Directed Graph-based Researcher Recommendation by Random Walk with Restart and Cosine Similarity
Kanta Nakamura, Kazushi Okamoto · 2020
In this study, we propose a researcher recommendation model using random walk with restart on a directed graph. The directed graph is constructed from the grants-in-aid for scientific research database, and three types of their edges are defined based on the history of research collaboration, the cosine similarity of research contents, and their combination. An evaluation experiment, which measures$\text{nDCG}@ k$, recommendation accuracy, for each model applied the three edge functions, is performed. As the experimental result, it is confirmed that$\text{nDCG}@ k$for the combination model of the history and cosine similarity edges achieves the best accuracy.