Study on a Hybrid Approach for Improving Clinical Behavior of Cancer by Assorting Informative Genes

In recent times in the classification and diagnosis of cancer nodules, gene expression profiling by micro array techniques are playing a fundamental role. A range of researchers have proposed a number of machine learning and data-mining approaches for identifying cancerous nodule using gene expression data. The process of gene selection for the cancer classification encounters with some major problems due to the properties of the data such as the small number of samples compared to the huge number of genes, irrelevant genes, and noisy data. Hence, this paper aims at selecting a near-optimal subset of informative genes that is most relevant for the cancer classification. This paper also proposes an efficient BFSS (Boost Feature Subset Selection) technique to improve the performance of single-gene based discriminative scores using bootstrapping techniques. The proposed hybrid approach (Filter-Wrapper) will be implemented on three publicly available microarray datasets. These microarray datasets are: Acute Lymphoblastic Leukemia Cancer (ALL), Lung Cancer and Colon Cancer. General Terms Genetic Algorithm, Cyclic Genetic Algorithm, Boost Feature Subset Selection, Filter-Wrapper Technique.