Data Mining for Outlier With Minimum Vector Variance

Detecting outliers are an important data mining task. In this paper the authors propose the Minimum Vector Variance (MVV) for mining outliers. The MVV is robust measure and it has the high breakdown point. The measure is generated from the 'New' multivariate dispersion, which is called as Vector Variance (VV). They introduced the MVV to improve the properties of Minimum Covariance Determinant (MCD). In their experiences for mining outliers, the MVV is an effective measure.