Unsupervised Contribution-Oriented Learning Model for Social Influence Detection

Computer Engineering

Understanding​‍​‌‍​‍‌​‍​‌‍​‍‌ the influence of individuals on each other's social media and online platforms has become an indispensable tool for marketing, recommendation systems, and social behavior analysis in the digital world of today. To discover and profile influential users, the present work introduces an unsupervised, contribution-oriented learning method that does not require labeled data. The method examines each user's contributions to the network and, based on their influence patterns, groups them thus allowing the identification of most influential opinion leaders and trendsetters. Contrary to conventional supervised methods, our technique reveals the hidden influence structures directly from the raw interaction data without any manual intervention, thus it can be extended to different social networks and is flexible. Results from different real-world social media datasets testify to the success of the model in pinpointing influential users and giving an understanding of the changes in the network. The method is valuable to enterprises and researchers in the social influence field as it provides a way in which the phenomena can be studied that is more driven by data and less ​‍​‌‍​‍‌​‍​‌‍​‍‌biased

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.