Adding additional features in a time series model?

[drewed206]

I have used TS model factory in the past and have been looking at some of the other time series tools (arima, ets) in case the answer is just building out a new batch macro to feed into those... but is there a way to add features (as an additional regressor) into my time series model?

[BS_THE_ANALYST]

@drewed206 

My understanding is that we have two Time Series models at our disposal in Alteryx: Arima and ETS. Those tools, alone, will allow you to process a single batch of time series data. If you're wanting to feed several batches in then the TS Model Factory comes into play. The Model Factory is much faster compared to batch macroing an Arima or ETS tool - it's like comparing a loop to parallel processing.

If you're wanting to add another regressor, you're looking at using Covariates with your models. If memory serves me correctly, the ARIMA model supports covariates (and I believe the TS Model Factory will support the batched variety in that case). Again, from memory, you'll just need to build the arima model with the history of your both regressors and then you'll also need to supply the TS Covariate Forecast tool with the additional regressors future data for the number of periods you want to forecast. For instance, if you have data surrounding Snowfall and Rainfall, you could supply the next 6 days expected (or known ideally) rainfall and you'd hope that could better explain your predicted snowfall for the next 6 days. (I think this is correct, do let me know if you encounter an issue).

Here's a nice article around covariates and Susan does a much better job at explaining the concept: https://community.alteryx.com/t5/Data-Science/Back-to-the-Future-ARIMA-and-Forecasting-with-Covariat...

Hopefully that's answered your question. If not, do shout.

All the best,

BS