粒子群算法优化RBF-SVM沙尘暴预报模型参数

为提高沙尘暴的预报准确率，针对目前已出现的RBF—SVM沙尘暴预报模型中的参数优化进行研究．利用基本粒子群优化算法（SPSO算法）中粒子速度及其位置与RBF—SVM模型中参数对相对应，对沙尘暴进行预报，为解决SPSO算法易陷入局部解的缺陷，提出了惯性权值自适应调节的改进粒子群算法（WPSO算法），并对沙尘暴RBF—SVM模型参数进行了优化．仿真结果表明，无论是SPSO算法，还是WPSO算法，在优化RBF—SVM沙尘暴预报模型参数方面都表现出了良好的性能，分别比已有的SVM方法的预报准确率提高了22．3％和45．3％．

Parameters Optimization of RBF-SVM Sand-Dust Storm Forecasting Model Based on PSO

To improve the accuracy of sand-dust storm forecasting, an RBF-SVM method with automatic parameter selection was presented in this paper.The proposed method used the simple particle swarm optimization （ SPSO ） algorithm to get the optimal parameter, in which the velocity and position of each particle correspond a group of RBF- SVM parameters.However, since the PSO tends to get into local optimal solutions, a weight particle swarm optimization （ WPSO ） algorithm was proposed, in which the weights changed dynamically with a liner rule, to optimize the parameters of RBF-SVM.The simulation results show that both PSO-RBF-SVM and WPSO-RBF-SVM can get high recognition accuracy and efficiency.And the accuracy ratios of two kinds of sand-dust storm forecasting are improved by 22.3% and 45.3% compared with the previous SVM, respectively.