基于Bayes准则的支持向量机

支持向量机( SVM)在处理分类问题时，纯粹从样本的角度出发，其分类效果取决于训练样本的特性，不考虑待分类问题的当前信息。本文从导弹武器系统的数据交叉现象出发，通过对支持向量机的决策函数增加反映待分类问题当前信息的先验概率项，将Bayes准则融于支持向量机算法中，提高支持向量机的分类效果；给出了算法的推导以及实现步骤。通过导弹武器系统中的两个实例对算法进行验证，新算法比传统支持向量机算法具有更好的分类效果，并且算法的鲁棒性和敏感性都得到提高。

Support Vector Machines Based on Bayes Criterion

In handling classification problems, support vector machines (SVM) purely processes sam?ples and its effect of classification lies on character of the samples without considering current informa-tion of the classified problems. For resolving the phenomenon of data crossing in missile weapon sys?tems, the Bayes criterion was fused into the SVM by adding the prior probability item reflecting classi?fied problem present information to its decision function, to improve its classification effect; the deduc?ing process and realizing steps of the algorithm were given. The algorithm was verified by two in?stances .Results show that the new algorithm has better classification effect, the robust and the sensi?tivity than the normal SVM algorithm.