一种改进的离群点检测方法

现有的离群点检测算法运用于规模较大的数据集时,其时间效率和检测效果通常不够理想.通过对离群点分布特征的分析,在计算每个数据点到其kth最近邻对象距离的同时,结合其k最近邻的分布情况,给出一种改进的离群点度量方法.基于上述思想构造的离群点检测算法DokOF能够处理混合属性数据.实验表明,该算法具有良好的适用性和有效性.

An improved efficient algorithm for detecting outliers

Existing algorithms for outlier detection are not efficient when facing large datasets.With analysis of features of outliers in datasets,we rank each point on the basis of its distance to its kth nearest neighbor and the distribution of its k nearest neighbors.A novel algorithm,called DokOF(outlier factor based on distribution of k nearest neighbors),is presented,which can deal with mixed attribute data.Results of experiments demonstrate promising availabilities of this approach.