目前很多已知的聚类算法对于异常点的处理存在不合理的问题,将模糊集和粗糙集的相关理论加入到支持向量聚类算法中,可增加异常点处理的合理性,并得到一种新的改进算法,将其称为模糊—粗糙支持向量聚类算法.当支持向量集作为一个特殊的聚类,通过元素间的亲密程度,模糊边界的隶属度可以被计算出来.而下近似集包含的样本点建立在算法训练阶段获得的超球体内.在检测异常值和计算任意轮廓的聚类方面,该算法具有较大的优势和潜力. At present, many known clustering algorithms have an unreasonable problem for the processing of outliers. The related theories of fuzzy sets and rough sets are added to the support vector clustering algorithm, which can increase the rationality of outliers processing and get a The new improved algorithm is called fuzzy-rough support vector clustering algorithm.When the support vector set is used as a special cluster, membership of the fuzzy boundary can be calculated by the degree of intimacy between the elements, while the lower approximation set The included sample points are built in the hypersphere obtained during the training phase of the algorithm, which has great advantages and potential in detecting outliers and computing the clusters of arbitrary contour.