Abstract:  A new error-based approach of hybrid classification is presented, when data sets with binary objective variables are classified and it could increase the accuracy of classification. The paper also uses data sets to test the proposed approach and compares with the single classification. The results show that this method greatly improve the property, especially when it is predicted by two methods and the rate of variance is higher, this hybrid approach had demonstrated impressive capacities to improve the prediction accuracy.

Key words: supervised learning, classification, hybrid model, error model