为解决支持向量机(SVM)参数在优化过程中存在的局部极值和收敛速度慢的问题,提出一种基于矢量矩免疫算法优化SVM参数的方法.通过将抗体距离与免疫网络原理中浓度调节机制相结合的方式,提高算法的局部搜索能力,通过引入免疫记忆单元加快算法搜索最优参数的速度,优化过程中用SVM的分类精度作为算法的循环条件,实现对不同分类问题SVM参数的自适应调节.最后,利用Matlab7.0软件进行计算机仿真并与遗传算法进行比较,结果表明前者在优化性能上具有一定的优越性,为应用提供了参考.%Aiming at local maximum and slow convergence rate in parameters optimization process of support vector machine ( SVM ), a method of parameters optimization was proposed for SVM based on immune algorithm of vector distance. The ability of local search was enhanced by vectors distances of antibody combining with the mechanism of antibody density regulation. The search speed of optimal parameters was accelerated by immune memory cells. The classification precision of SVM was used as circulating condition in the optimization process, SVM parameters self-adaptive adjustment were achieved for different classification. Finally, the computer simulation was done by Matlab7.0 software and the optimization algorithm was compared with the genetic algorithm. Results showed that the optimized performance of the former had some advantages.
展开▼
