Affiliations: [a] Centro de Investigación en Computación, Instituto Politécnico Nacional. Av. Juan de Dios Bátiz, Esq. Miguel Othón de Mendizábal. Col. Nueva Industrial Vallejo, Gustavo A. Madero, 07738, Ciudad de México, México | [b] Facultad de Ingeniería, Universidad Autónoma de Guerrero. Av. Lázaro Cárdenas s/n, Ciudad Universitaria Zona Sur, 39087. Chilpancingo Guerrero, México | [c] Centro de Innovación y Desarrollo Tecnológico en Cómputo, Instituto Politécnico Nacional, México. Av. Juan de Dios Bátiz, Col. Nueva Industrial Vallejo, 07700, Ciudad de México, México
Correspondence:
[*]
Corresponding author: Gustavo Adolfo Alonso, Laboratory of Computing Technologies and Electronics, School of Engineering, Universidad Autónoma de Guerrero, Av. Lázaro Cárdenas s/n, Ciudad Universitaria Zona Sur, Chilpancingo, Guerrero 39087, México. Tel.: +52 1 747 112 2838; E-mail: [email protected].
Abstract: Feature selection is a common solution to microarray analysis. Previous approaches either select features based on classical statistical tests that can be tuned up with a classifier, or using regularization penalties incorporated in the cost function. Here we propose to use a feature ranking and weighting scheme instead, which combines statistical techniques with a weighted k-NN classifier using a modified forward selection procedure. We demonstrate that classification accuracy of our proposal outperforms existing methods on a range of public microarray gene expression datasets. The proposed method is also compared to state-of-the-art feature selection algorithms by means of the Friedman test. Although a bunch of feature selection techniques has been used for genomic data, the experimental results show the classification superiority of our method on most of the present gene expression datasets.
