Detection of Unnatural User Behavior in Social Media Using Machine Learning Technologies

Abstract

Information from social media influence public sentiment, the spread of opinions and reactions to events. The spread of often false information and the incitement of people in sensitive topics are often influenced by artificially created users for certain events. In other words, such users are called bots, which should be quickly detected and blocked in order to stop the promotion of certain profitable questions to customers. The detection process can be automated with the help of trained models based on the collected data and clustering of users according to the features inherent in bots. Among the mentioned differences are information about the number of subscribers and followers, activity of publishing posts, photo, date of profile creation and others. For training and testing models, a method of automated data collection using the Selenium framework from user pages is described. Correlation of various attributes is shown, and classification results are shown for identifying users with unnatural behavior in social media.