Forecasting: Forecast Methods & Publishing

Forecasting: Forecast Methods & Publishing

This article is part of the Forecasting series, split from the original Forecasting article.

Pyramid Forecast (Standard View) 

The Pyramid Forecast is the forecast suggested by the ML engine. Arkieva uses machine learning to select methods based on accuracy and robustness over time. For this reason, the Pyramid Forecast may not always be the method with the lowest accuracy measure.





Arima with Season

Arima is used to calculate trend data, but does not support a seasonal time series. Arima with Season can support time series with a seasonal component. 




Robust Seasonal

Like Seasonal, this method fits a trend along with sine and cosine curves. This method uses linear programming to fit a seasonal series in a way that — compared to the regular seasonal method — is less likely to be thrown off by noisy values that depart from the trend or seasonality.




Fourier

Fourier is a special cyclical method which tries to find the cycle on its own. Fourier is a curve-fitting techniques are used to fit a combination of sine and cosine curves to the series. For very stable series with no trend, this can provide the best fit, but in most other situations one of the seasonal methods should work better.




Winters Optimum Additive

This method uses curve-fitting to attempt to select the values of the parameters alpha, beta, and gamma for the Winters additive method which give the “best” results. The algorithm does not use the given alpha, beta, or gamma parameters. The damping and MultiRegWeight parameters are used only in the case where there is not enough history to run Winters method, in which case a linear regression is used to forecast the series. This method will usually produce a good forecast, but the error criterion is based on fitting a curve to past history, so it doesn’t always have superior predictive ability to the ordinary additive Winters method.




Seasonal

The Seasonal method is a regression method fitting a linear trend (as with linear regression) along with sine and cosine curves that allow nearly any seasonal departure from the trend to be fit. This method is best for a series with a long term stable trend. For less consistent series, some of the other seasonal methods will probably be better.




Winters Optimum Multiplicative 

This method is the same as Winters optimum additive except the parameters Alpha, Beta and Gamma are chosen to minimize the MSE of the historic forecast.




3 - Period Average 

This Moving Average method only takes the average of three periods of historical data, but does not consider any pattern. 




Generated Forecast

Arkieva uses machine learning to select methods based on accuracy and robustness over time. Pick Best covers the scenario where if the number of history periods (after leading zeros are removed) are less than the window, the resulting forecast will be a straight constant.



  1. Apply Sequence: Runs the selected formulas in the specified order. The output generated by each formula is passed as the input to the next formula in the sequence. The final forecast is the result produced by the last formula in the sequence.
  2. Best candidate: Evaluates all selected formulas and automatically selects the formula that produces the best forecast based on the configured error criteria. If the number of available input data periods is less than the configured error window size, the selected formulas are ignored, and a separate fallback logic is used to generate the forecast.
  3. Average: Executes all selected formulas independently and calculates the forecast by taking the average of the output values for each period. The resulting forecast represents the mean of all selected formula outputs.

Publishing

Publishing the forecast strategy will publishes the values to the future and history. 




Forecasting series