API-PROGRAM: An API Package Recommendation Model Based on the Graph Representation Learning Method

To combine multiple services together using technologies such as mashup to produce a composite service has become a popular practice. However, with the increasing number of services and the diversification of service types, how to select suitable services and ensure these service combinations meet the needs of users has become an increasingly challenging topic. At present, although there are many recommendation algorithms for service selection, the semantics of the composed Web services have not been sufficiently modeled. This paper proposes an API package recommendation model based on the graph representation learning method (API-PROGRAM) which uses the historical data to learn more comprehensive semantics of Web APIs, construct the composite features of Web API collaborations and then recommend Web API packages for new mashups. The experimental results show that, compared with the existing algorithms, API-PROGRAM achieves better performance.