Semantic-based data mashups using hierarchical clustering and pattern analysis methods

Data mashups enable users to create new applications by combining Web APIs from several data sources. However, the existing data mashup framework requires some programming knowledge, hence it is not suitable for use by non-expert users. In this paper, we present hierarchical clustering and pattern analysis methods that build semantic ontologies automatically, and propose similarity searching algorithms that support the operation semantic matching and composable API discovery. These algorithms allow mashup developers to automate the discovery and composition of Web APIs eliminating the need for programmer involvement. We describe an experimental study on a collection of 168 REST APIs and 50 SOAP APIs. The experimental results show that our approach performs better in terms of both the rate of recall and precision performance compared with existing methods.