Research Partner Recommender System for Academia in Higher Education using Non-Linear Approach

Academic collaboration can be grouped by different faculty members. It involves searching for relevant research topics to collect and analyze information and establishing a new research partner collaboration. A research partner recommender system for academia in higher education can help reduce the time and effort required to establish a new collaboration. Naive Bayes classification makes recommendations with dynamic text search as input values that may extract not only relevant to research topics but also object locations and case studies into consideration. However, classification as output representation is not satisfactory for academia. Therefore, this paper proposes a non-linear approach to provide a score value instead of classes for more suitable relevant recommendations that can leverage a powerful Sigmoid activation function. We demonstrate our approach using actual data from faculty members at the Faculty of Information Technology and Business in a private university in Indonesia. The proposed web-based system helps increase recommendation accuracy for new users.