A Unified Neuro-Fuzzy Framework to Assess the User Credibility on Twitter

Abstract

With the tremendous proliferation of social media in day-to-day life, social content has become one of the potential information sources. Ensuring social trust is crucial to averting the negative impact on viral marketing, expertise retrieval, and recommendation systems, leading to ascertainment of credibility of social media users. A unified framework is proposed for evaluation of the user credibility through analysis of three fundamental credibility-related factors from social media and e-commerce sites, namely, probability of being a promoter, a spam bot, and/or a spammer. The factors are estimated using deep learning baseline model and/or further analysed using Fuzzy Inference System for making a decision on Tweeter credibility. The framework is demonstrated on Twitter and Amazon benchmark datasets for evaluating its efficacy in identifying credible users. The proposed system outperformed the baseline model and state-of-the-art techniques providing an accuracy of 97% in detecting the credible user.