1

I have problems calculating feature_matrix using cutoff_times table because of a type mismatch in cutoff times in the cutoff table and the time_type of the entityset.

I am trying to understand the predict-next-purchase example using synthetic data.

I got to the point of having cutoff_labels with 'datetime64[ns]' or pandas._libs.tslibs.timestamps.Timestamp type of cutoff time entries.

The dfs procedure gave me an error message:

cutoff_time times must be numeric: try casting via pd.to_numeric(cutoff_time['time'])

I figured that the problems lies in the entity set because the time_type of the entityset is not DatetimeTimeIndex but NumericTimeIndex?!

Filip Floegel
  • 95
  • 6

1 Answers1

2

found out...

I set a "wrong" time index defining an entity from a dataframe. So when one use a numeric variable as a time index then the time_type of the entity is numeric but if you have set the time index using a datetime variable the time_type of the entity is DatetimeTimeIndex!

Filip Floegel
  • 95
  • 6