A Combined Clustering and Ranking Based Gene Selection Algorithm for Microarray Data Classification

Biological information related to cancer patients are recorded as microarray data. Data mining plays important role in gene selection and classification of microarray data. The mining information obtained from cancer dataset should be precise or accurate as it is one of the critical diseases affecting living beings. This paper proposes a combined gene selection approach for selecting most promising genes from microarray cancer data that identifies genes based on Significance, T- statistics and Signal-to-noise ratio. Two variations of gene selection is used here; Clustered gene selection that uses clustering mechanism to cluster similar genes before applying gene selection, and Non-clustered gene selection that selects genes without clustering. Selected genes are sent to Support Vector Machine for classification. Experiments have been conducted on microarray cancer data that contains binary class. Comparison with existing methods shows that proposed gene selection algorithm increases overall classification accuracy up to 5%.