PAR Model cannot fit columns with dtype `period`

[andrewcz]

Hi All,

I have having a bit trouble with the time series generation module.

It looks to be an issue with the index but I'm not sure how to resolve it. Best, Andrew

[npatki]

Hi @andrewcz, there is an ongoing bug where the PAR model is crashing if the sequence index is anything other than a datetime. You can follow along in #808 for updates as we get to a resolution.

Due to this, I believe the error is caused by the following line:

DF.index = DF.index.to_period('M')

This is changing the column's data type from datetime to period.

Workaround: Is it possible for you to keep it as datetime (ie remove this line)? Verify that you have the supported data types by running the following line:

DF.dtypes

This will provide a list of column names and their dtypes. Make sure you see datetime and not period.

[npatki]

Quick update: Seems like this error is occurring during the fit, not the sample. There is a similar issue on this in #465, but the error is different. Let's keep this issue open as a separate bug, as the root cause seems different.