KDnuggets : Polls : Data Mining Methods (Mar 2007)
Poll
Data mining/analytic methods you used frequently in the past 12 months: [203 voters]

Decision Trees/Rules (127) 62.6%
Regression (104) 51.2%
Clustering (102) 50.2%
Statistics (descriptive) (94) 46.3%
Visualization (66) 32.5%
Association rules (53) 26.1%
Sequence/Time series analysis (35) 17.2%
Neural Nets (35) 17.2%
SVM (32) 15.8%
Bayesian (32) 15.8%
Boosting (30) 14.8%
Nearest Neighbor (26) 12.8%
Hybrid methods (24) 11.8%
Other (23) 11.3%
Genetic algorithms (23) 11.3%
Bagging (22) 10.8%

Comments

Omar Calzadilla, Other methods:
Rough Sets (The theory was originated by Zdzislaw Pawlak)

Will Dwinnell, Other methods:
GLMs, especially logistic regression
quadratic discriminant analysis

Matthias, Fuzzy Logic
You are missing fuzzy logic - a major tool for us at modern analytics

Note from the Editor
The average number of methods per voter was 4.1, significantly higher than 3.2 in 2006. In comparison with 2006, more data miners are using the classics methods of decision trees/rules and regression.

However, the highest growth was in use of other methods (see some mentioned above), Genetic algorithms, Boosting, Visualization, Hybrid methods, and Bagging.

The methods that suffered relative declines in popularity were SVM and Association Rules.

Here are the results of 2006 KDnuggets Poll on Data Mining Methods.


KDnuggets : Polls : Data Mining Methods

Copyright © 2007 KDnuggets. Subscribe to KDnuggets News!