Research Article
An API Recommendation Method Based on Beneficial Interaction
@INPROCEEDINGS{10.1007/978-3-031-24383-7_4, author={Siyuan Wang and Buqing Cao and Xiang Xie and Lulu Zhang and Guosheng Kang and Jianxun Liu}, title={An API Recommendation Method Based on Beneficial Interaction}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 18th EAI International Conference, CollaborateCom 2022, Hangzhou, China, October 15-16, 2022, Proceedings, Part I}, proceedings_a={COLLABORATECOM}, year={2023}, month={1}, keywords={Recommendation Beneficial feature interaction L0-Predictin GNN}, doi={10.1007/978-3-031-24383-7_4} }- Siyuan Wang
Buqing Cao
Xiang Xie
Lulu Zhang
Guosheng Kang
Jianxun Liu
Year: 2023
An API Recommendation Method Based on Beneficial Interaction
COLLABORATECOM
Springer
DOI: 10.1007/978-3-031-24383-7_4
Abstract
With the wide application of Mashup technology, it has become one of the hot and challenging problems in the field of service computing that how to recommend the API to developers to satisfy their Mashup requirements. The existing service recommendation methods based on Graph Neural Network (GNN) usually construct feature interaction graph by the interactions of service features, and regard it as the input of GNN to achieve service prediction and recommendation. In fact, there are some distinctions in the interactions between service features, and the importance of interactions is also different. To address this problem, this paper proposes an API recommendation method based on beneficial feature interaction, which can distinguish and extract beneficial feature interaction pairs from a large number of service feature interaction relationships. Firstly, feature extraction of Mashup requirements and API services is performed, and the correlation between API services is calculated based on the label and description document of the API services and used as a basis for recommending API services to Mashup requirements. Secondly, edge prediction component is used to extract beneficial feature pairs from input features of Mashup requirements and API services to generate beneficial feature interaction diagram between features. Thirdly, the beneficial feature interaction diagram is used as input of the graph neural network to predict and generate the API services set of recommendations for the Mashup requirements. Finally, the experiment on ProgrammableWeb dataset shows that the AUC of the proposed method has increased 20%, 24%, 27%, 13% and 21% respectively than that of AFM, NFM, DeepFM, FLEN and DCN, which means the proposed method improves the accuracy and quality of service recommendation.
