A comparative study using four machine learning classifiers to predict loan defaults on a dataset of 45,000 records. The study focuses on minimizing false negatives and reports the Random Forest model as the best performer with 90.4% accuracy.
From the README
Loan Default Classification – ML Algorithms Python Project Overview A comparative machine learning study on loan portfolio health, benchmarking four classification algorithms to predict the probability of default on an imbalanced dataset (22% default rate, 45,000 records). The study evaluates model performance with a focus on minimising false negatives — missed defaults — which carry the highest financial risk for lenders. Tools & Technologies • Python — pandas, numpy, scikit-learn, matplotlib, seaborn • Models: Logistic Regression, Decision Tree, Random Forest, AdaBoost • Techniques: One-Hot Encoding, Standard Scaling, IQR Outlier Detection, Feature Importance Analysis, Partial...
Skills
- Jupyter Notebook
