MAT: Effective Link Prediction via Mutual Attention Transformer

Quan Van Nguyen, Huy Quang Pham, Dan Quang Tran, Thang Kien-Bao Nguyen, Nghia Hieu Nguyen

Abstract: The Data Science and Advanced Analytics (DSAA) 2023 competition [1] focuses on proposing link prediction methods to solve challenges about network-like data structure, such as network reconstruction, network development, etc., from articles on Wikipedia. In this challenge, our “UIT Dark Cow” team proposes the Mutual Attention Transformer (MAT) method to predict if there is a link between two Wikipedia pages. Our method achieved the 5th and 4th position on the leaderboard for the public and private tests, respectively. Our source code is publicly available for the ease of experimental re-implementation at the following link: https://github.com/minhquan6203/source-code-dsaa-2023.