FML Assignment Demo

Predicting Credit Card Default: Logistic Regression vs Random Forest | MSc AI, SLIIT Presented by: Member 1: H.K.T. Bamunuge Member 2: C.P. Madubashana This video demonstrates our group assignment for the Fundamentals of Machine Learning module of the MSc in Artificial Intelligence at the Sri Lanka Institute of Information Technology. We predict whether a credit card customer will miss next month's payment, using 30,000 real customer records from a Taiwanese bank. We compare two supervised learning algorithms: Logistic Regression (a linear model) and Random Forest (a tree ensemble). Key results on 6,000 unseen test customers: • Random Forest: ROC AUC 0.778, F1 0.538 • Logistic Regression: ROC AUC 0.711, F1 0.464 • Recent repayment history was the strongest predictor of default for both models • Only 22% of customers default, so accuracy on its own is misleading. We evaluate with precision, recall, F1, and ROC AUC, and tune the decision threshold.