misleading error statement - Cannot convert non-finite values (NA or inf) to integer

raghutej
raghutej Registered Posts: 10

Hi ,

I am running into an issue while using evaluation recipe for a timeseries model.
I ran similar tests multiple times previous without any issue.
This time the error i see while running evaluation recipe is "Cannot convert non-finite values (NA or inf) to integer"
I checked all columns and i don't see any Null values in any columns.
Not to sure how to proceed. As the error message is not very useful as it doesn't give the the column name.

Any suggestions on how to proceed on this?

Thanks
RT

Answers

  • raghutej
    raghutej Registered Posts: 10

    Problem resolved.
    This issue is because of automatic type casting for some of the columns.
    Example for a column named hour_id datatype is defined as varchar.
    While training, the values were 1,2,3,...24.
    In the input passed during evaluate recipe, the values are being casted as 1.0000, 2.0000, ..... 24.0000 the casting is happening automatically.
    I have to cast the datatype of hour_id as integer to get data in same format as it was in training.

Setup Info
    Tags
      Help me…