Darth-Glennium/Mpesa-Loan-Default-Model

A Streamlit web app for predicting loan default risk using M-Pesa statements. # M-Pesa Credit Risk Scoring System ### Loan Default Prediction via Alternative Financial Data --- ## Overview This project is a machine learning-based credit risk scoring system that predicts the probability of loan default using alternative financial data extracted from M-Pesa mobile money statements. The system demonstrates how unstructured financial transaction data (PDF statements) can be transformed into structured behavioral features for use in predictive credit modeling. It is designed to simulate real-world underwriting systems used in fintech lending environments. The solution is deployed as an interactive **Streamlit** application that allows users to upload M-Pesa statements and receive real-time credit risk predictions alongside interpretable financial insights. --- ## Problem Statement In many emerging markets, traditional credit scoring systems are limited due to a lack of formal banking history. As a result, mobile money transaction data has become a critical alternative source for assessing creditworthiness. However, raw transaction data is unstructured and not directly usable for machine learning models. This project addresses the challenge of: - Extracting meaningful financial signals from mobile money statements - Engineering behavioral credit features from transaction data - Building a predictive model for loan default risk estimation --- ## Key Objectives - Extract financial features from M-Pesa PDF statements - Engineer behavioral indicators of financial health - Train a supervised machine learning model for default prediction - Provide interpretable risk insights for lending decisions - Deploy the model as an interactive web application --- ## Features ### PDF Data Extraction Parses M-Pesa statements using `pdfplumber`. ### Financial Feature Engineering - Income estimation - Expense tracking - Loan disbursement detection - Repayment behavior analysis - Cashflow stability indicators ### Credit Risk Indicators - Income-to-e …