粒子群优化算法工具箱及其在预测中的应用

A Prediction Method Based on Particle Swarm Toolbox to Optimize RBF

粒子群优化算法工具箱(PSOt)具有应用方便,无需复杂编程,功能强大等特点。该文提出了一种采用粒子群算法工具箱优化RBF网络的预测模型,即利用PSOt优化RBF网络隐层中心、基宽向量及输出权值等,将网络参数编码为PSOt算法中的粒子个体,在全局空间中搜索最优适应值参数,构成PSOt-RBF预测模型。将该模型应用于某矿生产数据的预测研究,并与常规粒子群优化算法预测结果进行比较。仿真实验表明PSOt-RBF模型预测精度优于PSO-RBF和单一的RBF预测模型,实例验证了所提方法的有效性和实用性。

Particle swarm optimization algorithm toolbox(PSOt) is easy to application, without complex programming, powerfulfunctions, etc. This paper proposes a particle swarm optimization toolbox to optimize RBF network forecasting model, namely PSOtis used to optimize the hidden layer of RBF network center, the wide base vectors and weight etc., the network parameters are encoded as PSOt algorithm particles in the individual, in the global space it is used to search the optimal parameters, so a PSOt-RBFprediction model is constituted. The model is applied to mine production data prediction research, and compared with conventionalparticle swarm optimization algorithm prediction results. Simulation results show that PSOt- RBF model prediction accuracy is better than PSO-RBF and the single RBF model, and proved the validity and practicability of the proposed method.