Automatic selection of a single solution from the Pareto front to identify key players in social networks

Social networks are increasingly growing and identifying the relevant and influential users within a network is becoming a problem of interest in many contexts. Although multi-objective optimization approaches have been proposed to address this problem, they identify a large number of valid and non-dominated solutions, so selecting a relevant single solution over the others is a difficult task. Several methods for post-Pareto optimality analysis have been considered to reduce the Pareto fronts identified by the multi-objective optimization approach to a single solution. Eleven methods have been implemented, tested, and compared. Most of them have never been used before for a reduction task and/or for a key player context. The highest hypervolume method and the method based on the Euclidean distance to the ideal point combine the best averages and the lowest dispersions, reporting statistically significant differences with the rest of the methods. Improvements up to 60.79% have been obtained. This methodology will be implemented in an e-learning platform in order to identify the most relevant and influential students in the social network of a course.