Hi everyone,
I am trying to incorporate some covariates to an Arima model I used to run without them, trying to achieve higher precision in their forecast results.
In the beginning, I thought only future data from my covariates were needed to input into the TS Covariate Forecast Tool (apart from the Arima model). But this was giving me the error: TS Covariate Forecast (7): TS Forecast: Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) :
So I started looking for answers both inside the Alteryx community and in Google in general. Some have encountered my same problem, and they realized the need for inputting all the data (past and future data of the covariates - the same data you introduce to the Arima tool plus the future data of the covariates). And they also realized the tool outpat forecasting data for the same number of periods as the periods of data existing in the input.
Then, I tried this method and yes, the tool doesn't return any error and outputs forecast data from the first period of the future to the period n (being n the number of periods of the input data).
But I am not at all convinced the results are good - they are not reasonable at all. It seems like the tool is using the covariate data from the first period of the data (past data) to predict the forecast of the first period of the future, and so on for the rest of periods.
Any one is using this tool successfully? Are you doing anything diferent? Any tips will surely be helpful.
I've attached an example of what I'm explaining: a flow returning an error and a flow with forecast data that have no sense at all.
Thank you,
Ramón
