SRRS: Design and Development of a Scholarly Reciprocal Recommendation System

The aim of this work is to propose a hybrid reciprocal recommendation algorithm for cold-start authors in a network based on text information and network-based features. The proposed algorithm is a novel collaborative filtering algorithm that combines text information with network features for more accurate and personalized recommendations. The feature importance values are used to understand the impact of each feature on the prediction and to identify the most important features for a given task. In the proposed algorithm, a community detection algorithm is used in addition to the baseline method, which uses a first-order neighborhood approach. Furthermore, varying T on edge weights in the co-author graph with optimal T is used to obtain hybrid recommendations in the same community. The results demonstrate that the proposed method is effective in predicting collaborators for cold-start authors in the network.