A Recommender System for Research Collaboration Matching in Nigerian Universities

Research collaboration in Nigerian universities is largely facilitated through informal networks and conference encounters, limiting interdisciplinary and cross-institutional partnerships that could strengthen research output quality and volume. This study designs and evaluates a recommender system for research collaboration matching in Nigerian universities, leveraging faculty research profile data and semantic similarity techniques. A dataset of 2,400 faculty research profiles collected from university repositories and Google Scholar across six Nigerian universities spanning all geopolitical zones is used as the recommendation corpus. Three recommender approaches are compared: collaborative filtering, content-based filtering using TF-IDF and cosine similarity, and a hybrid approach combining both methods with a BERT-based semantic embedding component. Evaluation uses precision at k, recall at k, and a user relevance survey administered to 60 faculty members who reviewed generated recommendations. The hybrid BERT-enhanced approach achieves the highest precision at 5 (P@5 = 0.73) and is rated most relevant by faculty. Analysis of recommended collaboration networks reveals strong within-zone clustering and limited cross-zone recommendations, suggesting geographic proximity continues to bias research networking even in digital systems. Recommendations target NUC investment in a national faculty research profile platform to power the recommender system. Keywords: recommender system, research collaboration, Nigerian universities, BERT, semantic similarity