Many-to-Many Collaborator Recommendation Based on Matching Markets Theory

As the academic data developing rapidly, it is crucially important to obtain more scientific achievements by helping scholars to make the right choices of potential collaborators among the big data. The conventional methods are to provide each scholar with a Top N recommendation list of candidates, in which some scholars are recommended to overmany collaborators and only guarantee the optimal results of them. To bridge this gap, in this work-in-progress, we present an academic collaborator recommendation method based on matching theory from the perspectives of maximizing preference and minimizing cost. The method adopts the multiple indicators extracted from the papers published by scholars to integrate the preference matrix among scholars. which applies matching theory to achieve a stable many-to-many matching of recommendation. It aims that each scholar can choose the collaborators they are keen on. After conducting abundant experiments on Microsoft Academic Graph, we demonstrate that this approach is definitely effective in improving recommendation accuracy and coverage.