Contribution (%)
Credit Utilization	35.2%
On-Time Payments	28.4%
Past Defaults	15.8%
Total Debt	10.6%

ROC Curve

AUC Score Example: 0.85 (Higher is better)

---

🛠 Future Improvements

• Using Actual CIBIL Dataset for better accuracy.
• Try advanced models like Random Forest, XGBoost.
• Deploy as an API or Web App.
• Optimize feature selection for better accuracy.

---

📜 License

This project is MIT Licensed.

---

🙌 Contributing

Feel free to contribute by creating issues or pull requests.

---

📩 Contact

For questions, reach out via yuvan7480@gmail.com or create a GitHub issue.

---