Exponential Smoothing and Holt-Winters
Explore exponential smoothing methods such as Simple Exponential Smoothing, Holt’s linear trend, and Holt-Winters seasonal models to forecast time series data. Understand how to adjust smoothing parameters and choose the appropriate model based on level, trend, and seasonality in your data. Learn to evaluate forecasts using RMSE and apply models to real-world demand scenarios for accurate replenishment planning.
For a single retail SKU, tomorrow’s replenishment quantity needs a forecast that responds to a real demand shift without treating one unusual day as a new normal. Unlike the SARIMA approach used previously, exponential smoothing represents the series through components that are easy to inspect: level, trend, and seasonality.
Exponential smoothing forecasts by combining past observations with unequal weights. Recent daily sales receive more weight than older sales, and those weights shrink exponentially as observations move further into the past. A sale of 42 units yesterday therefore affects the next forecast more than a sale of 42 units three weeks ago.
The level forecast updates after each observation. Its responsiveness is controlled by the smoothing parameter 0 and 1 for level smoothing. A larger
The tradeoff matters for replenishment. A high