期刊名称：International Journal of Advanced Computer Science and Applications(IJACSA)
印刷版ISSN：2158-107X
电子版ISSN：2156-5570
出版年度：2019
卷号：10
期号：12
DOI：10.14569/IJACSA.2019.0101242
出版社：Science and Information Society (SAI)
摘要：Edge computing extends cloud computing to enhancing network performance in terms of latency and network traffic of many applications such as: The Internet of Things (IoT), Cyber-Physical Systems (CPS), Machine to Machine (M2M) technologies, Industrial Internet, and Smart Cities. This extension aims at reducing data communication and transmission through the network. However, data processing is the main challenge facing edge computing. In this paper, we proposed a data processing framework based on both edge computing and cloud computing, that is performed by partitioning (classification and restructuring) of data schema on the edge computing level based on feature selection. These features are detected using MapReduce algorithm and a proposed machine learning subsystem built on user requirements. Our approach mainly relies on the assumption that the data sent by edge devices can be used in two forms, as control data (i.e. real-time analytics) and as knowledge extraction data (i.e. historical analytics).We evaluated the proposed framework based on the amount of transmitted, stored data and data retrieval time, the results show that both the amount of sending data was optimized and data retrieval time was highly decreased. Our evaluation was applied experimentally and theoretically on a hypothetical system in a kidney disease center.
关键词：Edge computing; cloud computing; data processing; data partitioning; MapReduce; machine learning; feature selection; user requirement
