Project Overview This project applies a cluster-based modeling approach on the UCI Credit Card Default dataset to predict the likelihood of a client defaulting on their payment. Instead of training a single global model, clients are first segmented into behavioral groups using HDBSCAN, then a dedicated prediction model is trained for each cluster. This approach yields more accurate predictions and cluster-specific financial risk insights.
Dataset Size: 30,000 clients, 23 features Target: DEFAULT — whether a client will default next month (binary: 0/1) Features: Credit limit, payment history (6 months), bill amounts, payment amounts, demographics
