Research on real-time outlier detection over big data streams

ABSTRACT

Nowadays technological advances have promoted big data streams common in many applications, including mobile internet applications, internet of things, and industry production process. Outliers should be detected from big data streams in many cases. However, the special characteristics of big data streams, such as transiency, uncertainty, multidimensionality, dynamic distribution, and dynamic relationship make outlier detection more challenging. This paper discusses the key issues, major challenges, and existing most frequently used methods for detecting outliers over big data streams, and then summarizes the directions for further investigation. This research can provide novel theoretical support and technical guidance for outlier detection over big data streams.

KEYWORDS:

• Big data
• data stream
• outlier detection
• streams mining
• real-time analytics

Acknowledgements

I sincerely express my gratitude to the reviewers for their constructive comments, which helped improve the quality of this paper.

Additional information

Funding

This work was partially supported by Fundamental Research Funds for the Central Universities of China University of Labor Relations [grant number 17ZY008], Education and Teaching Reform Project of China University of Labor Relations [grant number JG1640], Research Project of China University of Labor Relations [grant number 17YY004], National Natural Science Foundation of China [grant number 51479155], Natural Science Foundation of Hubei Province of China [grant number 2014CFB190], Science and Technology Department of Guizhou Province of China [grant number Qian Ke He Fu Qi[2015]4006], and finished under the help of my PhD supervisor and schoolmates. I am sincerely grateful to them for their aspiring guidance and support.