PSO-ε-SVM的回归算法

Regression Algorithm of PSO-ε-SVM

回归支持向量机的ε不敏感损失函数的参数寻优是一个重要的问题,它与支持向量机的行为特性有紧密关系。本文给出了一种基于粒子群优化算法的、对ε不敏感损失函数的ε参数寻优的方法,仿真结果表明:采用基于粒子群优化算法的寻优方法寻找ε参数,需要重复训练回归支持向量机模型的次数明显小于格点搜索方法,节省了大量的时间并且能找到较优的ε值。

It is an important problem to find the optimized parameter of ε-non-sensitive loss function, because it is closely related with the behavior of support vector machine. A method is put forward to find the better ε of the ε-non-sensitive loss function based on the particle swarm optimization. The simulation result shows the regression support vector machines take far fewer times than using the way of latticework searching to get the desired result the efficience is improved greatly and optimal ε is obtained.