What Characteristics Define a Good Solution in Social Influence Minimization Problems?

Abstract

The evolution of Social Networks has introduced significant challenges related to information overload. These challenges are covered in diverse areas, such as viral marketing or misinformation control. As social networks grow in complexity, the essential need to leverage data-driven insights becomes evident. The aim of the Social Influence Minimization Problem (IMP) is to identify and strategically block users to curtail information dissemination. Structural insights can be extracted through data-mining techniques to guide the design of efficient heuristics and the identification of influential users to be blocked. Considering good and bad quality solutions, a supervised learning approach is used to classify the extracted features, that allowed meaningful conclusions to be drawn regarding the features of these solutions. The IMP is addressed through the proposal of a robust heuristic method, based on the most relevant features, which is effective and efficient when compared with the state-of-the-art approaches.