Forecasting with Kats, Holt Winters Error about kats

[yangbk560]

Hi sreekarreddy21, thanks for the feedback. Could you provide more information about your setting, especially the version of statsmodels? Thanks.

[sreekarreddy21]

Hi,
statsmodels version 0.12.2, frequency of dates was not infered by the model, it generates dates of freq='D'.

from kats.models.holtwinters import HoltWintersParams, HoltWintersModel
warnings.simplefilter(action='ignore')


params = HoltWintersParams(
            trend="add",
            #damped=False,
            seasonal="mul",
            seasonal_periods=12,
        )
m = HoltWintersModel(
    data=air_passengers_ts, 
    params=params)

m.fit()
fcst = m.predict(steps=30, alpha = 0.1)
m.plot()

Output of fcst looks like,

time	fcst	fcst_lower	fcst_upper

144 1960-12-02 NaN NaN NaN
145 1960-12-03 NaN NaN NaN
146 1960-12-04 NaN NaN NaN
147 1960-12-05 NaN NaN NaN
.......
.......
.......

[iamxiaodong]

Hi @sreekarreddy21, we were not able to reproduce your error, do you mind sharing a notebook or colab notebook that we can reproduce? I believe it might because of the data or library versions (what's your Python version?). Also, can you reproduce other models with the same dataset?

[sreekarreddy21]

Hi @iamxiaodong, there was some problem with my conda env. I created a new environment and it worked.

Thanks.

[MoKazemi9]

Hi @ sreekarreddy21, can you go ahead and close this! Thanks