I am trying to do time series forecasting. I know datarobot creates its own derived features which can be lags of a given feature. I want to supply my own custom lag feature into the dataset (the rolling 14 day sum of 1 feature divided by the rolling sum from 21 days ago to 7 days ago of another feature). However this will cause the first 21 rows of this feature to be NA. I already told datarobot to use ~30 day feature derivation window so this doesnt seem like an issue.
Should I just leave NA in my custom derived feature and supply the feature into the dataset and assume that datarobot will handle it properly? Again, my feature derivation window is 30+ days so does the NAs in my feature basically get removed?
日本語
