Insufficient records in TS Model Factory tool

[dhavasa3]

Hi,

I am trying to predict sales bins for multiple regions at the same time using TS Model Factory tool.

My data has sales ranks for each region (denoted by postcode) from July 2016 to June 2019 at a monthly level. I am trying to predict the sales bin number for June 2019.  So there are more than two complete cycles (2*12=24 months) here.

Please see the attached data. My target is 2019-06-01 and the data is grouped by post_code.

However the model says there aren't sufficient records at post code level for prediction and asks to provide more. Help is much appreciated.

[lmorrell · Accepted solution]

Hi @dhavasa3 

Solution is attached. 

Your logic was all sound in grouping by postcode and batching models via the TS Model Factory, your workflow just required an extra step to transpose the dataset so that all the months were in a single column with their corresponding values.

The reason for the error is that the tool was getting confused as it was only reading a single month's data for each postcode and being unable to process the data from other months - resulting in the error "Error: TS Model Factory: None of the groups had sufficient records to produce an ARIMA model. Please supply more records." By transposing and providing all data for all months to build the model, it is able to forecast based on the full 24 months of data per postcode. 

Hope this helps!