Feature Selection for Intelligent Detection of Targeted Influence on Public Opinion in Social Networks

Abstract

AbstractNowadays social networks become the very important mean of communication between people. But such networks are also used for malicious information dissemination and targeted influence on public opinion. The paper analyzes the characteristics of information dissemination in social networks. The contribution of the paper is the set of features that allow using machine learning methods to detect targeted influence on public opinion in social networks and distinguish the profiles responsible for it. The proposed features include: the dynamics of the number of subscribers, the dynamics of the number of likes on posts, the dynamics of the number of commentators on posts, the coherence of likes, the coherence of commentators, the dates of user registrations. The peculiarity of the proposed features lies in the selection of only those features that do not require content analysis. Also the paper authors propose the classification of the subjects that influence on the public opinion in social networks. The experiments were performed and proposed features and classification were tested.KeywordsSocial media analysisPublic opinionSocial networkInfluence model