Distinguishability of Outlier Based on Correlative Analysis

This paper is the sequential research of the "correspondence based outlier analysis". For correlative observations, the partial correlation coefficients are used to distinguish the multidimensional outliers, and the compound correlation coefficients are used to proof-test the holistic prominence of multidimensional outliers. By computing the correlation coefficient time and again, the authors locate the outliers of multidimensional outliers more accurately.