Identifying a Team of Experts in Social Networks using a Cultural Algorithm

Abstract

We study the problem of identifying a team of experts in a social network. In this problem, each individual (i.e., expert) posses a set of skills based on its past experience or training. Experts form a social network based on their past collaboration with each other: if two experts work on the same project in the past, they are connected in the underlying social network. The success of a project depends on finding the right experts that cover all the required skills. In addition to covering the required skills, we prefer the experts in the same team have past collaboration. Past collaboration expedites the completion of the project since team members already know each other and get along well. In order to differentiate different teams from each other, we use the notion of the communication cost among team members. Clearly, the lower the communication cost, the more collaborative the team is. We use the diameter and the sum of distances as communication cost functions. Therefore, our goal is to find a team of experts that while covering all the required skills, also has minimum communication cost. Optimizing each of these functions are proved to be NP-hard. In this paper we propose a knowledge-based evolutionary optimization algorithm to solve this problem. We perform extensive experiments on large DBLP network to show the viability of our approach in comparison to the state of the art.