Abstract: In support vector classification，different features effect differently on classification.For this problem，a generalized Gaussian kernel function is proposed by using features weight，particle swarm optimization is used to do the model selection with the k-fold cross-validation error as the fitness.Then experiment results on UCI dataset show that the method can improve the performance of SVM.

Key words: support vector machine, feature weight, generalized Gaussian kernel, cross validation, particle swarm optimization

摘要： 支持向量机分类中，不同属性对分类的贡献往往不同，针对此问题，在核函数中引入属性权重，提出一种推广的高斯核模型，同时以最小化k-fold交叉验证误差为目标，采用粒子群算法进行推广高斯核的模型选择。最后通过UCI上标准数据集进行实验，证实该方法能够有效提高支持向量机的推广能力。

关键词: 支持向量机, 属性权重, 推广高斯核, 交叉验证, 粒子群优化