.png){kind=icon}
.png){kind=icon}
Return
Predicting cross-domain collaboration using multi-task learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2024.124570.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Collaboration across different domains has become a promising fashion in the academic field. However, it faces significant challenges that prevent researchers from forming new collaborative relationships across domains. Current methods only focus on the homogeneity between researchers, neglecting the diverse motivations and complex interactions that underlie cross-domain collaboration. As a result, they lost many opportunities to identify potential collaborators. In this work, we propose a novel Cross-Domain Collaboration Prediction (CDCP) model that captures the diverse motivations of researchers using semantic encoding based on metapaths and models the complex interactions between research topics using bibliometric methods. In the CDCP model, we propose to jointly optimize two tasks, i.e., collaboration prediction and collaborative pattern prediction, by adopting a multi-gate network that flexibly leverages both structural and semantic features. Our model can effectively learn from both tasks and enhance its predictive performance and generalization ability. Extensive experimental studies on large publication datasets from various domains show that our method outperforms state-of-the-art methods on multiple indicators and can effectively predict diversified collaborations for researchers.
Keywords:
Collaboration prediction
Cross-domain collaboration
Graph representation learning
Multi-task learning
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W