Publication Type

Conference Proceeding Article

Publication Date

11-2016

Abstract

Learning from data streams has been an important open research problem in the era ofbig data analytics. This paper investigates supervised machine learning techniques formining data streams with application to online anomaly detection. Unlike conventionalmachine learning tasks, machine learning from data streams for online anomaly detectionhas several challenges: (i) data arriving sequentially and increasing rapidly, (ii) highlyclass-imbalanced distributions; and (iii) complex anomaly patterns that could evolve dynamically.To tackle these challenges, we propose a novel Cost-Sensitive Online MultipleKernel Classification (CSOMKC) scheme for comprehensively mining data streams anddemonstrate its application to online anomaly detection. Specifically, CSOMKC learns akernel-based cost-sensitive prediction model for imbalanced data streams in a sequential oronline learning fashion, in which a pool of multiple diverse kernels is dynamically explored.The optimal kernel predictor and the multiple kernel combination are learnt together, andsimultaneously class imbalance issues are addressed. We give both theoretical and extensiveempirical analysis of the proposed algorithms.

Keywords

Cost-Sensitive Learning, Online Learning, Multiple Kernel Learning

Discipline

Categorical Data Analysis | Databases and Information Systems

Research Areas

Data Management and Analytics

Publication

Volume 63: Proceedings of The 8th Asian Conference on Machine Learning: Hamilton, New Zealand, 2016 November 16-18

First Page

65

Last Page

80

ISSN

1532-4435

Publisher

Microtome Publishing

City or Country

Brookline, MA

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Additional URL

http://www.jmlr.org/proceedings/papers/v63/

Share

COinS