Context-aware Dynamic Service Matchmaking Shan Liu, Member, IEEE, Yichao Yang, Student Member, IEEE, Wenfeng Zheng, Member, IEEE, Xiaolu Li School of Automation University of Electronic Science and Technology of China Chengdu, China shanliu@uestc.edu.cn Abstract—In pervasive environments, the current service matchmaking is lack of context information with machine understandable and unable to deal with uncertainty of service properties, so it cannot achieve intelligent service discovery. This paper presents a fuzzy rough set theory based context-aware dynamic service matchmaking approach that composes an application through combining semantic information and context information. The proposed approach consists of formalized service description model with semantics and context attributes, and fuzzy rough set based service matchmaking. By describing the context attributes, the proposed approach is capable of composing context-aware application. Through a transformation technique, the incomplete information system is converted into a simpler system and then reducts are obtained from the transformed system based fuzzy rough set theory. Afterwards, the candidate service sets are selected by the function of the degree of keyword match and ranked through the function of the degree of service match. This paper describes the design and mechanism of the proposed approach. The proposed approach is expected to increase users’ satisfaction in pervasive environments. Keywords-component; Service matchmaking; Fuzzy rough sets; Context awareness I. INTRODUCTION In pervasive environments, it is important for the requesters to be able to dynamically find and utilize these services available on the Internet. Service discovery enables services to properly discover, configure, and communicate with each other. Over the past few years, several service discovery protocols, such as Service Location Protocol (SLP), Jini, Universal Plug and Play (UPnP), Intentional Naming System (INS), have been proposed to explore the service discovery issues in pervasive computing environments [1]. While these protocols provide efficient and fast searching schemes for service discovery, they do not adequately solve the problems that arise in dynamic matchmaking between service requests and advertisements. On the other hand, Universal Description, Discovery and Integration (UDDI) has been proposed and used for Web service publication and discovery [2]. However, the search mechanism supported by UDDI is limited to keyword matches. With the development of the semantic Web, services can be annotated with metadata for enhancement of service discovery. DARPA Agent Markup Language - Service (DAML-S) uses semantic information for discovery Web services [3]. Ontology Web Language for Service (OWL-S) is an Ontology Web Language (OWL) based ontology for encoding properties of Web services [4]. In addition, several service matchmaking algorithms based on UDDI, DAML-S, or OWL-S, have already been developed and implemented (e.g., [5, 6]). However, although we spend precious time actively looking for services and manually configuring them, there are two major issues which remain unaddressed in the existing algorithms. First, the existing algorithms often assume that service advertisements and service requests use consistent properties to describe relevant services. But for a pervasive environment such as a smart space with a large number of resources and users which have their own predefined properties to describe services, it is impossible that service advertisements and service requests use consistent properties to describe services. In order to provide higher usability to end users, a service matchmaking algorithm should be taken into consideration to deal with uncertainty of service properties. Second, as the number of services available in the network increases, it becomes more likely that different services may satisfy the same request. When different services satisfy the given request, a service matchmaking algorithm should rank the possible services for the user, similarly to modern web search engines, which rank search results based on certain criteria. In order to allow users to deal with uncertainty in service properties and to rank possible services for users, the authors of this paper propose a context-aware dynamic service matchmaking algorithm. In order to deal with uncertainty in service properties, the fuzzy rough set theory is used in the proposed algorithm to abstract the most important influence factors. In addition, in order to rank possible services for users, the proposed algorithm computes the fuzzy similarity based on the context information of users (e.g., locations and profiles) and services (e.g., device capabilities). The remainder of this paper is organized as follows. Section 2 introduces fuzzy rough set theory. Section 3 is a description of the proposed service matchmaking algorithm, including the formalized service model with context attributes, transformation of the incomplete information system, fuzzy rough set based attribute reduction and service matchmaker through the degree of keyword match and the degree of service match. Finally, Section 4 is the conclusion and future works. II. FUZZY ROUGH SET THEORY Rough set theory proposed by Pawlak, has been conceived as a tool to conceptualize, organize and analyze various types of data, in particular, to deal with inexact, uncertain or vague The work was supported by Guangdong Province and the Ministry of Education cooperation project of China under Grant 2011B090400352. 2012 IEEE 12th International Conference on Computer and Information Technology 978-0-7695-4858-6/12 $26.00 © 2012 IEEE DOI 10.1109/CIT.2012.202 985 2012 IEEE 12th International Conference on Computer and Information Technology 978-0-7695-4858-6/12 $26.00 © 2012 IEEE DOI 10.1109/CIT.2012.202 985 2012 IEEE 12th International Conference on Computer and Information Technology 978-0-7695-4858-6/12 $26.00 © 2012 IEEE DOI 10.1109/CIT.2012.202 985 2012 IEEE 12th International Conference on Computer and Information Technology 978-0-7695-4858-6/12 $26.00 © 2012 IEEE DOI 10.1109/CIT.2012.202 985