A unified framework for effective team formation in social networks

作     者：Selvarajah, Kalyani Zadeh, Pooya Moradian Kobti, Ziad Palanichamy, Yazwand Kargar, Mehdi 

作者机构：Univ Windsor Sch Comp Sci Windsor ON Canada Ryerson Univ Ted Rogers Sch Management Toronto ON Canada 

出 版 物：《EXPERT SYSTEMS WITH APPLICATIONS》 (专家系统及其应用)

年 卷 期：2021年第177卷

页      面：114886-114886页

核心收录：

学科分类：1201[管理学-管理科学与工程(可授管理学、工学学位)] 0808[工学-电气工程] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：We would like to thank the anonymous reviewers for their comments that greatly improved the quality and presentation of this manuscript 

主　　题：Team formation Social networks Graph data management Multi-objective optimization 

摘      要：Collaboration networks are social networks in which nodes represent experts, and edges represent the interactions between them. Team Formation Problem (TFP) in Social Networks (SN) is to construct a group of individuals to work on complex tasks. Teams should satisfy the skill set required by the tasks and can collaborate effectively under multiple constraints. Although many algorithms have been proposed to confront the TFP, most of them optimize different criteria and various parameters (e.g. communication cost or expertise level). There is no unified framework to incorporate the most significant parameters towards formulating effective teams of experts. We propose a unified framework for the TFP in SN based on a multi-objective cultural algorithm that involves the integration of essential cost functions such as communication cost, expertise level, collective trust score, and geological proximity. Since these are conflicting objectives, we return a set of Pareto front of teams that are not dominated by other feasible teams with regards to any of the objectives. Moreover, we examine the temporal nature of both communication costs and expertise levels in our model and introduce a new method to formulate them. We introduce a profile similarity formula to express the trust score. We then discuss the importance of emotional index in TFP. Our model is tested on a benchmark table, which is generated with various criteria of social networks. Our model is then compared with NSGA II, Graph-Based and Exhaustive search.