Multi-team Formation using Community Based Approach in Real-World Networks

TL;DR

This paper introduces a community-based team formation algorithm that efficiently constructs teams within large social networks by leveraging community structures, outperforming traditional skill-centric methods in communication cost and scalability.

Contribution

The paper presents a novel community-based approach for team formation that is scalable and effective in large networks, unlike existing skill-centric algorithms.

Findings

TFC-N and TFC-R outperform standard algorithms in communication cost.

The proposed algorithms are several orders faster on community structures.

They achieve a good balance between speed and communication efficiency.

Abstract

In an organization, tasks called projects that require several skills, are generally assigned to teams rather than individuals. The problem of choosing a right team for a given task with minimal communication cost is known as team formation problem and many algorithms have been proposed in the literature. We propose an algorithm that exploits the community structure of the social network and forms a team by choosing a leader along with its neighbours from within a community. This algorithm is different from the skill-centric algorithms in the literature which start by searching for each skill, the suitable experts and do not explicitly consider the structure of the underlying social network. The strategy of community-based team formation called TFC leads to a scalable approach that obtains teams within reasonable time over very large networks. Further, for one task our algorithms TFC-R and TFC-N generate multiple teams from the communities which is show-cased as a case-study in the paper. The experimentation is carried out on the well-known DBLP data set where the task is considered as writing a research paper and the words of the title are considered as skills. Team formation problem is translated to finding possible authors for the given paper, who have the required skills and having least communication cost. In the process, we build a much larger bench-mark data set from DBLP for team formation for experimentation. Even though the benchmark algorithm Rarestfirst takes least time, our algorithms TFC-N and TFC-R give much better communication cost. They also outperform the standard algorithms like MinLD and MinSD with respect to the time taken in finding a team. The time taken by our algorithms on communities are several orders faster than the time taken on the larger network without compromising too much on the communication cost.