Adaptive k-Nearest Centroid Neighbor Classifier for Detecting Drifted Twitter Spam
L A Lalitha1, Vishwanath R Hulipalled2
1L A Lalitha, Department of C&IT, REVA University, Bengaluru (Karnataka), India.
2Dr. Vishwanath R Hulipalled, Department of C&IT, REVA University, Bengaluru (Karnataka), India.
Manuscript received on 05 June 2019 | Revised Manuscript received on 14 June 2019 | Manuscript Published on 29 June 2019 | PP: 235-243 | Volume-8 Issue-5S, May 2019 | Retrieval Number: E10480585S19/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: With the growth of Internet and its related technologies have resulted in increased usage of smart and Internet connected devices and large amount of time is spent on Social Network. Nonetheless, because of increase in attractiveness of Social Network, cyber offenders are spreading spam on these networks to exploit possible targets. The spammers trap users to malware downloads or external phishing URLs, which has been an enormous problem for online safety and user quality of exposure. However, the existing research fails to detect spam in Twitter and has become a key issue in recent times. Recent work [14], focused on using Machine Learning (ML) approach for detecting spam in Twitter, by making use of the statistical features of Twitter data. However, adoption of such method affects the classification accuracy of ML algorithm. Because the Statistical Feature characteristics of spam tweets vary with respect to time. This problem is known as “Twitter Spam Drift”. To address this problem, we present a novel non-parametric Adaptive K-Nearest Centroid Neighbor (AKNCN) Classifier. Further, for meeting real-time requirement the AKNCN is trained using one million spam tweets and one million non-spam tweets data. The AKNCN model can discover spam more efficiently than the state-of-the-art model. Experiment outcome shows the AKNCN attains significant performance with reference to Accuracy (A), F-Measure (F) and Detection Rate (DR) in real-world scenarios.
Keywords: Nearest Centroid Neighbor, Machine Learning, Social Networks, Statistical Features, Spam Drift, Twitter Spam Detection.
Scope of the Article: Adaptive Networking Applications