In order to avoid the shortcoming of lower learning efficiency, slow rate of convergence and tend to trap into local extremum of BP algorithm, the technology combined with taboo search (TS) in forward neural network is proposed for the parameters adjustment, namely parameters of neural network are guided by TS to avoid searching in local neighbor region. The method is applied to runoff prediction in Yama Du station of Yili River in Xinjiang, and is compared with traditional BP algorithm.Results show that forward neural network based on TS not only increases the algorithm efficiency, but also improves the forecasting precision.
