SVM参数优化方法分析与决策

针对支持向量机应用过程中的参数选择问题,从UCI数据库选择样本集,分别采用传统的网格法、智能优化算法中的粒子群法及遗传算法实现核函数参数寻优过程,将所得最佳参数应用到样本测试中。在深入分析优化过程中各参数关系、参数对支持向量机性能的影响以及传统与智能优化算法的优劣后,得出了核函数优化策略。即先使用智能优化算法初步确定最优解范围,再结合网格法进行高精度寻优。实验数据验证了参数优化策略的有效性,为扩大支持向量机泛化率、提高应用性做了铺垫。

Analysis and Strategy for Parameter Optimization Methods of SVM

To select parameters of support vector machine (SVM) during application, we choose sample sets from UCI database. The parameters optimization of SVM with kernel function is achieved through traditional grid method, particle swarm optimization algorithm and genetic algorithm in intelligent optimization algorithms. And the optimum parameters are applied to testing samples. The kernel function optimization strategy is obtained after a thorough analysis of the relationship between parameters, the influence of parameters on the performance of SVM and the pros and cons of both traditional and intelligent optimization algorithms. The users should use intelligent optimization algorithm preliminarily to make sure the scope of parameters, and couple it with grid method to obtain a high degree of accuracy. Experimental datas verify the effectiveness of the strategy, which expands the generalization and application of SVM.