AI

Most machine learning techniques quietly assume that rows in a dataset are independent and identically distributed, so shuffling them changes nothing about what a model can learn. Time series data breaks that assumption on purpose: order is the signal. This course covers what makes sequential data fundamentally different, the classical toolkit built specifically to handle it (moving averages, exponential smoothing, ARIMA), why the standard train/test split from other ML courses on this site is actively wrong here, how to backtest a forecaster the way it will actually be used, and when a modern deep learning or LLM-based approach earns its extra complexity versus when a well-tuned classical baseline just wins.