Outlying subspace detection for high-dimensional data

Abstract

[Conclusion]: This article formulates the outlying subspace detection problem and provides a survey of the existing methods for solving this problem. In particular, it focuses on the metrics used to measure the outlier quality of given data points in different subspaces and the searching strategies employed by the existing techniques for exploring high-dimensional space lattices. We have also pointed out the major limitations of the existing techniques, and some important issues to be considered in developing a better outlying subspace detection method in future research work.