Evolutionary Neural Network And Its Application Research In Enterprise Order Forecast

For time series modeling and forecasting application, and for the RBF neural network training in structural design and weights fixed problem, this paper puts forward respectively single-objective RBF neural network design method and improved multi-objective RBF neural network design method. Main work is summarized as follows:(1)The value of the RBF network structure ratio optimization problems is as a minimum error of the sample training problem, and an improved single-objective evolutionary design of RBF neural network method is proposed which optimizes the neural network weights set with a structure of adaptive adjustment capability.(2) Based on the network structure and the weights optimization of synthesis multi-objective evolutionary, an improved RBF neural network multi-objective evolutionary design method is put forward. This method takes into account the RBF neural network generalization ability training error, the network structure and optimizing targets, and a genetic algorithm of the non-dominated sorting order is proposed,which increases the population and the elite to retain the procnd the elite to retain the process of sorting, maintains diversity of population and prevent premature convergence for obtaining a uniform distribution and very good optimal solution set.(3) The RBF network prediction model,which bases on the prediction of enterprise order and radial basis function, is established by the above proposed two kinds of optimization algorithm for the training. Data forecast experiment for an enterprise's orders and application results show that the improved method to optimize the RBF network model has better prediction ability.Based on evolutionary neural network and its dynamic time series prediction study, this paper puts forward an improved evolutionary training methods, which shows the effectiveness of the method for predicting the actual order of the enterprise application. The paper's researches and achievements have also some theoretical significance and practical value for other application fields.