Web Services Clustering Approaches Based On Probability Topic Model And Neural Network

With the development of Web 2.0 technology,the number of Web services is becoming larger and larger.Clustering services together is an effective way to improve service discovery performance.This paper studys how to accurately cluster Web services by exploiting machine learing algorithms and neural networks algorithms.The main work of this paper can be divided into the following three parts:(1)This paper proposes a Web service clustering method based on Word2 Vec and LDA topic model.The method first uses the Wikipedia corpus as an extension source,it uses word2 vec algorithm to augment the content of the Web service description document,then uses the topic model to model the extended description document.It converts the short text into long text.It can realize service content theme expression more accurately and complete clustering according to the topic distribution matrix of the document.The experimental results show that compared with TFIDF-K,LDA,WTLDA and LDA-K methods,the proposed method has 419.74%,20.11%,15.60% and 27.80% improvement on F value.Therefore,the method of Web service clustering based on the expanded description document by using Word2 Vec and LDA topic model can improve the clustering effect of the Web service effectively.(2)This paper presents a Web service clustering method based on HDP and SOM neural network.This method combines the HDP and SOM neural network models.The HDP model can automatically determine the optimal number of topics without manual adjustment;the SOM algorithm can automatically determine the optimal number of clusters,and does not need to manually select the initial cluster center.Combining these two algorithms can reduce the cumbersomeness of the adjustment and the problem that the experimental results are affected by human factors.Compared with TFIDF-K,LDA,HDP,WT-LDA,HDP-K,HDP-S,ELDA-K,the method is different in F value.There are 486%,54.0%,35.8%,47.1%,39.0%,28.6%,9.4% improvement,so we conclude that the use of HDP and SOM neural network for service clustering effectively improves the clustering effect of Web services.(3)We designed and developed a visualization tool of service clustering.The tool implements the service clustering algorithm proposed in(1)and(2),and uses the pyecharts data visualization tool to display the clustering results of Web services,and exploits Flask technology to implement the service recommendation based on clustering results.