Noname manuscript No. (will be inserted by the editor) An Evolutionary Non-Linear Ranking Algorithm for Ranking Scientific Collaborations Fahimeh Ghasemian · Kamran Zamanifar · Nasser Ghasem-Aghaee Received: date / Accepted: date Abstract The social capital theory motivates some researchers to apply link-based ranking algorithms (e.g. PageRank) to compute the fitness level of a scholar for collaborating with other scholars on a set of skills. These algorithms are executed on the collaboration network of scholars and assign a score to each scholar based on the scores of his/her neighbors by solving a linear system in an iterative way. In this paper, we propose a new ranking algorithm by focusing on link-aggregation function and transition matrix. The evolution strategy technique is applied to find the best aggregation function and transition matrix for computing the score of a scholar in the collaboration network which is modeled by a hypergraph. Experiments conducted on two datasets gathered from ScivalExpert and VIVO show that the new non- linear ranking algorithm acts better than the other iterative ranking approaches for ranking scientific collaborations. Keywords Scientific Collaboration · Hypergraph · Ranking Algorithm · Evolution Strategy · Collaboration Network 1 Introduction Scientific collaboration is increasing in frequency and importance [41]. Scholars require to collaborate with each other to share and complement their knowledge and skills and access to specialized equipment and data. The complexity of unsolved scientific problems is one of the reasons for this trend. Therefore, collaboration among the right individuals is essential for progress in science [37]. To answer this need, scientific collaboration is studied in diverse disciplines [41]. Especially the field of Science of Team Science (SciTs), focuses on understanding and enhancing collaborative process and outcomes and uses both qualitative and quantitative methods such as case studies, social network analysis and bibliometric analysis to study the scientific collaboration [42]. Identifying appropriate collaborators for teams is one of the key motivations for many of these studies and its prerequisite is determining the set of collaborators’ features that affect the collaboration outcome. Social theories from great scientists like Burt, Coleman and Granovetter motivate some researchers to model scientific collaboration as a graph and use the network analysis measures to study the scholar and collaboration performance [44], [5], [32], [29], [28]. Most of the findings indicate that social network structures are effective in collaboration performance since individual features are not the only factors in individual or group success and social capital which is determined based on social network structure of people is another important factor [2]. One of the most popular approaches for encountering the effect of social capital is ranking scholars using algorithms like PageRank. In this algorithm, the collaboration relations among scholars is modeled as a graph and the rank or score of each scholar is computed based on the scores of scholars who are connected Fahimeh Ghasemian Department of Software Engineering, Faculty of Computer Engineering, University of Isfahan, Isfahan, Iran E-mail: ghasemianfahime@gmail.com Kamran Zamanifar Department of Software Engineering, Faculty of Computer Engineering, University of Isfahan, Isfahan, Iran E-mail: zamanifar@eng.ui.ac.ir Nasser Ghasem-Aghaee Department of Software Engineering, Faculty of Computer Engineering, University of Isfahan, Isfahan, Iran E-mail: aghaee@eng.ui.ac.ir