基于模糊分割和邻近对的支持向量机分类器

支持向量机算法对噪声点和异常点是敏感的，为了解决这个问题，人们提出了模糊支持向量机，但其中的模糊隶属度函数需要人为设置。提出基于模糊分割和邻近对的支持向量机分类器。在该算法中，首先根据聚类有效性用模糊c-均值聚类算法分别对训练集中的正负类数据聚类；然后，根据聚类结果构造c个二分类问题，求解得c个二分类器；最后，用邻近对策略对样本点进行识别。用4个著名的数据集进行了数值实验，结果表明该算法能有效提高带噪声点和异常点数据集分类的预测精度。

Support vector machine classifier based on fuzzy partition and neighborhood pairs

Support Vector Machine (SVM) is sensitive to noises and outliers. To overcome this drawback, Fuzzy Support Vector Machine (FSVM) is developed, in which the fuzzy membership function is set subjectively. In this study, support vector machine classifier based on fuzzy partition and neighborhood pairs (FPNP-SVC) was presented to deal with the classification problems with noises or outliers. In the proposed algorithm, fuzzy c-means clustering was firstly adopted to cluster each of two classes from the training set based on the clustering validity; Then c binary classification problems were formed based on the clustering results; Finally, based on neighborhood pairs strategy, for each sample a binary classifier constructed by two nearest subsets from two classes was chosen to identify it. The experiments were conducted on four benchmarking datasets for testing the generalization performance of FPNP-SVC. The experimental results show that FPNP-SVC is valid for improving the prediction accuracy of the classification problems with noises or outliers.