The best trial method is used to automatically select exponential smoothing parameters.
Denote the i-th approximation:
,
where:
ai. Parameter α.
di. Parameter δ.
gi. Parameter γ.
fii. Parameter ϕ.
n+1-th approximation un+1 with the n-th approximation un available is searched as follows:
Introduce some S implementations of random vectors:
Find the index i0 using the criterion:
Where J is the criterion that defined optimal one-step-forward predictions in an exponential smoothing problem.
Find n+1 approximation:
Move to the first step.
See also:
Exponential Smoothing | Seasonal Effects Model | Growth Models | IExponentialSmoothingBestTrialMethod