Anomaly detection in social media using deep learning approach

Abstract

Todayand#8223;s modern world makes the utilization of social media newlineinevitable, as it helps to connect with people. Besides, the social media are newlinepublic platforms that can reach huge crowds as well as allow people to share newlinetheir personal perspectives, news, opinions and so on. Hence, it is a powerful newlineweapon of this era. Despite all the merits, social media also suffer from newlineseveral issues. One of the major issues is the detection of abnormal activities. newlineUnusual behaviour in social networks refers to abnormal and illegal acts that newlinebehave differently from other users of the same structure. newlineMalicious users often involve in abnormal activities, which could newlineimpact on the society. There are many reasons for unusual behaviour and the newlinepredominant reasons are commercial, political and clueless. As social media newlineare open platforms, different classes of users including malicious users newlineparticipate in them. Subsequently, it cannot be claimed that all the users in newlinesocial media are reliable and genuine. Yet, differentiation of reliable and newlineillegal user of the social media is quite difficult upon large size and complex newlinenature of social media. newlineAs a result, anomaly detection in social media is one of the popular newlineresearch trends of 21st century. Considering the importance of anomaly newlinedetection in social media, three solutions have been proposed for the same in newlinethis research. The initial stage of this research presents a rumour detection newlinescheme using different machine learning classifiers such as K-Nearest newlineNeighbor (KNN), Naive Bayes (NB) algorithm, Support Vector Machine newline(SVM) and Random Forest. The experimental results are compared with newlineEnsemble model and they have shown better performance results upon the newlinePHEME dataset extension. newline newline