This project analyzes U.S. small-denomination time deposits using classical models and machine learning methods. It evaluates forecast accuracy, volatility behavior, and the impact of monetary policy, showing that flexible models like XGBoost adapt best to post-2022 structural changes.