Web services clustering using SOM based on kernel cosine similarity measure
Lei Chen, Geng Yang, Yingzhou Zhang, Zhengyu Chen · 2010
with the rapid growth of Web services and the need of quickly finding the right services, automatically clustering Web services becomes exceedingly important and challenging. The performance of Web services clustering relies closely on services representation, the similarity measure, and the clustering algorithm. This paper first presents a WordNet-VSM (W-VSM) model for Web services representation which not only enriches the conventional VSM feature vectors' semantic information but also reduce their dimension and sparsity. Then a set of kernel cosine similarity measures are proposed to well estimate the similarity of the Web services. Furthermore, an unsupervised SOM neural network algorithm based on aforementioned kernel cosine similarity measure (KCSOM) is presented to automatically cluster Web services. Finally, the preliminary experiments using real-world Web services demonstrate the feasibility of the proposed approach.