Web Service Classification With Class-dependent Feature Selection

With the development of cloud computing Web service technology becomes more and more important.The surge of Web services in the service registry(UDDI)brings a list of problem which need to be paid attention to.Web service classification is the first step after the service registered in UDDI.Now the classification method in UDDI is still manual assignment.This approach lack systematic management and the classification information is not applicable of the follow-up work such as Web service discovery and Web service combination.To solve the problems mentioned above,this paper presents an automatic classification mechanism for Web services.Firstly,the Web service document modeling method is proposed.Web service document describes the function of the service in detail.Key words extracted from the describe document serve as attributes for document modeling.TF-IDF is a popular method to get the vector space model of documents.The main idea of TF-IDF is to evaluate the importance of words in documents according to the word frequency.However,with the development of the Internet word frequency method is proved to be not powerful to express the characteristic of document.Using the semantic relations among terms is a new way to reflect the link of documents in specific areas.Based on the method of text mining,this paper utilizes TF-IDF and semantic similarity as an integrated method to establish the vector space model of the document.In this paper,a new method to calculate the information content of the concept in document is proposed to get the semantic similarity.This method uses the hierarchy structure of otology to calculate the information content.The position and different link of concept in the otology are all considered.The effectiveness of the proposed method is illustrated by comparing the correlation coefficient of this method with the results of artificial judgment.Secondly,in order to improve the classification accuracy,reduce data dimension and identify different discriminability of feature in each class,class-dependent feature selection method is used.Multi-objective genetic algorithm is used to select feature subspace.RBF is served as the classifier to evaluate selection scheme.Finally,the optimal feature subspace and RBF classifier is obtained.The test set is classified by using the feature subspace and classifier.In the experiment,the data of OWLS-TC 4 are used to validate the algorithm.The experimental results show that the proposed method can improve the accuracy of classification and improve the ability to identify the current class by the integrated similarity method.At the same time,the addition of feature selection improved the performance of the classifier on Web service.