How to predict a discrete attribute with Decision Trees?


  • So this is my first time trying to implement Data mining algorithms with SQL server. I've working with the AdventureWorks DW sample data from Microsoft and I've implemented a couple of examples which seem to give reasonable results.

    Now I'm trying to predict the attribute "Sales Reason" based on several attributes in the Customer dimension. My main problem is the following:

    When I was predicting "Unit Price" for example, the model would give me results where it sums the unit prices of all cases in each node, which is perfectly fine. But with "Sales Reason", I want it to give me the exact ReasonID and not sum it so I would get results like "0.564" or "1.432" because that doesn't make any sense.

    I've tried this with both Decision Trees and Association Rules and tried setting the Sales Reason attribute in the model to Discreet and Discretized but still won't get any meaningful results.

    Does anyone have an idea about this? Thanks!

    Tuesday, December 10, 2013 6:28 PM

All replies