Gene Filtering and Classification of Microarray data

Abstract

Microarrays have been widely used by scientific community to study and analyze the expression of large number of genes simultaneously. The advance of microarray technology provides a huge amount of genomic data which leads to the necessity of efficient methods for its analysis. This technology gains special attention in the field of cancer research because with better classification of tumors, it would be easier to efficiently diagnose and treat cancerous cells. Efficient classification of microarray data is still a problem because of increased dimensions of the feature space and a very small sample size. Effective methods are required to reduce the dimensionality and improve the classification accuracy to by extracting meaningful information from the datasets. In this study, we aim to find a combination of different feature selection and classification methods that work best in terms of accuracy and number of features selected. Our proposed approach uses Correlation Based Feature Selection CFS (using Forward Search) as feature selection method combined with an ensemble based on SVM , Random Forest, Bagging and Bayesian Generalized Linear Models (BayesGLM). A number of experiments are conducted on the benchmark datasets: colon cancer, prostate cancer, leukemia and breast cancer. We demonstrated that our proposed approach outperforms or give comparable results with already published approaches in literature.