A Framework for Global Outlier Detection from Distributed Data Streams

Abstract

Distributed data streams are continuous streams of data that come from different sources across
the globe in real time. It is different from traditional centralized data streams and throws challenges in
detecting outliers from such dynamic data streams. The existing methods used to handle such data streams
are built from different perspectives. However a framework that can perform global outlier detection from
such distributed data streams is still desired. In this paper, we proposed and implemented a framework
known as Distributed Global Outlier Detection (DGOD) framework. The generated global outliers are
consistent with the centralized paradigms for outlier detection. The framework also takes care of privacy
while performing the detection process. Optimizations are suggested to further improve the performance of
the framework. The experimental results revealed the significance of the proposed framework in mining
outliers from distributed data streams.
Keywords – Outlier, outlier detection,