Effective Features to Classify Big Data Using Social Internet of Things

被引：84

作者：

Lakshmanaprabu, S. K. ^{[1
]}

Shankar, K. ^{[2
]}

Khanna, Ashish ^{[3
]}

Gupta, Deepak ^{[3
]}

Rodrigues, Joel J. P. C. ^{[4
,5
,6
]}

Pinheiro, Placido R. ^{[7
]}

De Albuquerque, Victorhugo C. ^{[7
]}

机构：

[1] BS Abdur Rahman Crescent Inst Sci & Technol, Dept Elect & Instrumentat Engn, Madras 600048, Tamil Nadu, India

[2] Kalasalingam Acad Res & Educ, Sch Comp, Krishnankoil 626126, India

[3] GGSIP Univ, Assistant Maharaja Agrasen Inst Technol, Delhi 110078, India

[4] Natl Inst Telecommun, BR-37540000 Santa Rita Do Sapucai, MG, Brazil

[5] Inst Telecomunicacoes, P-1049001 Lisbon, Portugal

[6] ITMO Univ, St Petersburg 197101, Russia

[7] Univ Fortaleza, Grad Program Appl Informat, BR-60811905 Fortaleza, Ceara, Brazil

来源：

IEEE ACCESS | 2018年 / 6卷

关键词：

Internet of Things; social Internet of Things; machine Learning; big data; feature selection; DATA ANALYTICS; CLASSIFICATION;

D O I：

10.1109/ACCESS.2018.2830651

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

Social Internet of Things (SIoT) supports many novel applications and networking services for the IoT in a more powerful and productiveway. In this paper, we have introduced a hierarchical framework for feature extraction in SIoT big data using map-reduced framework along with a supervised classifier model. Moreover, a Gabor filter is used to reduce noise and unwanted data from the database, and Hadoop Map Reduce has been used for mapping and reducing big databases, to improve the efficiency of the proposed work. Furthermore, the feature selection has been performed on a filtered data set by using Elephant Herd Optimization. The proposed system architecture has been implemented using Linear Kernel Support Vector Machine-based classifier to classify the data and for predicting the efficiency of the proposed work. From the results, the maximum accuracy, specificity, and sensitivity of our work is 98.2%, 85.88%, and 80%, moreover analyzed time and memory, and these results have been compared with the existing literature.

引用

页码：24196 / 24204

页数：9