Explainable Machine Learning Framework for Financial Fraud Detection and Risk Analysis

Project Synopsis & Overview

This project develops an explainable machine learning system for detecting potentially fraudulent financial transactions. Transaction data is preprocessed through feature engineering, timestamp extraction, categorical encoding, and class imbalance handling. A Random Forest classifier predicts fraudulent and legitimate transactions, while SHAP explains individual predictions and feature importance. A Streamlit interface provides fraud probability, prediction results, and understandable transaction-level explanations for users.

Complete Technical Specifications