Use of Classification Algorithms for Semantic Web Services Discovery Martha Varguez-Moo, Francisco Moo-Mena and Victor Uc-Cetina Facultad de Matematicas, Universidad Autonoma de Yucatan, Merida, Mexico Email: msvarmoo@gmail.com, {mmena, uccetina}@uady.mx Abstract—Web services (WS) are used in different environments like enterprises, government and industry, providing tools for implementing complex distributed systems. Web service discovery allows a system to find services that meet the requirements of the users. One way to improve this type of discovery would consider not only the functional aspects of the required service, but also relevant aspects such as performance or availability to perform its functions. Moreover, this approach would allow a more efficient discovery process to obtain results closer to the user needs. In this paper we present an approach for Web service discovery through the use of machine learning algorithms for classification of Web services. For Web services matching our proposal takes into account quality of service (QoS) parameters, which include semantic information about each Web service. For semantic Web service specification we use the SAWSDL standard. Whereas the proposed UDDI standard was discontinued, among other reasons, due to limited syntactic Web service discovery, our approach brings important elements to the consolidation of semantic Web service discovery. Index Terms—Web services discovery, semantic Web services, non-functional properties, naïve bayes, support vector machines, adaboost I. INTRODUCTION Web Services (WS) are large distributed applications used under different environments such as enterprises, industry and government, in order to implement complex systems. Like any distributed system, Web services are susceptible to failures, for instance the failure of the server where a service resides or a failure in the channel of communication between a service provider and the requester. Another important problem taking place within Web services systems is that of finding the most appropriate service according to the needs of the customers [1]. The Universal Description Discovery and Integration (UDDI) registry was proposed twelve years ago for the publication of services. WS clients can access this registry in order to find the best service that meets their needs. Since UDDI uses syntactic information and does not use semantic information, the search for the most appropriate service is limited in the sense that clients cannot make requests asking for specific desirable properties such as quality of service parameters related to reliability, performance, security, response time, etc. These are some of the reasons that caused UDDI to be discontinued since 2006 [2]. With the disuse of UDDI, several proposals have emerged such as URBE [3]. URBE analyzes the structure and terms of the Web Service Description Language (WSDL) file that defines the Web service. This descriptor file is now employed to replace a failed WS. The combination of the theory of Semantic Web and Web services gives rise to what is known as Semantic Web Services. There are several approaches to add semantic information to services such as: OWL-S [4], WSDL-S [5] and WSMO [6]. Although it is possible to model ontologies to incorporate quality of service information, there are no commonly accepted techniques for finding and comparing the degree of similarity between Web services. In this paper we propose the use of machine learning algorithms as a better alternative to classify and match WS. Machine Learning has been used in the phase of composition of Web services using algorithms such as: ant colony and genetic algorithms [7][8]. Other approaches have used the Naive Bayes algorithm for discovering Web services together with WSMO [9][10]. After mentioning an overview of the service discovery problem, the Section II describes the related works. Section III gives details of the methodology here proposed. Finally, we present the experiments and conclusion. II. RELATED WORK Web service discovery consists in retrieving Web services from a registry, which meets the requirements of the request [11]. The approach in [11] specifies a classification of Web services discovery in which the authors define two categories: syntactic and semantic discovery. In the first category, they present techniques of Information Retrieval, based on schemes and links. In the second category semantic discovery is implemented by means of concepts and ontologies. The discovery based on concepts retrieves the semantic information from WSDL files to measure the similarity of parameters. The discovery based on ontologies make a model of semantic information using some approaches like OWL-S and WSMO. Manuscript received October 20, 2012; revised November 15, 2012; accepted November 25, 2012 1810 JOURNAL OF COMPUTERS, VOL. 8, NO. 7, JULY 2013 © 2013 ACADEMY PUBLISHER doi:10.4304/jcp.8.7.1810-1814