Theta Method Forecasting

The Theta method forecasts a series by combining simple exponential smoothing with half of the linear trend, optionally after seasonal adjustment, and is equivalent to SES with drift.

Theta 방법은 단순 지수평활에 선형 추세의 절반을 더해(필요하면 계절 조정 후) 예측하는 방법으로, 표류항이 있는 단순 지수평활과 같다.

What simple, reliable method can forecast next year's monthly demand?

내년 월별 수요를 간단하면서도 믿을 만하게 예측하는 방법이 있을까요?

The classical Theta forecast adds half of the OLS drift to a simple exponential smoothing forecast and reseasonalises the result if the series was seasonally adjusted. It performed strongly in the M3 forecasting competition despite its simplicity.

When to use Theta Method Forecasting

• You need a robust benchmark forecast for many series.
• The series has a roughly linear trend and possibly stable seasonality.
• Short to medium horizons are of interest.

When to use something else

• The series has complex dynamics or covariates; use an ARIMA or SARIMAX model.
• Demand is intermittent with many zeros; use Croston Intermittent Demand Forecasting.

Assumptions and requirements

• A stable linear drift.
• Multiplicative or additive seasonality removed before smoothing.
• Errors approximately uncorrelated.

What the analysis reports

• Point forecasts by horizon
• Prediction intervals
• SES smoothing parameter α and drift b₀

How to report it

A Theta model (SES α = [value], drift = [value]) forecast [value] at h = [value] (95% PI [[value], [value]]).

Related methods

• Exponential Smoothing (Holt-Winters)
• ARIMA Model
• Croston Intermittent Demand Forecasting

Run Theta Method Forecasting on Quorant