Credit Risk Prediction System

An end-to-end Machine Learning project that predicts whether a customer is a Good or Bad credit risk based on financial and demographic attributes.

This project follows a structured ML pipeline including data preprocessing, feature engineering, model training, evaluation, and deployment using Streamlit.

---

Problem Statement

Banks need to identify high-risk customers before approving loans.

The objective of this project is to build a classification model that segments customers into:

• Good Customer (Low Risk)
• Bad Customer (High Risk)

This helps financial institutions reduce default risk and improve decision-making.

---

Dataset

The dataset contains customer financial and demographic information including:

• Debt Ratio
• Monthly Income
• Number of Open Credit Lines
• Real Estate Loans
• Number of Dependents
• Education Level
• Region
• NPA Status (Target Variable)

🎯 Target Variable

• Good → 1
• Bad → 0

---

Machine Learning Pipeline

The project strictly follows a production-style ML workflow:

Data Preprocessing

• Removed invalid rows
• Handled missing values using random sampling
• Log transformation on skewed features
• Outlier treatment using quantile capping
• Train-test split (before feature engineering to prevent data leakage)

---

Feature Engineering

• One-Hot Encoding for nominal features
• Ordinal Encoding for ordered categorical features
• Variance Threshold feature selection
• Hypothesis testing (p-value filtering)
• SMOTE for handling class imbalance
• StandardScaler for feature scaling

---

Model Training

Models implemented:

• K-Nearest Neighbors
• Naive Bayes
• Logistic Regression
• Decision Tree
• Random Forest

---

Model Evaluation

• Accuracy Score
• Confusion Matrix
• Classification Report
• ROC Curve comparison

---

Final Model Selection

Based on performance comparison and ROC analysis, the best performing model was selected and saved as:

credit_card.pkl

---

Deployment

The trained model is deployed using Streamlit.

The web app allows users to:

• Enter financial details
• Automatically scale inputs
• Predict credit risk
• Display risk level with confidence score

Installation

git clone https://github.com/hemamalini0708/Credit-Card-Risk-Prediction.git
cd Credit-Card-Risk-Prediction
pip install -r requirements.txt

Run Locally

pip install -r requirements.txt
streamlit run app/credit_risk_streamlit.py

Tech Stack

• Python
• Pandas
• NumPy
• Scikit-Learn
• Imbalanced-Learn (SMOTE)
• Matplotlib
• Streamlit

credit-risk-prediction/

│
├── data/
├── models/
├── src/
├── app/
├── README.md
├── requirements.txt
└── .gitignore

Business Impact

• Reduces loan default risk
• Supports automated credit approval decisions
• Enables risk-based customer segmentation
• Improves financial risk management strategy

Author

Hema Malini Gangumalla

Aspiring Data Scientist

📧 hemamalinig07@gmail.com

License

MIT License