n_lags
Lagged regressor to n_forecasts
forecast steps ratio
#320
-
What is the optimal ratio when using n_forecasts and n_lags? Should n_lags always be greater than n_forecasts? What does n_lag really mean? Lagging by n rows? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
@veioxve3r1d23seb2g6b
There is no rule set in stone. Generally, it should be at least same.
The conventional ratio is around 1:1 to 1:2, sometimes higher. |
Beta Was this translation helpful? Give feedback.
@veioxve3r1d23seb2g6b
n_lags
is the number of most recent observations to include in the model. e.g.n_lag=3
uses all of the last 3 hours as input to the model.There is no rule set in stone. Generally, it should be at least same.
The conventional ratio is around 1:1 to 1:2, sometimes higher.
e.g. hourly data: when predicting next 6h, use last 24h
e.g. hourly data: when predicting next 72h, use last 72h