A Multi-Scale Hypergraph-Based Approach for Third-Party Library Recommendation in Mobile App Development

In mobile app development, selecting the right third-party libraries (TPLs) is crucial to enhance functionality, improve code quality, and speed up the development process. However, recommending appropriate TPLs remains challenging due to the complexity of app-library interactions and the need to capture high-order relationships. Existing methods, such as collaborative filtering and graph-based approaches, often fail to adequately address these complexities. To address this challenge, we propose MsRec, a multiscale hypergraph neural network-based approach for TPL recommendation. MsRec uses hypergraphs to model interactions in groups of different sizes, leading to a detailed representation of the relationship between the application and the library. By modeling the strength, category, and functionality of interactions within each category, our framework improves the accuracy and diversity of recommendations. The multiscale hypergraph structure supports fine-grained relational reasoning, making it particularly effective in this context. Extensive experiments on real-world datasets demonstrate that MsRec outperforms current state-of-the-art methods, providing relevant and diverse TPL recommendations. Additionally, our model shows strong performance on various benchmarks, highlighting its ability to effectively handle complex app-library interactions.