An Intuitionistic Fuzzy Approach for Ranking Web Services under Evaluation Uncertainty George Kakarontzas Dept. of Computer Science and Engineering Technological Education Institute of Thessaly Larissa, Greece gkakaron@teilar.gr Vassilis C. Gerogiannis Dept. of Business Administration Technological Education Institute of Thessaly Larissa, Greece gerogian@teilar.gr Abstract— Many researchers have suggested fuzzy-based methods to derive rankings of services based on the fuzzy degree that each service satisfies a set of weighted quality attributes. Most of these methods assume a complete set of candidate services completely assessed. However, the candidate service set may include services which have not been fully assessed yet with respect to all quality attributes. Unassessed candidates introduce hesitation regarding the ranking of already evaluated services. This paper suggests Intuitionistic Fuzzy Sets (IFSs) to handle these sources of hesitation when assessing quality of services. IFS score functions are used to rank services with regard to each quality attribute. The final ranking can be derived by applying an objective method based on entropy weights for the quality attributes. Keywords—service evaluation; objective ranking; quality attributes; intuitionistic fuzzy sets I. INTRODUCTION & RELATED WORK The problem addressed in this paper is ranking of web services when assessment of their quality attributes is characterized by uncertainty. Consider, for example, a broker evaluating services. The broker may have evaluated a subset of candidate services (already present in the service registry) which provide a similar function. Thus, the evaluator’s assessment horizon includes a number of services, but not all of them have been fully evaluated on their quality attributes. Even if all candidate services have been assessed, a new service may appear in the evaluator’s horizon also providing the same function. Most objective ranking methods cannot handle these sources of hesitation, since they often assume a closed set of completely assessed services [1, 2]. To allow service quality assessments to be as reliable as possible, despite unassessed candidate services, we introduce an objective service ranking approach based on Intuitionistic Fuzzy Sets (IFSs) [3]. In particular, we describe an approach for mapping quantitative quality attribute assessments to IFSs and then we discuss how this mapping can be useful to indicate rankings of candidate services with regard to quality attributes. IFS score and accuracy functions are used to derive service rankings with regard to each attribute. We also present how we can derive the final ranking of services with respect to all quality attributes by applying an objective method [4] based on entropy weights. Service ranking techniques are broadly classified into objective, subjective or hybrid [5], according to the paradigm they follow to determine attribute weights. Subjective techniques are based on unquantifiable qualitative attributes. Objective methods are suitable for quantifiable attributes [6] and determine attribute weights directly from existing data and quality measurements, relying on quality attribute values provided by service providers or trusted service brokers [7]. These values can be stored in the UDDI (Universal Description, Discovery, and Integration) since the UDDI registry has the capability for storing quality information using the tModels feature [8]. Subjective techniques specify attribute weights based on preferences given by users, providers or domain experts. Hybrid techniques [5] are also proposed to determine synthetic weights for attributes by combining available objective information and experts’ preferences. Many techniques rely on fuzzy logic for ranking services based on the fuzzy degree that each alternative service satisfies weighted quality attributes [2, 5, 6, 9]. Fuzzy-based techniques handle uncertain ratings of quality attributes, imprecise preferences for attribute weights and vague relationships/trade-offs among quality attributes. In the literature there are also subjective ranking approaches which adopt fuzzy set generalizations, such as IFSs [10] and vague sets [11], to represent imprecise perceptions and hesitation of users and providers regarding quality attributes. We also adopt IFSs following an objective approach but with a different aim: to consider hesitation on service assessments due to the fact that all candidate services are not fully evaluated with respect to their all quality attributes. The remainder of the paper is structured as follows. In Sect. II we describe the proposed approach for mapping quality attribute assessments to IFSs. In Sect. III we give a simple example that exemplifies the mapping approach. In Section IV we demonstrate how service ranking is performed by examining an exemplar set of services. In Section V we conclude the paper. II. MAPPING QUALITY ASSESSMENTS TO IFSS An IFS A in a finite set X is defined as [3]:         (1) where  ,   ,       and       . Functions    and    denote, respectively, the degree of membership and the degree of non-membership of x to A.    is the hesitation degree of whether x belongs to . 2015 IEEE International Conference on Services Computing 978-1-4673-7281-7/15 $31.00 © 2015 IEEE DOI 10.1109/SCC.2015.105 742 2015 IEEE International Conference on Services Computing 978-1-4673-7281-7/15 $31.00 © 2015 IEEE DOI 10.1109/SCC.2015.105 742 2015 IEEE International Conference on Services Computing 978-1-4673-7281-7/15 $31.00 © 2015 IEEE DOI 10.1109/SCC.2015.105 742 2015 IEEE International Conference on Services Computing 978-1-4673-7281-7/15 $31.00 © 2015 IEEE DOI 10.1109/SCC.2015.105 742