Forecast Pluging:Forecasting multiple time series

Ouma
Ouma Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Registered Posts: 12 ✭✭✭✭

Hi Dtaikers!

I want to forecast multiples Time Series using The new Forecast plugin, my time series have different history length, I already interpolated missing values for each one, but I don't want to use extrapolation because it will bias my study, so as a result, I have a dataset with multiple Time Series, identified by store ID, and having different history-length.

For instance: Store x has history from 2011 TO 2015 AND Store y has history from 2014 to 2020 and so on...

I want to know if by activating the Long format and adding identifiers, the Forecasting Horizon will be adjusted for each TS, or I should partition my input dataset and then train partitioned forecasting models?

Thank you!

Best Answer

  • StanG
    StanG Dataiker, Registered Posts: 52 Dataiker
    Answer ✓

    Hi,
    The multiple time series can have different history lengths but the forecasting horizon is the same for all time series, starting after the last day of each time series (in your example, if the forecasting horizon is 1 year, then the models will forecast year 2016 for store x and 2021 for store y).

Answers

Setup Info
    Tags
      Help me…