IFS-LSSVM及其在时延序列预测中的应用

IFS-LSSVM and its application in time-delay series prediction

针对最小二乘支持向量机预测模型中最优参数难以确定的问题,提出一种基于改进的自由搜索算法确定最小二乘支持向量机最优参数的方法(IFS-LSSVM)。对标准自由搜索算法进行改进,使之可应用于最小二乘支持向量机的参数优化,改进之后的算法具有更好的优化性能。将具有时间序列性质的网络时延作为预测对象,利用本文的IFS-LSVM算法进行预测。在仿真中与遗传算法优化的最小二乘支持向量机(GA-LSSVM)、粒子群优化算法优化的最小二乘支持向量机(PSOLSSVM)、标准最小二乘支持向量机工具箱中的网格搜索算法(Grid-LSSVM)进行了对比。仿真对比结果表明本文的方法具有更高的预测精度与更小的预测误差。

It is difficult to determine the optimal parameters of least squares support vector machine prediction model,so a prediction method based on improved free search algorithm（ IFS-LSSVM） was proposed to determine the optimal parameters of least squares support vector machines. First,the standard free search algorithm was improved so that it can be applied to the parameter optimization of least squares support vector machines,the improved harmony search algorithm has better optimization performance.Then the least squares support vector machines was applied to predict the time-delay series of the network based on improved free search optimization algorithm. Finally,time-delay series was used as prediction simulation object,genetic algorithm optimized least squares support vector machines（ GA-LSSVM）,particle swarm optimization algorithm optimized least squares support vector machines（ PSO-LSSVM）,standard grid search method of least squares support vector machines（ Grid-LSSVM） toolbox were compared. Simulation comparison results show that the proposed method has higher prediction accuracy and smaller prediction error.