We have a process in which we need to forecast future numerator and denominator scores. There is a pretty strong seasonal relationship to these scores but with in-season trending more important so we are running an ARIMA (300,010) model. This all works well when testing for January-December with a target field frequency of "monthly" however we have not added runout data yet: the true data includes 3 more months of runout for each year, so we also need to forecast December +1, December +2, and December +3.
Before I add this runout data to the inputs, I can not quite figure out how to modify the ARIMA model to change the target field frequency (m value) to 15. It seems pretty easy to parse the date field into separate year and period fields so now instead of data like this:
| Report Date | Score |
| 2021-01-01 | .10 |
| 2021-02-01 | .20 |
| XXX | XXX |
| 2021-12-01 | .67 |
| 2022-01-01 | .11 |
We would now have:
| Year | Period | Score |
| 2021 | 1 | 0.10 |
| 2021 | 2 | 0.20 |
| XXX | XXX | XXX |
| 2021 | 12 | 0.67 |
| 2021 | 13 | 0.68 |
| 2021 | 14 | 0.69 |
| 2021 | 15 | 0.70 |
| 2022 | 1 | 0.11 |
But will the ARIMA module pick up that these are separate months in a 15 month period with no date value field in the input data any longer? If not, any advice on how to set this up?
Thanks in advance!
