Unexpected Interpolation with Visual Time Series and Monthly Granularity

zzinda
zzinda Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Adv Designer, Registered Posts: 4 ✭✭✭

I'm using the visual time-series functionality for a monthly series. The dates are all set to the last day of the month, and the data contains no missing values. After training a model using the default settings and viewing the predicted data, there is numerical interpolation happening that I would not expect since there are no missing values. I've attached a picture of some mock data to try to help explain. For example, my 8/31/23 Y value in the predicted data has a value of 20, which is the average of the 7/31 and 8/31 values in the training data. Changing the numeric interpolation method to 'previous' fixes this, but what is going on under the hood that there is interpolation happening at all when there are not missing values? I found this post which might be related, but wasn't able to make total sense of it.


Operating system used: Windows

Answers

Setup Info
    Tags
      Help me…