An Online Support Vector Machine Algorithm for Dynamic Social Network Monitoring

Abstract

This paper introduces a novel One-Class Support Vector Machines (OC-SVM) algorithm for online social network monitoring, overcoming limitations of existing methods. Utilizing nodal and network-level attributes, it excels in diverse applications and efficient change point detection. The algorithm’s well-defined training data dictionary and updating procedure enhance memory and time efficiency. Experimental results, using an EpiCNet model and Enron Email network, highlight superior accuracy in detecting change points, affirming OC-SVM’s significant advancement in social network monitoring for real-world applications.