A Cloud API Personalized Recommendation Method Based on Multiple Attribute Features and Mashup Requirement Attention

Abstract

In current mashup-oriented cloud API recommendation systems, insufficient attention to personalized development requirements remains a common issue, particularly regarding developers’ needs for attributes such as functionality similarity and complementarity. This paper proposes a novel approach for personalized cloud API feature representation and recommendation. We construct a graph of the cloud API ecosystem with rich side information and design metapaths to capture and characterize various API features. To fully leverage information from intermediate nodes in the metapaths and emphasize the significance of different instances, we employ a translational distance model and graph neural network techniques to aggregate cloud API feature information. Furthermore, we introduce mashup requirement attention, a mechanism that customizes recommendations based on the specific needs of each mashup project, thereby enhancing the accuracy and personalization of API recommendations. Extensive experiments on real-world datasets demonstrate the effectiveness of the proposed method.

Keywords

• Attention mechanism
• cloud application programming interface
• mashup-oriented
• multiple attribute features
• personalized recommendation