Alteryx designer Discussions

Find answers, ask questions, and share expertise about Alteryx Designer.
Register for the upcoming Live Community Q&A Session - and don't forget to submit your questions for @DeanS regarding the future role of analytics here.

Customising ARIMA models for covariate forecasting

8 - Asteroid



Does anyone have experience customising ARIMA models and can help me understand the impact and appropriate use of the custom options?  Alternatively, can you point me to good documentation / training resources?


Hoping someone can help.




Alteryx Alumni (Retired)

@DrDan is this something you could help with?



Andy Cooper
Senior Solutions Engineer - EMEA

I would suggest looking through the online book Forecasting: Principles and Practice which is co-written by Rob Hyndman, who is also the author of the forecast package which we use. This should provide the needed background. Having said this, the important thing to remember is that time series methods basically find systematic patterns in the underlying data, and if there is not a sufficiently long time series (for data measured on a weekly, monthly, or quarterly basis, we recommend three years or more), then it is difficult for the methods to find the underlying patterns. Moreover, for time series data that is sufficiently long, research shows that automated methods (which is the default in the Alteryx time series tools) out perform manually specified models except for the most experienced users of time series methods (where the performance differences between the automated methods and the most experienced users are statistically insignificant). In cases where the forecast is an unsatisfying straight line, it typically means that either the time series is not sufficiently long enough to find the patterns in the data, or there are no systmatic patterns in data, meaning that an estimate at about the mean of the series is the best estimate possible.