Research On The Mashup Service Clustering Based On Tag Recommendation

With the development of cloud computing and services, many kinds of network services appeared, which greatly promoted the construction of network applications and software systems for services. Traditional Web services generally based on SOAP protocol and described by WSDL documents, they are widely used in various fields of the Internet, but, on another hand, a lot of problems appear because of its disadvantages like complicated technical system and poor scalability. Light weight Mashup service based on R ESTful is composed by different Web API, users can develop software application to satisfy the individuality demand. However, compared to the traditional Web services, there is no formal description model in Mashup service, lead finding and discovering its service more difficult. Programmable Web is a popular online community that allows users to publish, annotate and sort Mashup, but, because everyone can join in this web to complete some operations, the accuracy and efficiency of it is poor. So, Web services discovering and mining become a research hotspot, using label information to cluster rationally and effectively to improve efficiency of Web service get more and more attention. In this paper, we did the following works based on tag recommendation and service clustering:(1) Proposed ainnovative Mashup services clustering method by fusing K-Means and Agnes(MSCA). First, expanding and sorting Tag labels in Mashup services; Second, compute intergraded similarity of Mashup services; then, using K-Means algorithm to cluster similarity matrix of Mashup services, once find out similar Mashup services, divided them into N atom clusters, and using Agnes algorithm clustering these N clusters, compared with traditional method, the effect and precision greatly improved.(2) Proposed a Mashup services clustering method based on LDA tag(MT-LDA), first, preprocessing the data, using LDA a s corpus to modeling, using Gibbs sampling method to infer, indirect calculation model parameters, finding out the relationship between different themes and words in a document to get the distribution of its themes, and compute the similarity between the text.Finally, the text similarity matrix to clustering Mashup services and experiments are carrieel out to assess the effect of clustering.Compared with another method, using LDA method to find out the implied topics in resource and service clustering with tag information can significantly improve the performance of the service clustering.