Abstract—Classification is one of the most important data mining techniques especially in the area of bioinformatics. This can be clearly seen in cancer classification which is recently addressed by many researchers specially after emerging of microarrays. This technology opens the area for computer researchers to classify cancer samples without any previous biological knowledge. Microarrays high dimensionality problem forces scientists to design gene selection techniques as a preceding step to the implemented classifier. Gene selection techniques behavior varies according to the combined classifier. In this paper we are proposing a new gene selection technique which combines F-score and entropy-based methods. The output of the combined gene selection technique is fed into two different classifiers resulting in two hybrid cancer classification systems. The proposed systems achieved reliable classification accuracies when tested on two different microarray datasets. This validates the success of the proposed gene selection technique as it efficiently reduced the original number of genes by 71.29%.

Index Terms—Bioinformatics, Classification, Data mining, Gene selection, Microarrays.

D. A. Salem is with the Department of Computer, Faculty of Engineering, Misr University for Science and Technology, Egypt (e-mail: dena.salem @mail.com).
R. A. Abul Seoud is with the Department of Electrical Eng.-Comm. and Electronics Section-Faculty of Eng - El Fayoum University-Fayoum, Egypt.
H. A. Ali is with the Computer Engineering System Department, Faculty of Engineering-Mansoura University, Egypt.



Cite: Dina A. Salem, Rania Ahmed A. A. Abul Seoud, and Hesham A. Ali, "A New Gene Selection Technique Based on Hybrid Methods for Cancer Classification Using Microarrays," International Journal of Bioscience, Biochemistry and Bioinformatics vol. 1, no. 4, pp. 261-266, 2011.