We use accuracy for evaluation. Accuracy = R / N, where R is the number of test instances with correct label and N is the total number of test instances.