Journal
MOBILE NETWORKS & APPLICATIONS
Volume 28, Issue 1, Pages 348-358Publisher
SPRINGER
DOI: 10.1007/s11036-023-02091-0
Keywords
Heterogeneous data; Intelligent semantic annotation; Graph neural network; Mobile services; IoT computing
Ask authors/readers for more resources
The rapid development of IoT computing has led to many problems, particularly in the management of massive mobile services. It is crucial to assign proper semantic annotations to these services in order for developers to find suitable services and providers to generate revenue. Existing approaches lack the utilization of the natural association between services, providers, and users. In this study, we propose a new model called GoT, which constructs a HIN for service data and fully leverages structural and semantic information to improve recommendation accuracy and alleviate the cold-start problem.
The rapid development of Internet-of-Things (IoT) computing leads to many problems, among which the management of massive mobile services has attracted much attention. When developers are looking for a service for IoT computing from mobile services, they typically try to discover the services according to annotations. If the mobile services are not assigned by proper annotations, it will be difficult for developers to find the suitable service. For providers, if the services that they provide cannot be used by developers, there will be no revenue. Therefore, it is a critical to assign proper semantic annotations to the mobile services. Existing approaches usually use the call records between services and developers to construct a score matrix, and compute the similarity between services and semantic annotations. However, these approaches do not leverage the natural association between services, providers and users. To make full use of the information inherent in services, we construct a heterogeneous information network (HIN) for service data, and propose a new model named GoT, which fully utilizes the structural and semantic information. GoT contains four components, which are the metapath construction, the intra-metapath fusion, the inter-metapath fusion, and the semantic annotation recommendation. We collected a real-world Web API dataset and performed adequate experiments. The experimental results show that our model produces superior recommendation accuracy and alleviates the cold-start problem.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available