Research of Gravity-Based Outliers Detection

A gravity-based outliers detection algorithm GODA is presented. Considering that for a data point, both of the density around it and the distance between it and others can influence the outlier´s definition, the algorithm can detect the crytic outliers in the dataset. This paper proposes the definitions and techniques firstly and then introducts the algorithm detailedly. Experiments have been carried out with real data, and the results indicates that not only the new algorithm is of goog extensible ability but also it has higher efficiency of detecting outliers. It points out the outlier´s outlying degree in the dataset as well.