NCAA-Was I surprised my team lost? SAP Predictive Analytics Explains
So last week I shared my initial clustering analysis using SAP Predictive Analytics 2.0 (Expert Analysis) Looking back at NCAA Basketball Clustering Analysis – Reviewing the Results
Sadly, it did reveal that my team, Kansas, didn’t stand a chance of winning yesterday’s game against Wichita State (congratulations to Wichita State). Kansas was the only team in cluster 4 who won the first round. Did this help heal my wounds of losing? No…I’m reminded of Dr. Smith from Lost in Space:
“The pain of it all” – but life goes on, and there’s more data to review. So today, I took the winning “Sweet Sixteen” teams and ran autoclustering again
I reduced the number of clusters and looked at winning percentage this time.
I ran the analysis successfully, as shown above
Above shows my clusters resulting from Predictive Analytics.
Parallel Coordinates chart is a little easier to read this time (you can tell Kentucky because they continue to have 0 losses)
Above is the scatter matrix chart from SAP Predictive Analytics
How did the results look?
So 11 teams are in cluster 1 and 5 are in cluster 3
See teams in the crosstab below – first is cluster 1:
Cluster 3 has 5 teams below:
Should I predict based on this? Kentucky (Cluster 3) over West Virginia (Cluster 1)?
NC State over Louisville? Utah over Duke?
I am not sure.
Also it doesn’t help that Oklahoma and Michigan State are in the same cluster.
So I go back and look at the old clusters (pre-tournament):
Cluster 8 continued:
Cluster 8 had the most wins, so based on this Michigan State is favored over Oklahoma
We’ll see what happens this coming weekend.
BI pre-conference session at ASUG Annual conference
Monday, May 4. (extra registration fees apply).
Featuring Hands-on SAP BusinessObjects BI 4.1 w/ SAP NetWeaver BW Powered by SAP HANA – Deep Dive
See details here: http://bit.ly/ASUGPreConBI
Focus on Analysis Office, Lumira, and Design Studio. You get to work with these for 7 hours! Full day BI workshop. Limited to 30 people. One person per machine (no sharing).
Also some upcoming ASUG Predictive Analytics webcasts are listed below: