Water pollution forecasting model of the back-propagation neural network based on one step secant algorithm

Abstract

To overcome the shortage of the conventional Back-propagation (BP) Network, the BP network is trained by using one step secant (OSS) algorithm. According to the Yangtze River water statistics reported from 1995 to 2004, the BP neural network model for water quality evaluation was established to predict the consequence of water pollution in the next ten years. The result shows: (1) With no effective measures, the Yangtze River water pollution will be on drastic rise in ten years; (2) This model can predict development tendency in ten years and its result is reasonable and also proves that it has strong generalization ability. It is a very valid model of estimating non-linear problem. (3) BP neural network based on OSS algorithm possesses the advantage of high accuracy and high speed for convergence. © Springer-Verlag 2010.