This document summarizes an analysis of loan data to classify customers as likely to default or make payments on time. It performed data cleaning including missing value imputation and outlier treatment. Feature selection identified important variables. SMOTE addressed class imbalance, resulting in a balanced 50:50 classification model. Stochastic gradient boosting was the best model with 10% misclassification, 0.61 precision and 0.57 recall on the test data. Important variables included delinquencies, loan amounts, and credit history factors.