Transcript Toward A Unifying Taxonomy for Feature Selection

Toward A Unifying Taxonomy for
Feature Selection
• Feature selection is to select relevant features for
effective data mining.
• There are many good feature selection algorithms.
• This SGER project attempts to
– define a common platform for all algorithms
– build a taxonomy to discover how they complement
each other and what’s missing
• Ultimately, we wish to solve automatic selection
problem.
Achievements
• Project web site with a software package
www.public.asu.edu/~huanliu/feature_selection.html
• Publications:
–
–
–
–
–
One paper accepted by ICML 2002
An updated survey draft
One book chapter by LEA Handbook of DM
One invited paper at a data mining workshop
One submission to SIGKDD 2002
• One full proposal to NSF IDM
• A short Data Mining course is planned.