A Cluster-based Outlier Detection Method without Pre-Clustering

摘要

Outlier detection is critically important in the information-based society. It can lead to discovering useful anomalies, such as criminal activities in electronic commerce, terrorist threats, and agricultural pest infestations. This paper focus on cluster-based outlier detection approaches. We present a fast outlier detection method. Unlike the current cluster-based outlier detection approaches, where the outlier detection procedure follows a clustering process, our method avoids the pre-clustering procedure. Our method is tested experimentally over the real datasets and it shows around three-to-four-time improvement in speed compared to contemporary cluster-based outlier detection approaches.