Elsevier

Knowledge-Based Systems

Volume 77, March 2015, Pages 1-15
Knowledge-Based Systems

A new fuzzy hybrid technique for ranking real world Web services

https://doi.org/10.1016/j.knosys.2014.12.021Get rights and content

Abstract

We propose in this article a new fuzzy hybrid ranking technique, which is based on a linear combination of two new ranking techniques we devised: an objective Fuzzy Distance Correlation Ranking Technique (FDCRT) and a subjective Fuzzy Interval-based Ranking Technique (FSIRT). The objective technique leverages the distance correlation metric to derive weights of quality attributes directly from the available data. The subjective technique computes weights from opinions of domain experts, which are specified via two ingredients: intervals representing acceptable ranges of values for quality attributes and importance values of a quality attribute with respect to the other attributes. We show that the linear combination of these two techniques allows to overcome the shortcomings of objective and subjective techniques. Our experiments are performed on a dataset of real world Web services. The empirical results show that a tuning of the proposed linear combination gives better ranking results than Entropy and Fuzzy AHP separately and even than a linear combination of these two well-known techniques.

Introduction

Nowadays, Web services are considered one of the most successful and popular technologies thanks to their contribution to the deployment of legacy systems through the internet and their convenience for the development of loosely coupled applications. The drastic growing of the deployed Web services made the selection of best Web services an issue for the consumers. In fact, several Web services are similar in functionality, which requires a mechanism for ranking them. Such a mechanism can be based on non-functional quality attributes such as response time, reliability, availability, throughput, and security. Depending on the domain of application, quality attributes may have different weights. Such weights are paramount for ranking Web services. Several ranking techniques were proposed to determine attribute weights. They are classified in three major categories: objective, subjective, and hybrid techniques. Objective techniques [19], [20], [22], [24], [26], [37] determine weights directly from the data. Well-known techniques of this category are: principal element analysis, entropy, and multiple objective programming model. Subjective techniques [13], [15], [16], [23], [33] derive quality attribute weights from preference information given by domain experts. This second category includes well-known techniques such as Analytic Hierarchy Process (AHP), eigenvector technique, weighted least square technique, and Delphi technique. The last category is a hybrid combination of objective and subjective techniques [14], [17], [28], [30], [32], [34], [35], [38]. Algorithms using the hybrid techniques determine weights directly from the data as well as from the preference information given by domain experts or the users.

It is well-known that objective techniques do not take into consideration the opinions of experts. This leads sometimes to unrealistic or biased results for some domains of application since the automatic collection of data is done at a precise time period and for specific candidates. Besides, regarding Web services there are some attributes that cannot be accurately measured such as security, and flexibility. On the other hand, weights determined by subjective techniques reflect the subjective judgement of domain experts. However, analytical results or ranking of alternatives based on such weights can be influenced by the decision maker due to his/her lack of expertise.

To overcome shortcomings of both objective and subjective techniques, we propose a fuzzy hybrid ranking technique that is a linear combination of a new fuzzy objective technique called FDCRT and a new fuzzy subjective technique called FSIRT that we devised. Both techniques are based on fuzzy theory in order to deal with imprecise Quality Of Service (QoS) constraints and opinions of experts. The experiments are performed on a dataset of real world Web services. They show that the combination of these two techniques provides better ranking results than Entropy and Fuzzy AHP separately or combined together.

The remainder of this article is organized as follows. Section 2 explores the related work. Section 3 provides background knowledge about fuzzy logic. Section 4 presents the new fuzzy objective technique (FDCRT) we propose while Section 5 is dedicated to the presentation of the new fuzzy subjective one (FSIRT). A new hybrid technique that is a combination of objective and subjective techniques is proposed in Section 6. In Section 7, we present our experimental design and we provide our experimental results and analysis in Section 8. Finally, the conclusion and future work are given in Section 9.

Section snippets

Related work

We review in this section the related work regarding objective, subjective and hybrid techniques for ranking Web services with a special focus on those using Fuzzy logic.

Fuzzy theory

Fuzzy systems use possibility theory to handle uncertainty in their reasoning process [2]. The following traditional example clarifies the fuzzy theory: let us assume a glass filled with unknown liquid. In this case, “the glass is filled with water with the possibility of 70%” implies that 70% (volume wise) of this liquid is definitely water and 30% something else. In contrast, crisp systems handle uncertainty using probability theory. In crisp systems the above statement “the glass is filled

FDCRT: A new fuzzy distance correlation based ranking technique

In this section, we propose a new fuzzy objective technique for ranking Web services. Such ranking leverages the non-functional quality attributes of Web services. We provide in what follows a description of these attributes along with description of the data used for in the experiments.

FSIRT: A new fuzzy subjective interval-based ranking technique

From a weighting techniques perspective, a technique can be categorized based on the input source. Objective techniques include all techniques that derive weights for specific qualities based on a set of sample historical data of those qualities. Whereas, subjective techniques derive weights based on users/experts’ opinions. These opinions are not related to any dataset but to human judgment. In this section, we propose a new fuzzy subjective weighting technique, which we call Fuzzy Subjective

A new fuzzy hybrid ranking technique

Subjective ranking techniques have a major weakness in being dependent on the experts’ opinions, which may provide bad results when these opinions are not accurate due to the lack of experience of some experts. Henceforth, we resort to the combination of FDCRT and FSIRT to overcome the limitations of both objective and subjective techniques. This leads to the elaboration of a new hybrid ranking technique. Before computing the utility value of a Web service which is needed for ranking Web

Experimental design

For the assessment of our ranking results, we designed experiments to compare the performance of our approach to the aforementioned well-known objective (Entropy) and subjective technique (Fuzzy AHP) and to a linear combination of them. We provide hereafter a quick overview of Entropy and Fuzzy AHP, which are used in our comparative study.

Experimental results and analysis

To the best of our knowledge, there are so far no published experimental results on ranking real world Web services based on fuzzy logic. Our experiments are done on the QWS dataset [22], which is a dataset of real world Web services where nine quality attributes are evaluated for each Web service. The experiments are conducted for the purpose of comparing the performance of our approach to the aforementioned well-known objective (Entropy) and subjective technique (Fuzzy AHP) and to a linear

Conclusion and future work

We proposed in this article a new fuzzy hybrid ranking technique, which is based on a linear combination of an objective ranking technique (FDCRT) and a subjective ranking one (FSIRT). The objective technique leverages the distance correlation metric to derive weights of quality attributes. The subjective technique computes such weights from opinions of domain experts, which are specified via two ingredients: intervals representing acceptable ranges of values for quality attributes and

References (38)

  • T.W. Malure et al.

    Expert systems: the new challenge for managers

    Sloan Manage. Rev.

    (1996)
  • B. Kang, P. Compton, P. Preston, Simulated expert evaluation of multiple classification ripple down rules, in: 11th...
  • V. Leekwijck et al.

    Defuzzification: criteria and classification

    Fuzzy Sets Syst.

    (1999)
  • K. Jarvelin et al.

    Cumulated gain-based evaluation of IR techniques

    ACM Trans. Inform. Syst.

    (2002)
  • M.F. Chen et al.

    Fuzzy MCDM approach to select service provider

  • Sh. Ran

    A model for web services discovery with QoS

    ACM SIGecom Exchanges

    (2003)
  • M. Lin, J. Xie, H. Guo, H. Wang, Solving QoS-driven web service dynamic composition as fuzzy constraint satisfaction,...
  • L.-H. Vu et al.

    QoS-based service selection and ranking with trust and reputation management

  • P. Wang et al.

    A fuzzy model for selection of QoS-aware web services

  • Cited by (22)

    • A formal approach for matching and ranking trustworthy context-dependent services

      2018, Applied Soft Computing Journal
      Citation Excerpt :

      However, they are very limited to the functional aspects of services. Approaches such as [8–11], and [12] use non-functional features to improve service ranking. Mirmotalebi et al. [10] simplified the process of specifying non-functional requirements by building a user profile for non-functional preferences; in that way requesters are not required to specify preferences for every single query.

    • TAP: A personalized trust-aware QoS prediction approach for web service recommendation

      2017, Knowledge-Based Systems
      Citation Excerpt :

      Section 5 concludes the paper. QoS has been widely discussed in Service-Oriented Computing domain and Cloud Computing domain, covering the topic of service selection [15–18], service recommendation [19–21], service discovery [22], service composition [23,24], composite service reliability prediction [25,26], etc. Most existing QoS-based approaches assume that the QoS of services are pre-existing and easily obtained from a third-party organization or service providers with guaranteed accuracy.

    • Lattice-based ranking for service trust behaviors

      2016, Knowledge-Based Systems
      Citation Excerpt :

      The author uses trust-based ranking to favor or diminish candidates in service composition. Almulla et al. [2] combine both objective and subjective trust-based ranking in a fuzzy hybrid technique for web services. Objective techniques determine the weights of quality attributes directly from the existing data, whereas subjective techniques derive these weights from the opinion of domain experts.

    • Uncertain Service Skyline Queries Based on Cloud Model in Mobile Application*

      2021, 2021 11th International Conference on Information Science and Technology, ICIST 2021
    View all citing articles on Scopus
    View full text