Title: Improved cyberbully detection techniques using multiple correlation coefficient from forum corpus
Authors: J.I. Sheeba; S. Pradeep Devaneyan; Prathyusha Tata
Addresses: Department of Computer Science and Engineering, Pondicherry Engineering College, Puducherry, India ' School of Mechanical and Building Sciences, Christ College of Engineering and Technology, Puducherry, India ' Sapient Corporation, Publicis.Sapient, Bangalore, India
Abstract: Today, there are many prominent online sites where people share their experiences regarding crimes and anti-social behaviour. In this regard, a major unaddressed and even unidentified problem that is experienced in the social network websites is cyberbully. This proposed framework primarily targets the cyberbullying in the crime investigation forum since a high degree of cyberbully is common in crime forums. In this paper, a highly furnished representational framework is proposed that is specific to cyberbully detection using hybrid techniques (multiple correlation coefficient - MCC and support vector machine - SVM). The bag of words are given individual weights to examine their correlations using MCC algorithm before feeding them into a linear SVM classifier that identifies and classifies the cyberbully words. The efficiency of the system developed can be enhanced by analysing the evaluation metrics and the dataset validation metrics.
Keywords: cyberbully detection; cyberbully classification; multiple correlation coefficients; MCCs; support vector machines; SVMs; data analytics; DAs.
International Journal of Autonomic Computing, 2018 Vol.3 No.2, pp.152 - 171
Accepted: 06 Nov 2018
Published online: 31 Jan 2019 *