Clustering - Volume Purchase Pattern/Periodicity

[M23]

I have the data of my monthly purchases for the last 4 years by SKU. I need to make a forecast of purchases for next year by SKU; However, prior to that, I want to study if the forecast should be done monthly, quarterly, every quarter, or yearly depending on the pattern of my purchases.

How can I create a workflow so that SKUs are grouped according to the regularity with which they are purchased? The behavior is not perfect, but a clustering could be made.

Example of the data I have:

SKU #	Year	Month	Volume purchased
12312	2018	01	5
96854	2018	02	0

[caltang]

I think your post is better suited for the ML Forum.

[gawa]

hi @M23 

You can try Clustering tools like this. In this example, I categorized into 2 clusters with yearly basis.

If you change variables for Clustering tool to another one(Year-Month, Quarter, etc), then you can analyze from different perspectives.

Hope this could somewhat help you out.