Research Article
Improving Personalized Project Recommendation on GitHub Based on Deep Matrix Factorization
@INPROCEEDINGS{10.1007/978-3-030-92635-9_19, author={Huan Yang and Song Sun and Junhao Wen and Haini Cai and Muhammad Mateen}, title={Improving Personalized Project Recommendation on GitHub Based on Deep Matrix Factorization}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 17th EAI International Conference, CollaborateCom 2021, Virtual Event, October 16-18, 2021, Proceedings, Part I}, proceedings_a={COLLABORATECOM}, year={2022}, month={1}, keywords={GitHub Project recommendation Personalized recommendation User behavior Deep matrix factorization}, doi={10.1007/978-3-030-92635-9_19} }- Huan Yang
Song Sun
Junhao Wen
Haini Cai
Muhammad Mateen
Year: 2022
Improving Personalized Project Recommendation on GitHub Based on Deep Matrix Factorization
COLLABORATECOM
Springer
DOI: 10.1007/978-3-030-92635-9_19
Abstract
GitHub is a hosting platform for open-source software projects, where developers can share their open-source projects with others in the form of a repository. However, as the software projects hosted on the platform increase, it becomes difficult for developers to find software projects that meet their need or interest. Considering the practical importance of software project recommendations, we propose a recommendation method based on deep matrix factorization and apply it to GitHub, which is used to recommend personalized software projects in GitHub. With the use of deep neural network, we learn a low dimensional representation of users and projects from user-project matrix in a common space, in which we can capture the user’s latent behavior preference of each item, and automatically recommend the top N personalized software projects. The experiments on use-project data extracted from GitHub shows that the proposed recommendation method can recommend more accurate results compared with other three recommendation methods, i.e., UCF (user collaborative filtering), ICF (item collaborative filtering) and PPR (a personalized recommendation method).
