基于核可能性c-均值算法的支持向量机

提出一种基于核可能性c-均值算法的支持向量机分类算法,该算法改进了SVM训练过程中噪声和孤立点的敏感问题.其基本思想是:用核可能性c-均值算法对每个模式类训练样本进行聚类,得到不同的可能性度量值,根据得到的可能性度量值对训练样本进行修剪,删除具有较低可能性度量值的训练样本,最后用生成的新训练样本训练支持向量机.实验结果表明,该算法可以有效地解决由噪声和孤立点引发的分类错误问题以及重要样本的错分问题.

Support vector machine based on kernel possibilistic c-means algorithm

This paper proposes a support vector machine based on kernel positivistic c-means algorithm (KPCM-SVM) to improve the noises and outliers sensitivity problem of SVM for data classification. The basic idea is to assign different possibility metric to different training sample,the possibility metric used in KPCM-SVM is generated by kernel-based possibilistic c-means algorithm,whose partition gen- erates relative high values for important data points but low values for outlier.According to the possibili- ty metric,it prunes the training sample,deletes the sample whose possibility metric is smaller,then trains the SVM using the new training sample.Experiment results show that the proposed method re- duces the affect of outliers and yields higher classification rate than standard SVM to outliers in the training sample.