🌟 Credit Prediction Machine Learning API

🚀 Project Overview

This project is a state-of-the-art Machine Learning API designed to assist financial institutions with credit risk assessment. It leverages advanced neural networks for:

• Credit Card Default Prediction: Assess the probability of a customer defaulting on their credit card payments.
• Credit Limit Recommendation: Estimate an optimal credit limit based on customer financial behavior.

With a clean architecture and scalable design, this API bridges the gap between machine learning models and practical financial solutions.

---

🔧 Tech Stack

Category	Technology
Programming Language	Python
Machine Learning	TensorFlow/Keras, Scikit-learn
Data Manipulation	Pandas, NumPy
API Framework	FastAPI
Data Scaling	StandardScaler, Random Oversampling

---

📊 Features

Default Prediction Model

• Binary Classification: Predicts the likelihood of credit card default.
• Probability Output: Provides interpretable percentage risk scores.
• Class Balancing: Handles imbalanced datasets using Random Oversampling.

Credit Limit Prediction Model

• Regression-based: Estimates appropriate credit limits for customers.
• Regularization: Incorporates techniques to avoid overfitting.
• Custom Feature Engineering: Tailored financial predictors for better performance.

---

🛠️ Core Capabilities

Machine Learning Highlights

• Deep Neural Networks: Custom layers with dropout and regularization.
• Feature Scaling: Scaled inputs ensure consistent model performance.
• Model Evaluation: Metrics like Mean Squared Error and Classification Accuracy guide training.

API Development

• FastAPI Framework: Efficient, asynchronous API handling.
• Pydantic for Validation: Validates incoming request payloads.
• Robust Error Handling: Ensures predictable and stable responses.

---

🚦 Getting Started

Prerequisites

Ensure you have the following installed:

• Python 3.8+
• pip (Python package manager)
• Virtual environment (recommended)

Installation

1. Clone the repository:

   git clone https://github.com/austinLorenzMccoy/credit-prediction-api  
   cd credit-prediction-api  

2. Install dependencies:

   pip install -r requirements.txt  

Running the API

To start the API server locally:

python cli.py api  

📡 API Endpoints

1️⃣ Default Prediction

• URL: /api/v1/predict/default
• Method: POST
• Input: Simplified customer financial data
• Output: Default risk as a percentage probability with risk factors

2️⃣ Credit Limit Prediction

• URL: /api/v1/predict/credit-limit
• Method: POST
• Input: Simplified customer financial data
• Output: Recommended credit limit with adjustment factors

3️⃣ Health Check

• URL: /api/v1/health
• Method: GET
• Output: API health status and model status

🔍 Example API Usage with curl

Check API Health

curl -X GET http://localhost:8000/api/v1/health

Response:

{"status":"healthy","model_status":{"predictor_ready":true}}

Predict Default Risk

curl -X POST http://localhost:8000/api/v1/predict/default \
  -H "Content-Type: application/json" \
  -d '{
    "credit_limit": 100000,
    "age": 25,
    "gender": 2,
    "education": 2,
    "marital_status": 1,
    "payment_status": 0,
    "bill_amount": 10000,
    "payment_amount": 2000
  }'

Response:

{"prediction":"Low Risk of Default (Probability: 15.00%)","probability":0.15,"is_high_risk":false,"risk_factors":["Low payment to bill ratio"]}

Predict Credit Limit

curl -X POST http://localhost:8000/api/v1/predict/credit-limit \
  -H "Content-Type: application/json" \
  -d '{
    "credit_limit": 100000,
    "age": 35,
    "gender": 2,
    "education": 1,
    "marital_status": 1,
    "payment_status": 0,
    "bill_amount": 10000,
    "payment_amount": 8000
  }'

Response:

{"predicted_credit_limit":150000.0,"adjustment_factor":1.5,"recommendation_factors":["Good payment history","High payment to bill ratio","Age factor positive","Education level positive"]}

📊 Input Parameters Explained

Parameter	Description	Values
credit_limit	Current credit limit	Numeric value
age	Customer age	Integer
gender	Customer gender	1=male, 2=female
education	Education level	1=graduate, 2=university, 3=high school, 4=others
marital_status	Marital status	1=married, 2=single, 3=others
payment_status	Last month's payment status	0=paid duly, 1=1 month delay, 2=2 months delay, etc.
bill_amount	Last month's bill amount	Numeric value
payment_amount	Last month's payment amount	Numeric value

---

📡 API Endpoints

1️⃣ Default Prediction

• URL: /predict/default
• Method: POST
• Input: Customer's financial features (list of floats)
• Output: Default risk as a percentage probability

2️⃣ Credit Limit Prediction

• URL: /predict/credit-limit
• Method: POST
• Input: Customer's age, payment history, and bill amount
• Output: Recommended credit limit (numeric)

---

📈 Model Performance

Default Prediction Model

• Accuracy: Validated on a balanced dataset with enhanced features.

Credit Limit Prediction Model

• Metric: Mean Squared Error minimized through iterative fine-tuning.

---

📂 Project Structure

credit-prediction-api/  
├── app/  
│   ├── app.py             # FastAPI application  
│   ├── main.py            # Model training script  
│   ├── models/            # Pre-trained models  
│   ├── scalers/           # Saved scalers  
├── data/  
│   ├── raw/               # Original datasets  
├── notebooks/             # Jupyter notebooks for EDA and model training  
├── frontend/              # HTML, CSS, and JS files for the user interface  
├── tests/                 # Unit tests for endpoints  
├── requirements.txt       # Python dependencies  
├── README.md              # Project documentation  
└── .env                   # Environment variables  

---

🔒 Data Privacy

• Models trained on anonymized data.
• No sensitive information is stored or processed beyond the scope of prediction tasks.

---

🤝 Contributing

We welcome contributions! To get started:

1. Fork this repository.
2. Create a feature branch: git checkout -b feature-name.
3. Commit your changes: git commit -m "Add new feature".
4. Push the branch: git push origin feature-name.
5. Open a pull request for review.

---

📄 License

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

---

👥 Contact

• GitHub: @austinLorenzMccoy
• Email: chibuezeaugustine23@gmail.com