Link Prediction Regression for Weighted Co-authorship Networks

P. 667–677.

Герасимова О. А., Макаров И. А.

In this paper, we study the problem of predicting quantity of collaborations in co-authorship network. We formulated our task in terms of link prediction problem on weighted co-authorship network, formed by authors writing papers in co-authorship represented by edges between authors in the network. Our task is formulated as regression for edge weights, for which we use node2vec network embedding and new family of edge embedding operators. We evaluate our model on AMiner co-authorship network and showed that our model of network edge representation has better performance for stated regression link prediction task.

Ключевые слова: машинное обучение, рекомендательные системы, machine learning, link prediction, Recommender Systems, Co-authorship networks, Network embedding, граф соавторства, векторное представление графа

ПУБЛИКАЦИЯ ПОДГОТОВЛЕНА ПО РЕЗУЛЬТАТАМ ПРОЕКТА: Приобретение и представление знаний для рекомендательных систем (2019)

В книге

Advances in Computational Intelligence. IWANN 2019

Berlin: Springer, 2019.