Siphesihle-Merile/credit-risk-scorer

# Credit Risk Scorer A machine learning project that predicts whether a loan applicant is a good or bad credit risk — built using the German Credit Risk dataset. ## Project Overview Financial institutions face significant losses from high-risk loans. This project builds and compares two ML models to predict credit risk, identifying key factors that drive loan defaults. ## Dataset - **Source:** German Credit Risk Dataset (UCI Machine Learning Repository) - **Size:** 1,000 loan applicants - **Features:** Age, Sex, Job, Housing, Saving Accounts, Checking Account, Credit Amount, Duration, Purpose ## Models Built | Model | Accuracy | |-------|----------| | Logistic Regression | 83.50% | | Random Forest | 84.00% | ## Key Findings - **Random Forest** outperformed Logistic Regression - **Top 3 predictors of credit risk:** 1. Credit Amount 2. Duration 3. Age - Dataset imbalance (87% good vs 13% bad) highlights the importance of looking beyond accuracy alone ## Technologies Used - Python - Pandas, NumPy - Scikit-learn - Matplotlib, Seaborn ## Business Relevance This type of model is directly applicable to: - Bank loan approval systems - Enterprise risk management - Financial advisory services ## Author **Siphesihle Merile** BSc Computer Science — University of the Witwatersrand LinkedIn | GitHub