The improvement of accuracy of gene expression data classification with gene ontology

Gene selection is one of important research issues in analysis of gene expression data classification. Current methods try to reduce genes by means of statistical calculations and have used semantic similarity under gene ontology. In this article a technique has been presented based on which in addition to considering biological relation among genes, redundant genes by means of hierarchical clustering are omitted and the accuracy of classification increases. The structure and function of this technique have also been explained. The experiments using a single real data set indicate that the proposed technique in addition to selecting fewer genes, have higher accuracy of classification (Loocv), comparing to the technique that is based on semantic similarity.