基于改进粒子群优化RBF神经网络的地理信息预测

本文首先针对标准粒子群优化算法容易陷入局部最优的缺点,采用动态自适应调节策略,使得粒子的惯性权重随群体聚集程度而适时变化,从而调整粒子群搜索的速度和方向以跳出局部最优;然后将粒子群算法的全局搜寻能力和RBF网络的局部优化能力相结合,利用改进的粒子群优化算法优化RBF神经网络的关键参数;并将其应用于地理信息的预测,得到满意的结果。

Geographic information prediction based on RBF neural network optimized by improved particle swarm algorithm

In the paper,first,the dynamic adaptive strategy was introduced into the PSO algorit,because of the shortcoming of vulnerable to local optimum,the inertia weight of the particles changes with the degree of population to adjust the particle swarm speed and direction of the search.Combining the particle swarm global search capability and capacity of local optimization of RBF network,the key parameters of the RBF neural network were optimized by improved particle swarm optimization algorithm.It was applied in geographic information prediction,and has obtained satisfactory results.