The Bipolar Flexible Neural Forecasting Model and Its Application

Abstract

A flexible neural network which includes flexible, S parameter-varying function is proposed owing to the defect of the tradition neural network in order to enhance the study speed and generalization of the flexible neural network. Action function of flexible function is called S-type function that contains monopole and bipolar. There, The bipolar flexible neural S-type function is adopted. It gives the basic principle of flexible neural network and learning algorithm. To illustrate the effectiveness of the proposed flexible neural network, we get two application examples, one is forecasting power load of a certain electric network and another is forecasting floods of Taiyangtuo Rever, the results show that the model is accurate in forecast.