Open Source Software Supply Chain Recommendation Based on Heterogeneous Information Network

In the GitHub open-source collaborative development scenario, each entity type and the link relationship between them have natural heterogeneous attributes. In order to improve the accuracy of project recommendation, it is necessary to effectively integrate this multi-source information. Therefore, for the project recommendation scenario, this paper defines an open source weighted heterogeneous information network to represent the different entity types and link relationships in the GitHub open source collaborative development scenario, and effectively model the complex interaction among developers, projects and other entities. Using the weighted heterogeneous information network embedding method, extract and use the rich structural and semantic information in the weighted heterogeneous open source information network to learn the node representation of developers and projects, and fuse the personalized nonlinear fusion function into the matrix decomposition model for open source project recommendation. Finally, this paper makes a large number of comparative experiments based on the real GitHub open data set, and compares it with other project recommendation methods to verify the effectiveness of our proposed open source project recommendation model. At the same time, it also explores the impact of different metapaths on the effect of project recommendation. The experimental results show that the recommendation method based on heterogeneous information network can effectively improve the recommendation quality.