Optimal alpha in exponential smoothing models

[matsuobasho]

I realize this is likely out-of-scope for this package, but are there any ready-to-use techniques for finding the optimal alpha based on minimizing the SSE of the residuals?

I understand I can just write my own optimization code but curious what techniques practitioners turn to?

[jmoralez]

I think that's what this model does

statsforecast/statsforecast/models.py

Line 1989 in 7f60571

class SimpleExponentialSmoothingOptimized(_TS):

do you want to extract the value of alpha?

[matsuobasho]

Ah, great, I should have looked at those functions before asking. Extracting alpha would be beneficial - is that feature available in that function?

[jmoralez]

No, the alpha is found here

statsforecast/statsforecast/models.py

Lines 1783 to 1791 in 7f60571

	def _optimized_ses_forecast(
	x: np.ndarray, bounds: Sequence[Tuple[float, float]] = [(0.1, 0.3)]
	) -> Tuple[float, np.ndarray]:
	r"""Searches for the optimal alpha and computes SES one step forecast."""
	alpha = minimize(
	fun=_ses_mse, x0=(0,), args=(x,), bounds=bounds, method="L-BFGS-B"
	).x[0]
	forecast, fitted = _ses_forecast(x, alpha)
	return forecast, fitted

but it isn't stored. You can open a feature request for us to store it in that model if that'd be useful for you.

[matsuobasho]

Thanks @jmoralez will do if I actually have need for it. This answers my initial request, glad these functions are available!