Business analytics
Forecasting
Carry a seasonal series forward with a range, after checking the method on a season it never saw.
A seasonal series carried forward with an honest range, after the method proves itself on a season it never saw.
Forecast
Holt-Winters exponential smoothing carries a series’ level, trend and seasonal pattern forward, with a range for how far off the forecast could be. Before trusting it, the tool hides the last season, forecasts it, and compares the result with simply repeating last year.
periods
12 for months in a year, 4 for quarters, 52 for weeks, 7 for days in a week.
Seasonal swings
—next period
- Error on the hidden season
- —
- Repeating last season
- —
- Hidden values inside the 95% range
- —
- Seasonal swings
- —
The shaded band is where 95% of future values should fall if the pattern holds. It widens with every step ahead.
Sources
- Hyndman, R. J. & Athanasopoulos, G. Forecasting: Principles and Practice, 2nd ed., chapter 7 (Holt-Winters); prediction intervals for the additive model after Hyndman, Koehler, Ord & Snyder (2008), Forecasting with Exponential Smoothing. Seasonal swings that grow with the level are modelled on logarithms.
- Statistics here are computed in your browser and checked against SciPy and statsmodels in the site’s automated tests.
- All example datasets are synthetic, generated for this page.