一种基于PSO的混合核支持向量机算法

摘要

As a new machine learning method, support vector machine algorithm has many obvious advantages in solving the problem of small samples. However, it is important to select an optimal kernel function and parameters in order to enhance the performance of support vector machine algorithm. In this paper, the support vector machine algorithm of hybrid kernel based on PSO is proposed by the mixed kernel function method combined with global kernel function and local kernel function. After Matlab simulation experiment, the results show that the improved algorithm is superior to standard SVM in respect of classification accuracy, learning and generalization ability.%支持向量机算法作为一种新的机器学习方法,在处理小样本分类问题上具有明显优势,但核函数和参数的选取的好坏直接影响支持向量机算法的性能.针对该问题,通过组合全局核函数和局部核函数的混合核函数方法,建立了基于粒子群算法的混合核支持向量机算法,并经过Matlab仿真实验,表明该改进算法较支持向量机算法具有更高的分类准确率和更好的学习及泛化能力.