Machine learning on big data: Opportunities and challenges

NEUROCOMPUTING

卷:237

Article

Zhou, Lina¹  Pan, Shimei¹  Wang, Jianwu¹  Vasilakos, Athanasios V.² 

[1] UMBC, Dept Informat Syst, Baltimore, MD 21250 USA

[2] Lulea Univ Technol, Dept Comp Sci Elect & Space Engn, SE-93187 Skelleftea, Sweden

关键词: Machine learning;    Big data;    Data preprocessing;    Evaluation;    Parallelization;   

DOI  :  10.1016/j.neucom.2017.01.026

来源: Elsevier

【 摘 要 】

Machine learning (ML) is continuously unleashing its power in a wide range of applications. It has been pushed to the forefront in recent years partly owing to the advent of big data. ML algorithms have never been better promised while challenged by big data. Big data enables ML algorithms to uncover more fine-grained patterns and make more timely and accurate predictions than ever before; on the other hand, it presents major challenges to ML such as model scalability and distributed computing. In this paper, we introduce a framework of ML on big data (MLBiD) to guide the discussion of its opportunities and challenges. The framework is centered on ML which follows the phases of preprocessing, learning, and evaluation. In addition, the framework is also comprised of four other components, namely big data, user, domain, and system. The phases of ML and the components of MLBiD provide directions for identification of associated opportunities and challenges and open up future work in many unexplored or under explored research areas.

【 授权许可】

Free