A Rough Set Approach to Spatio-temporal Outlier Detection

Abstract

Detecting outliers which are grossly different from or inconsistent with the remaining spatio-temporal dataset is a major challenge in real-world knowledge discovery and data mining applications. In this paper, we deal with the outlier detection problem in spatio-temporal data and we describe a rough set approach that finds the top outliers in an unlabeled spatio-temporal dataset. The proposed method, called Rough Outlier Set Extraction (ROSE), relies on a rough set theoretic representation of the outlier set using the rough set approximations, i.e. lower and upper approximations. It is also introduced a new set, called Kernel set, a representative subset of the original dataset, significative to outlier detection. Experimental results on real world datasets demonstrate its superiority over results obtained by various clustering algorithms. It is also shown that the kernel set is able to detect the same outliers set but with such less computational time.