forecast with regressors

[tyokota]

Is it possible to forecast with regressors such as in the Google example? I first split the data into a train and valid set. I fit the train set. Then I conform the valid set regressors into a dictionary as such:

new_dict = {k:list(df[k].unique()) for k in df.columns}

However, I am seeing a NameError: The step is out of range error.

[wwrechard]

Yes, it's possible, but you need to supply the data to the regressor before making prediction. The out of range error is caused by insufficient data in the regressor.

May I know how you supply the new features into the model? One way to do is via predictN and featureDict: https://github.com/wwrechard/pydlm/blob/a2bbfd97fb9285d131a6a8be82e0728ff2d4f567/pydlm/predict/dlmPredictMod.py#L86

[logisticregress]

I'm having the same issue.
I have one predictor, and I'm trying to forecast out 3 days, so I've newf = {'x2':[[-0.13064217],[-0.13064217],[-0.13064217]]}

(predictMean, predicrVar) = drm.predictN(N=3,date=drm.n-1, featureDict=newf)

NameError: The step is out of range

I'm not sure what I'm doing wrong in setting up the future values of the predictor variable.

[Jill07217]

Same here.

I try to output 12 months ahead (N=12) with the featureDict having 12 dynamic components. But I still get the NameError: The step is out of range. Could you please provide any insights?

[ramdhan1989]

same here. I am confused to use featuredict. any example of using this will be good