Forecasting & Time Series

Time series data breaks many of the assumptions ordinary machine learning relies on, which is why forecasting deserves its own study. These courses cover decomposition into trend and seasonality, stationarity and differencing, and the model families that follow: ARIMA and SARIMA, exponential smoothing, Prophet, and sequence models such as LSTMs. Just as important is the evaluation side, where you learn backtesting with rolling origins, why a naive baseline is the benchmark to beat, and how forecast error translates into inventory or budget cost. Relevant to anyone forecasting sales, demand, headcount, energy load or financial series.