Protein structural can be regarded as one of the most significant elements in protein function. In this work, the amino acid residues’ frequency featuring set and the E-H description method have been employed as the classification features. And then, support vector machine and neural network have been employed as the classifier in this work. In these employed features, some novel features have the ability to making differences between α/β type and α+β one. In order to show the performance of the proposed method, the three employed benchmark datasets are utilized to train and test the proposed approach. The results demonstrated that some performances of the proposed method are better than existing method ones, especially in α/β and α+β types classification.