Forecasting model on rice blast based on BP neural network and chaos theory

To foresee rice blast efficiently, a new forecasting model was established by integrating the chaos theory into the BP artificial neural network (ANN). The BP neural network was optimized by the QPSO algorithm. In this new model, the shortcoming that the BP network algorithm is easy to fall into local minimum value has been avoided. Based on the data of rice blast occurrences over the years in Fengqing County, Yunnan, China, the G-P algorithm was applied to study their K entropy and inserting space dimension. It was found that their values were all positive. So the rice blast occurrence has some chaos characteristics, and the number of model input layer can be determined. This model was used to make forecasts of rice blast in this county during 2001-2009, and was compared with other forecasting models. The results proved that the accuracy rate and convergence velocity of this model were higher than other forecasting models, and its forecasting results were also effective and believable. The new model can provide a new way to solve the problems, such as pattern recognition, classification, and forecasting of rice blast.