Time Series resampling using EOM dates, but I want month start
Is there a way to do the time series resampling by month where it uses the first day of the month (i.e. 2022-01-01) rather than the EOM? My data is already in first of month but resampling is forcing it all to the end. I played around with some option in the resampling but couldn't find a solution. Thanks!
Operating system used: Windows 10
Answers
-
Alexandru Dataiker, Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Adv Designer, Registered Posts: 1,226 DataikerHi @Vaslo
,
You will need to use a recipe before training, where you add a step "Increment date with a specified value", and specify DATUM / static value / -1 / days. It should pre-realign all your existing series to the end of the previous month. -
Sorry I guess I don’t understand - would this give me all the dates starting on the first?
-
Alexandru Dataiker, Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Adv Designer, Registered Posts: 1,226 Dataiker
Sorry, perhap I misunderstood your question.
Are you referring to the https://www.dataiku.com/product/plugins/timeseries-preparation/ plugin?The resampling https://doc.dataiku.com/dss/latest/machine-learning/time-series-forecasting/settings.html#time-series-resampling that happens at the beginning of the training for time series, only aligns the data with the end of the requested period (end of the month) this behavior can't be changed currently.
Are you having a specific issue with the current behavior? -
I think you have explained it correctly. Yes, it's that function I was using and yes when I run the plugin it only gives me an end date. The tools I use externally to Dataiku are looking for start of the requested period. I know a lot of tools out there allow passing a variable like 'MS' instead of 'M' to consider month start but it sounds like this does not do it.