Loan Default Risk Predictor

This project provides a web-based tool for predicting the probability of loan default using three machine learning models: Logistic Regression, XGBoost, and a Feedforward Neural Network (FFNN). The goal is to assist in evaluating borrower risk based on financial and demographic information. Read the paper here.

Try the Loan Default Risk Predictor on HuggingFace Spaces 🤗

Features

• Interactive Gradio Interface: Easily input borrower data through dropdowns and number fields.
• Multi-Model Predictions: View default probabilities from Logistic Regression, XGBoost, and FFNN models.
• Derived Financial Metrics: Automatically computes ratios like Affordability Ratio, Total Interest, Debt-to-Income Ratio, and Average Borrowed per Credit Line.
• Data Preprocessing: Includes standardization and Box-Cox transformation for skewed features.
• Model Interpretability: Outputs both probability scores and binary classification (Default / No Default) for each model.

Technologies Used

• Python (Pandas, NumPy, Scikit-learn, XGBoost, TensorFlow/Keras)
• Gradio for the web interface
• Joblib / Pickle for model and transformer serialization

Getting Started Locally

Prerequisites

• Python 3.9+

Running the App

1. Clone the repository:

   git clone <repository-url>

2. Install the required packages:

   pip install -r requirements.txt

3. Navigate to the project directory:

   cd loan_default_prediction

4. Run the app:

   python app.py