首先讨论支持向量机( SVM)的基本思想和实现过程,随后着重对SVM核函数进行探讨,从理论上研究常用核函数的选择优化问题.采用UCI数据库中的玻璃识别数据、菖蒲植物数据以及汽车评估数据分别对选择不同的核函数情况进行实验仿真分类和比较.仿真结果表明,同类数据选择不同核函数会产生不同的分类效果,选取合适的核函数对分类效果有很大的影响.%The basic theory and realization of support vector machines are described. Then the kernel functions of SVM and its optimization problem are discussed. The different kernel function selections are carried out simulation experiments with glass identification database, iris plants database and car evaluation database of UCI. The simulation results show that the data of same type with different kernel function have different classification efficiency and there is great influence for classification efficiency with right selection of kernel function.
展开▼
