Neural Network Ensemble Method Based on Improved Sort Learning Algorithm

Based the analysis of the deficiency existing in current neural network ensemble method, a new method based on sort learning algorithm was proposed, which contains several predictors. This is true provided the combined predictors are accurate and diverse enough, which posses the problem of generating suitable aggregate members in order to have optimal generalization capabilities. According to the new algorithm, the data used in the training have been discriminated using different strategies firstly. And then the weights of the participated neural networks have been optimized to obtain the minimum estimate error. Finally the classified results were presented after the ensemble process of them. A significant advantage of this algorithm in the classification accuracy and speed has been demonstrated experimentally and theoretically, comparing with the classical model.