| In the view of disadvantages of uncertain network structure in the BP neural network to establish model and big influence of different parameters selection on prediction accuracy for SVM model, a novel combining model was proposed. The parameters of penalty parameters C and σof the SVM model were optimized by particle swarm optimization algorithm with the chaos system (CPSO). Utilizing the uncertainty of chaotic systems, the shortcomings of slow convergence speed and easy to find the local optimal value were conquered by traditional PSO algorithm. In this way, the global optimal value could be faster and more accuracy to find with the novel approach of CPSO. The SVM model with parameter optimization effectively improved the prediction accuracy. The popular class of wheat stripe rust in Ningxia was predicted with proposed novel approach (CPSO-SVM). The results showed that, compared with the SVM model, PSO-SVM model and PSO -BP neural network model, the combining model were better than the above three kinds of models in prediction accuracy and generalization capability. The prevalence levels of wheat stripe rust can be accuracy predicted by the CPSOSVM algorithm and it can provide effective theory for agricultural pests and diseases research department. |
