Partitioning models
Antoine_Auby
Registered Posts: 2 ✭✭✭✭
Hello,
Is there a way to partition a model the same way we partition datasets?
In our case, we have forecast models for the travel industry, and the feature importance could differ massively depending on the season (i.e. summer VS ski holidays). We would use the same schema as input, with the same model architecture, but trained separately for every season/month to account for their differences.
Thanks,
Antoine
Is there a way to partition a model the same way we partition datasets?
In our case, we have forecast models for the travel industry, and the feature importance could differ massively depending on the season (i.e. summer VS ski holidays). We would use the same schema as input, with the same model architecture, but trained separately for every season/month to account for their differences.
Thanks,
Antoine
Tagged:
Answers
-
Hello,
That is something on which we are working, but is not currently possible.
You can work around this by splitting your dataset and using the ML API to train separate models. You can also of course create several models manually, and since DSS 5.1.3 you can copy the feature handling & algorithm settings from one ML task to another.
-
Thanks
