C-KPCA: Custom Kernel PCA for Cancer Classification

Abstract

Principal component analysis (PCA) is an effective and well-known method for reducing high-dimensional data sets. Recently, KPCA (Kernel PCA), a nonlinear form of PCA, has been introduced into many fields. In this paper, we propose a new gene selection, namely Custom Kernel principal component analysis (C-KPCA). The new kernel function for KPCA is created by combining a set of kernel functions. First, Singular Value Decomposition (SVD) is used to reduce the dimension of microarray data. Input space is then mapped to a higher-dimensional feature space using the proposed custom kernel function. The main objective of our method is to extract nonlinear features for classification process. In order to test the accuracy of our method, a number of experiments are carried out on four binary gene datasets: Colon Tumor, Leukemia, Lymphoma, and Prostate. The experimental results show that our proposed method results in a higher prediction rate as comparing with several recently published algorithms.