A Quality-Based Web API Selection for Mashup Development Using Affinity Propagation

The rising interest in web APIs and mashups have led to a myriad of web APIs with similar functionality. Due to this reason, it is challenging to select relevant and quality web APIs for mashup developers, to compose quality and valuable mashups. On the other hand, clustering has proven to be one of the effective ways to select web APIs. However, methods, models and approaches that attempt to cluster web APIs for selection, by providing distinction between similar web APIs, focus either on their functionality or popularity and seldom consider quality of these web APIs. It is for this reason that this work proposes a method, based on topic modeling and clustering, to select quality web APIs for mashup development. First, we use Hierarchical Dirichlet Process (HDP) to identify a set of Web APIs that match a mashup developer's requirement, using the semantic distances between web API and developer's requirement topic distributions. Next, we use a black-box approach to analyze the quality of the subset of web APIs that match the mashup developer's requirement and employ Affinity Propagation (AP) clustering algorithm to cluster web APIs based on their quality. We perform experiments using dataset crawled from programmableweb.com and compare our results to other clustering-based selection methods.