Recombination Operators for the Multi-Objective Team Formation Problem in Social Networks

Julio Juárez, Carlos A. Brizuela, Hugo Terashima-Marín, Carlos A. Coello Coello

Abstract: The Team Formation Problem in Social Networks (TFP-SN) describes the process of finding an effective group of people, drawn from a network of experts, to perform a particular task. For a team to be considered as effective, it requires to comply with a task-specific skills set while also showing a high degree of cohesiveness. Although team effectiveness is subject to multiple criteria, the study of the problem from a multi-objective (MO) perspective is still scarce. In this paper, we focus on an MO TFP-SN whose objective is to maximize the team's level of expertise and the team's density, simultaneously. To solve this problem, we introduce two novel recombination operators to be used within the framework of the well-known NSGA-II. Our proposed crossover operators act as heuristics that compute the parents' unique and shared information, which is then combined for generating potentially improved offspring. Our experiments show that each of the two proposed crossover operators lead to significantly better results when compared to a naive crossover operator taken from the specialized literature. Particularly, the results consistently show higher hypervolume values when compared to the use of an adaptation of a simple recombination operator commonly used for this problem. The good performance of our proposed operators may be attributed to the incorporation of knowledge that exploits the structure of the problem.