Ensemble algorithm of neural networks and its application

Neural network ensemble is a very hot topic in both neural networks and machine learning communities (A. Sharkey, 1999). A new approach named BAGAEN is proposed, in which adaptive genetic algorithm and bootstrap algorithms are employed to increase the different degrees among individual RBF neural networks in order to enhance the generalization ability of a neural network system. The training set for individual RBF neural network is generated by the algorithm based on bootstrap and the result can be obtained by using majority voting method or simple averaging method. Experimental results show that BAGAEN has preferable performance in generating ensembles with strong generalization ability. Finally, BAGAEN is applied to predict the magnitude of earthquake.