A Kolmogorov-Smirnov Correlation-Based Filter for Microarray Data.


Jacek Biesiada1 and Wlodzislaw Duch2.
1Division of Computer Studies, Department of Electrotechnology, The Silesian University of Technology, Katowice, Poland.
2Department of Informatics, Nicolaus Copernicus University, Grudziadzka 5, 87-100 Torun, Poland.

Abstract.

A filter algorithm using F-measure has been used with feature redundancy removal based on the Kolmogorov-Smirnov (KS) test for rough equality of statistical distributions. As a result computationally efficient K-S Correlation-Based Selection algorithm has been developed and tested on three high-dimensional microarray datasets using four types of classifiers. Results are quite encouraging and several improvements are suggested.

Preprint for comments in PDF, 96 KB.

Reference: Biesiada J, Duch W, A Kolmogorov-Smirnov Correlation-Based Filter for Microarray Data.
Lecture Notes in Computer Science, Vol. 4985, pp. 285–294, 2008.

BACK to the publications of W. Duch.
BACK to the on-line publications of the Department of Informatics, NCU.