This project presents a complete machine learning pipeline to predict the likelihood of loan default (credit risk) using supervised classification models. It includes data preprocessing, handling class imbalance, model training, evaluation, and ROC curve analysis.
credit_risk_modeling_assessment.ipynb: Main Jupyter notebook with end-to-end workflowdata/: Folder containing the credit risk dataset (optional if hosted externally)
- Logistic Regression
- Random Forest
- XGBoost
- Exploratory Data Analysis (EDA)
- One-hot encoding of categorical features
- Standardization of numerical features
- Upsampling of minority class to handle class imbalance
- Model evaluation using accuracy, precision, recall, F1-score, and ROC-AUC
- ROC Curve comparison and feature importance analysis
- Best Accuracy: 93%
- Best ROC-AUC: 0.93
- Recall for Defaults Improved via upsampling and feature engineering
- Python, Pandas, Scikit-learn, XGBoost, Seaborn, Matplotlib, Jupyter
Aditya Akilesh Mantha