Options
What time series method does the "Function and Seasonal Component Forecast" operator use?
Hello,
What time series method does the "Function and Seasonal Component Forecast" operator use? Here is what the help description says:
"The model contains of a polynomial function which describes the trend of a time series and a seasonal component with length seasonality which describes the seasonal effects of a time series"
This is clearly neither ARIMA nor HoltWinters (since there are separate operators for those), but what method exactly is this, in terms that are understood in time series analysis terminology outside of RapidMiner? Is this exponential smoothing with seasonality? Here are two lists of standard approaches for comparison:
 Wikipedia list of time series forecasting methods: https://en.wikipedia.org/wiki/Forecasting#Time_series_methods
 A list of seven common methods: https://www.analyticsvidhya.com/blog/2018/02/timeseriesforecastingmethods/
Regards,
Chitu
Tagged:
0
Best Answer

Optionstftemme Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, RMResearcher, Member Posts: 164 RM ResearchHi Chitu,
Thanks for your interest in the forecasting methods in RapidMiner. As you correctly identified, the Function and Seasonal Component Forecast is not based on one of the commonly known standard approaches for time series forecasting. It is our own developed method (not claiming that no one did this before, just we didn't used any reference for implementing this).
Keep in mind that any function which is based on past values and tries to predict future values is a forecasting method. Even if a function would only predict 42 for all future values, it could be considered a forecasting method (though a not so precise one, if it is not forecasting the answer to the ultimate question of life, the universe, and everything).
The Function and Seasonal Component Forecast combines two techniques to compute predicted future values. It first uses a decomposition method, to decompose the input series into a trend and seasonal component. Then a polynomial function (degree can be specified) is fitted to the trend component. The predicted future values are computed by evaluating the function for the corresponding index values and adding the corresponding values of the seasonal component.
You can evaluate the performance of your forecasting method with the Forecast Validation operator. I would always recommend to base your decision on which forecasting method you use, on the performance evaluated in the validation.
Best regards,
Fabian4
Answers