Loan Risk Predictor

A complete Python package to predict loan default risk using a Decision Tree model. This project is built from a Kaggle notebook and has been structured as a robust, installable package published on PyPI.

Overview

This project provides a full pipeline for a machine learning task, including data loading, preprocessing, model training, and evaluation. It is designed to be both a ready-to-use tool and a clear example of how to convert a data science notebook into a distributable software package following modern best practices like src-layout and pyproject.toml configuration.

Key Features

• Data Preprocessing: Handles missing values, encodes categorical features, and scales numerical data.
• Model Training: Implements a DecisionTreeClassifier from scikit-learn.
• Packaged for Distribution: Published on PyPI and installable with a single pip command.
• Modular Code: Logic is separated into modules for data handling and model training.
• Command-Line & Library Usage: Can be run directly from the terminal or imported into other Python projects.

Technology Stack

• Python 3.7+
• Pandas: For data manipulation and loading.
• Scikit-learn: For data preprocessing and machine learning modeling.
• Numpy: For numerical operations.
• Setuptools & Build: For packaging and distribution.
• GitHub Actions: For Continuous Integration and automated publishing to PyPI.

Installation

You can install the Loan Risk Predictor directly from PyPI:

pip install loan-risk-predictor

Usage

There are two primary ways to use this package.

1. As a Command-Line Tool

The simplest way to run the full training pipeline is to execute the package as a Python module. This will load the included dataset, preprocess it, train the model, and print the evaluation results.

python -m loan_risk_predictor.train

You should see an output like this:

Loading data...
Preprocessing data...
Training model...
Model Accuracy: 0.8754
Classification Report:
              precision    recall  f1-score   support
           0       0.88      0.99      0.93     44123
           1       0.49      0.08      0.14      6277
    accuracy                           0.88     50400
   macro avg       0.68      0.54      0.53     50400
weighted avg       0.83      0.88      0.83     50400
Training complete.

2. As a Python Library

For more advanced use, you can import the functions directly into your own scripts or notebooks. This allows you to integrate the components into a custom workflow.

Here is an example:

# main_script.py
from loan_risk_predictor.data_preprocessing import load_data, preprocess_data
from loan_risk_predictor.model import train_model

def run_custom_pipeline():
    """
    An example of how to use the library functions programmatically.
    """
    print("--- Starting Custom Pipeline ---")
    
    # 1. Load the data
    print("Step 1: Loading data...")
    raw_df = load_data()
    print(f"Loaded dataset with {raw_df.shape} rows.")
    
    # 2. Preprocess the data
    print("\nStep 2: Preprocessing data...")
    X_scaled, y, scaler, encoders = preprocess_data(raw_df)
    print("Data has been cleaned, encoded, and scaled.")
    
    # 3. Train the model
    print("\nStep 3: Training the model...")
    trained_model = train_model(X_scaled, y)
    print("Model training is complete.")
    
    print("\n--- Custom Pipeline Finished ---")
    # You can now use the `trained_model` object for predictions,
    # or the `scaler` and `encoders` for new data.

if __name__ == "__main__":
    run_custom_pipeline()```

Project Structure

This project follows the modern src-layout to ensure a clean and robust package structure.

loan-risk-predictor/
├── src/
│   └── loan_risk_predictor/      # Main package source
│       ├── data/
│       │   └── training.csv      # Packaged data
│       ├── data_preprocessing.py
│       ├── model.py
│       └── train.py
│
├── .github/
│   └── workflows/
│       └── publish-to-pypi.yml   # CI/CD workflow
│
├── notebooks/
│   └── exploration.ipynb         # Original exploratory notebook
│
├── .gitignore
├── LICENSE
├── pyproject.toml                # Main packaging configuration
└── README.md

Contributing

Contributions are welcome! If you have suggestions for improvements or want to fix a bug, please follow these steps:

1. Fork the repository on GitHub.
2. Clone your fork locally: git clone https://github.com/shanujans/loan-risk-prediction.git
3. Create a new branch for your feature: git checkout -b feature/loan-risk-prediction
4. Install dependencies in a virtual environment.
5. Make your changes and commit them with a clear message.
6. Push your branch to your fork: git push origin feature/loan-risk-prediction
7. Create a Pull Request from your fork to the original repository.

Roadmap

Future improvements could include:

• Trying different models (e.g., RandomForest, Gradient Boosting) for comparison.
• Adding hyperparameter tuning using GridSearchCV or RandomizedSearchCV.
• Implementing a prediction script to score new, unseen data.
• Adding a comprehensive test suite with pytest.
• Containerizing the application with Docker.

License

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