An Improved Fuzzy Support Vector Machine Algorithm for Highly Imbalanced Datasets in the Co-Authorship Recommendation Problem

The co-authorship recommendation problem aims to suggest authors join research groups based on their research topics, expertise, and previous collaborations. To address this problem, we first model it as a co-authorship network, where each author is represented as a vertex, and collaborations between authors are represented as edges. This allows us to generate two-class imbalanced datasets derived from the co-authorship networks. We then propose an adaptive weight adjustment algorithm based on FSVM-CIL to classify highly imbalanced two-class datasets. To evaluate the performance of our algorithm, we conducted experiments using self-built co-authorship datasets of various sizes and imbalance ratios. Our experimental results show that our algorithm outperformed FSVM-CIL in solving the co-authorship recommendation problem.