Hello there,
I've spent about 8 hours watching videos and trying all kinds of ways to get a TS model that looks good. I can't seem to get my predictions close to the actuals.
I set up a workflow based on a combo of videos. Can anyone maybe just look at the flow and give me any advice or pointers?
I did a crosstab to aggregate the data. Do I even have to do a crosstab? My raw data is by date and transaction level. Could I just insert a 'year-month' column and feed the raw data into a model.
Thank you in advance.
Solved! Go to Solution.
Time Series is not my strong suit, but It seems like there is some seasonality that needed to be accounted for in your estimation. In the ARIMA tool, I modified the seasonal component differencing and saw a significant change in model performance.
Alteryx documentation on the ARIMA tool recommends this online book Forecasting: Principals and Practice for "A detailed discussion of the ARIMA model, along with a description of the automated methods used in this tool".
Thank you so much. This helped me out a lot.