Skip to Content
Author's profile photo Mitesh Patel

2015 College Basketball Predictive Analysis

Predictive Analytics 2.0 – College Basketball Analysis by Charles Gadalla

It’s that time of your again!  Spring is in the air, well at least for the west coast, and there is a hint of craziness about to hit us.  Did I say crazy? Yep. A whole lot of it. So much that this entire month has been dubbed March Madness!  And it’s time to make some Picks!!  Your friends on the Lumira and Predictive team joined forces to bring you a solid dataset, and some tips on teams to pick for Selection Sunday.

Taking this dataset, I turned to SAP Predictive Analytics 2.0 to crunch through the data. First I built a Cluster analysis to get an idea of how the teams split out based on their rankings:


I then took a look at each Cluster (I started with 10) to see where their strengths/weaknesses were:


I then ran a Regression analysis to see which Variables were most contributive to success – and found that “Net Points vs Average” was a clear winner:


I built a Decision Tree on the variables so I could get an idea if this lined up with the Cluster break out I found in the first 2 steps, but using the variables I got from the regression analysis:


Good news! It matched up well, so I could start to make my selections based on the stack rank of the teams.  Below is a map of the US showing how the conferences I picked stacked up.


In summary, here’s the list of teams I believe will make it to Selection Sunday (March 15). Stay tuned, once the brackets are announced the Lumira/Predictive team will put out a bracket sheet for you to challenge.  Use our data, use your own, use your gut too!

Click here for full details on the 2015 #VizTheMadness Challenge powered by SAP Lumira & SAP Predictive Analytics.

Good luck!!

Disclaimer: this is for entertainment purposes only. Please do not use these for gambling purposes, as this is not condoned and SAP cannot be held responsible.

Assigned Tags

      Be the first to leave a comment
      You must be Logged on to comment or reply to a post.