Features

The annual KDnuggets data mining tool poll results were reported in KDnuggets News May 16 issue.

Each year it's an interesting and controversial poll. Comparing the 2005 & 2006 poll results, I draw different conclusions about the tools that show the biggest increases and decreases:

To me, a combination of four factors tell the story:

1. Substantial change in 2005-2006 market share (calculated as either share of votes or share of people)
2. Substantial change in 2005-2006 tool ranking
3. Movement in or out of the top 10 rank
4. Substantial change in number of votes

1) KXEN

• Market share increased from 2% in 2005 to 16% in 2006
• Rank jumped from 22nd up to 5th
• Moved into the top 10
• 1186% increase in votes (7 votes in 2005, 90 votes in 2006)

• Market share increased from 18% in 2005 to 28% in 2006
• Rank moved from 4th up to 1st
• 130% increase in votes (69 votes in 2005, 159 votes in 2006)

1) SAS-Enterprise Miner

• Market share dropped from 13% in 2005 to 7% in 2006
• Rank dropped from 6th to 12th
• Dropped out of the top 10
• 24% decrease in votes (49 votes in 2005, 37 votes in 2006)

• Market share dropped from 9% in 2005 to 4% in 2006
• Rank dropped from 9th to 16th
• Dropped out of the top 10
• 38% decrease in votes (32 votes in 2005, 20 votes in 2006)

1) It is a competitive market. The most popular tool was selected by only 28% of the people, and got only 14% of the total votes. Most people reported using multiple tools (the average was slightly over 2 tools per voter). The market share numbers above are calculated as share of people.

2) Comparing percentage gains in votes is potentially misleading due to the small number of votes some tools received in 2005. E.g., MATLAB's 156% gain from 16 to 41 votes is potentially a more meaningful gain than Angoss's 167% increase from 3 to 8 votes.

3) When looking at the change in tool ranks, it is extraordinary that KXEN moved from 22nd in 2005 up to 5th in 2006. It will be interesting to see if its popularity is maintained in the 2007. Movement near the top of the rankings is especially noteworthy, as it reflects the shift in many more votes. E.g., it's very striking that the Salford tools moved up from 4th in 2005 up to 1st in 2006. It's not as striking that Oracle moved from 20th up to 15th.

4) I am interested in hearing people's opinions on vendors reminding their employees and users to vote. I've heard some say this is cheating, but others say these people are all legitimate members of the data mining community and their votes should be heard.

Karl Rexer, PhD
www.RexerAnalytics.com