Rogerlytics/loan-default-prediction-kenya

Predicting loan default risk for logbook loans using machine learning and Kenyan market data. # 💰 Loan Default Prediction Model (Kenya) ## 📌 Project Overview This project builds a machine learning model to predict loan default risk for logbook loans in Kenya. The goal is to help financial institutions identify high-risk borrowers and reduce lending losses. --- ## 🎯 Business Problem Lenders in Kenya face significant risk when issuing logbook loans. This project uses data analysis and machine learning to: * Identify high-risk customers * Understand factors influencing default * Improve lending decisions --- ## 📊 Dataset The dataset simulates real-world lending scenarios and includes: * Customer income * Employment type * Loan details * Vehicle information * Previous loan history --- ## 🛠 Tools & Technologies * Python (Pandas, NumPy) * Seaborn & Matplotlib (Visualization) * Scikit-learn (Machine Learning) * Jupyter Notebook / Google Colab --- ## 🔍 Exploratory Data Analysis (EDA) Key findings: * Lower income customers have higher default rates * Informal employment increases risk * Older vehicles correlate with higher defaults * High loan-to-value ratio strongly increases risk --- ## 🤖 Machine Learning Models ### 1. Logistic Regression * Baseline model for prediction ### 2. Random Forest (Final Model) * Improved accuracy * Captures complex patterns --- ## 📈 Key Insights * Loan-to-value ratio is the strongest predictor * Previous defaults significantly increase risk * Income level affects repayment ability --- ## 💼 Business Impact This model can help: * Reduce loan default rates * Improve credit risk assessment * Support better lending decisions --- ## 📂 Project Structure ``` loan-default-prediction-kenya/ │ ├── data/ ├── notebooks/ ├── visuals/ ├── src/ └── README.md ``` --- ## 🚀 Future Improvements * Use real financial datasets * Deploy model as a web app * Add more advanced models (XGBoost) --- ## 👤 Author Data Analyst Portfolio Project ## 📓 Project Notebook View full analysis in the notebook: Open Notebook