How to make qlik handle variables correctly?

[keithyowell1]

I'm having a pretty basic issue that I'd really love a bit more flexibility in the automl application. for one experiement, I use zip code as a feature which Qlik interprets as numeric. This is not correct so before I try to train the model i tell Qlik that the field is categorical. in another case qlik interprets percent of purchases made in-person as a  categorical field. this is also not correct so I respecify it into a numeric field. when I ask to run the experiment it tells me"AutoML could not convert PCT_IN_PERSON to numeric. Please start the experiment over or create a new experiment from the hub." How do I get around this? If I decide to not convert that field to numeric, and the experiment runs, I can't upload new data for the apply data and tell qlik to treat zip code as categorical. When I try to deploy the model with new data, there is no option to respecify the zip code field in the apply data set I upload, and the deployment fails. How do I get around this?

[Kyle_Jourdan]

For the percentage field issue, you will need to modify the dataset to remove the % symbol, as this is what is causing the conversion to fail. It’s best to store these in their numeric representation (i.e. 0.90 for 90%) for use in AutoML.

As to the issue with zip code, you are correct to convert it to a categorical. The deployment should handle the conversion automatically if it was setup to be changed as part of the experiment. You will initially get a warning about a data type mismatch, but it shouldn’t stop the deployment from saving and running.

If you are experiencing an issue where there is an error in the experiment running and it is specific to the zip code field, there may be another issue that needs to looked at.

[keithyowell1]

There is no "%" on the number. it is calculated from a numeric field in the
SQL script that pulls the data.

[igoralcantara]

Is your file used for training a QVD? Regardless, make sure to check the data provided if all rows are numeric. Maybe there is one or more rows where data is text. For example, if the value is N/A. That might mess up the whole column type. You can also try in your script to specifically set to numeric and format it, using the Num() function.