sion trees are widely used for classification. Several approaches exist to induce decision trees. All these methods vary in attribute selection measures i.e., in identifying an attribute to split at a node. ID3 is the classic and popular decision tree algorithm which uses Information Gain based on Entropy as the node splitting criteria. This paper proposes a novel splitting criteria based on Coefficient of Variation and it is named as Coefficient of Variation Gain (CvGain). Empirical analysis based on standard data sets revealed that Coefficient of Variation based decision tree (CvDT) has less computational cost and time.