Time Series Forecasting - Values Seem very wrong

Question

I'm new to using the time series tools.  I attempted to use the Time Series Forecasting tool example for some data that I am working with but the results from the output tab of the forecast tool seem very wrong. For instance it is forecasting 16 streams a period when the data shows over 20,000 streams per day. Is something wrong with the period? Or am I messing something up in how the tools are set up?

Stream_Forecast.yxzp

ArtApa · Answer

Hi @JohnRichey - I suggest you to go through the Time Series training, if you haven't done it already: https://community.alteryx.com/t5/Interactive-Lessons/tkb-p/interactive-lessons/label-name/Time%20Series%20Forecasting

AngelosPachis · Answer

Hi @JohnRichey ,

I think they issue lies in the data being fed into the ETS tool. In time series forecasting, your target variable should be sorted in ascending chronological order (older dates first) for the tool to create a proper model.

This is how your dates looked like, but I would suggest you add a sort tool in there to sort the dates as described above

If you don't sort on chronological order, the ETS tools will train the model as the days feed into it, and as a result the oldest dates (around 2015-16 with a stream of 0) appearing if you scroll down will be treated like they are the most recent ones.

That will explain why you get those weird figures.

Please test it and let me know if that worked for you.

Cheers,

Angelos