A micro array represents thousands of gene expression levels across a few samples. Determination of an optimal set of features from such a high dimensional dataset requires a good feature selection method. Based on statistical significance of the features, an elimination of insignificant genes can be performed. However such methods lack biological validation. In this paper we propose a method where statistically reduced gene set is biologically verified with the help of Gene ontology (GO). With this verified feature set classification is performed on three micro array datasets using Support vector machine (SVM) and Random forest (RF) classifiers. The classification accuracy determined using same test sets for both without and with ontological verification are found to be significantly improved.