Elsevier

Information Sciences

Volume 530, August 2020, Pages 43-65
Information Sciences

A novel linguistic approach for multi-granular information fusion and decision-making using risk-based linguistic D numbers

https://doi.org/10.1016/j.ins.2020.04.006Get rights and content

Highlights

  • The interval-valued linguistic D numbers are developed

  • Linguistic D numbers possible risk scenarios are investigated.

  • A risk-based linguistic D numbers model is developed by taking into account the risks of beliefs and linguistic elements.

  • A multi-attribute decision-making (MADM) framework is extended using the proposed risk-based linguistic D numbers.

Abstract

The D numbers methodology is a new mathematical approach that has been developed to improve some constraints surrounding evidence theory by managing information uncertainty and incompleteness. Various studies have been conducted on developing D numbers. One of the main extensions of the D numbers methodology is linguistic D numbers, which employs linguistic terms as a set of evaluations of D numbers. In this study, linguistic D numbers are further extended to an interval-valued belief structure. Additionally, to consider the various risk scenarios of each linguistic D number, a risk-based linguistic D numbers model is presented, based on proposed interval-valued linguistic D numbers. The efficiency of the proposed model is investigated by applying it to numerical examples and considering a case study. The results show the robustness of the risk-based linguistic D numbers methodology while simultaneously applying various risk scenarios.

Introduction

Granular computing paradigms were developed to enable decision-makers to have more flexibility and accuracy in dealing with different kinds of uncertain information. One of the frameworks applied among granular computing models is fuzzy sets. Fuzzy information granules consist of fuzzy sets or linguistic terms that are sufficiently interpretable [1]. In many decision-making problems, the probabilities differ with regard to various linguistic information [2]. Therefore, the concept of probabilistic linguistic expressions was proposed to deal with such situations. Several research studies have aimed to combine, compute, and analyze these probabilistic-based models, such as probabilistic linguistic term sets (PLTSs) [3], distribution assessment [4], probabilistic hesitant fuzzy sets [5], probabilistic interval-valued intuitionistic hesitant fuzzy sets [6], evidential reasoning [7], and linguistic D numbers [8].

When adopting multi-granular fuzzy information, it is crucial to know how credible the data is. In many decision-making and engineering problems, especially in the case of forecasting and decision-making for future events, the data analyzed contains some percentage of errors and risk [9]. These errors are caused by several factors, such as predicted/unpredicted effective factors, which result in a situation where knowledge of the present cannot be generalized in a confident and reliable way to a future situation. These errors are recognized as risk factors [10]. Risk factors can be considered as effective factors that cause variations in  evaluated values and main values [11]. In order to model the reliability of multi-granular data obtained from different sources, a large number of studies have been conducted in which the reliabilities of different sources have been obtained by using dynamic reliability measures [12] or prior knowledge [13]. However, these methods are not suitable for risk modeling of data, since they only report on how reliable the source of the data is, and do not model variations in the data due to possible future-based risk factors. In addition, their reliability and credibility are commonly pessimistically estimated. Based on these approaches, unreliability likely worsens any data obtained, which is not necessarily the case in reality. Risk factors affect the information, i.e., ending up with an overestimation or underestimation [13], as compared to the actual value. Therefore, in order to account for the entire set of available data and reach to a robust solution, both overestimated (optimistic) and underestimated (pessimistic) outputs – which lead to local ignorance or interval uncertainty [14] – should be considered simultaneously. Thus, it is more rational to use interval-valued multi-granular models. A few studies have been proposed to model different risk evaluation scenarios, within which risk modeling has been conducted in two ways: (I) by considering the pessimistic-optimistic range of an evaluation due to risk factors [10]; and (II) by supposing a range indicating the degree of belief in an assessment [15]. However, there are cases where the risk factors pose risks to the evaluation and its belief simultaneously; it can be seen from the literature that no study has been carried out which considers the variations of assessment and belief due to risk factors simultaneously, especially when the data are in the form of linguistic terms.

D numbers were initially proposed to overcome some issues with the Dempster-Shafer theory (DST) [16]. There are strong hypotheses and hard constraints that limit the further development and application of DST in working with incomplete and fuzzy data. In DST, the elements are required to be mutually exclusive in the frame of discernment (FOD). Another constraint of the DST is that the sum of basic probabilities of a mass function is required to be equal to one, a condition that is called the completeness constraint [17]. Incompleteness of beliefs might occur due to a variety of reasons, including time pressure or lack of data or knowledge [18]. The linguistic D numbers approach can model incompleteness of beliefs, and does not involve the completeness constraint of the mass function, as compared to the classical DST [15]. Moreover, unlike other methods, it does not use the normalizing approach for dealing with incompleteness. Another characteristic of D numbers, as compared to the other multi-granular methods like distribution assessment, evidential reasoning, and DST, is that the aggregation set can be different from the basic set. This attribute gives more flexibility to the D numbers to model the variability of elements to capture risks and errors of the basic set itself. Therefore, different types of risk modeling can be considered, including pessimistic and optimistic modes of elements and beliefs.

The current study aims at developing a comprehensive framework based on linguistic D numbers, the so-called risk-based linguistic D numbers, to consider various risk scenarios by modeling intervals of pessimistic and optimistic modes of incomplete and probabilistic multi-granular linguistic information. For this purpose, firstly, the D numbers are extended to an interval-valued structure, then the pessimistic and optimistic modes of linguistic data are considered by proposing risk-based linguistic D numbers.

The paper is organized as follows: a literature survey on risk modeling in evaluations and studies on D numbers is provided in Section 2. The concepts associated with additive linguistics term sets, DST, and D numbers are briefly described in Section 3. The interval-valued D numbers model is explained in Section 4. Risk-based linguistic D numbers methodology and the multi-attribute decision-making (MADM) framework corresponding to the risk-based linguistic D numbers are discussed in Section 5. An example of failure modes and effects analysis using the proposed risk-based linguistic D numbers model is presented in Section 6. A summary and discussion on the model proposed in the present work and recommendations for future research are presented in Section 7.

Section snippets

Literature review

Various research studies have been performed in the areas of risk modeling of evaluation risks and D numbers methodology, which are discussed briefly as follows.

Preliminaries

This section briefly describes the additive linguistic terms, DST and D numbers, which are prerequisites for modeling the proposed interval-valued and risk-based linguistic D numbers.

Interval-valued linguistic D numbers

In this section, we develop the interval-valued linguistic D numbers model. Let us first present the main definitions and computation steps of the interval-valued linguistic D numbers (IVLD numbers) method. In this regard, by assuming that the IVLD numbers belief values are expressed as intervals, we then have the following:

Definition 9

For a discrete linguistic set Ω={bs1,bs2,,bsi,,bsn}, where bsi ∈ S and bsi ≠ bsj if i ≠ j, the IVLD numbers, represented by IDS({bsi}), are defined as follows:IDS({bsi})=[v

Risk-based linguistic D numbers and their application to MADM problems

This section aims at proposing a new concept called risk-based linguistic D numbers (RBLD numbers) for risk modeling of linguistic D numbers due to future-events risk factors, as well as a new MADM framework based on this novel concept. In this regard, the risk modeling of linguistic D numbers is investigated in two different schemes, i.e., risk modeling of the beliefs and risk modeling of linguistic elements. This section is structured as follows. Risk modeling of beliefs is investigated in

Case study

In order to evaluate the applicability of the proposed model, it was tested on the rotor blades of an industrial turbine, as detailed in this section. Being thin and subjected to excessive loading in tough working environments and rotating at high speed, the rotor blades of an industrial turbine are at a high risk of failure, making the so-called "failure mode and effects analysis" (FMEA) a required design solution for improving their safety and reliability [36]. Commonly used across various

Conclusion

In this study, an interval-valued linguistic D numbers method was developed for dealing with incomplete interval belief structures. Moreover, in order to avoid loss of output information and to improve the accuracy of the information, the risk modeling of linguistic D numbers and various possible scenarios were examined through a new concept called risk-based linguistic D numbers. In this new methodology, pessimistic-optimistic models were proposed for beliefs and the linguistic elements of the

Declarations of Competing Interest

none.

Compliance with ethical standards

This article does not contain any studies with human participants or animals performed by any of the authors.

CRediT authorship contribution statement

Hamidreza Seiti: Conceptualization, Methodology, Visualization, Writing - original draft, Writing - review & editing. Ashkan Hafezalkotob: Data curation, Writing - original draft, Writing - review & editing, Validation. Enrique Herrera-Viedma: Investigation, Methodology, Supervision.

Acknowledgments

This work was supported by Spanish Ministry of Science and Universities with FEDER funds in the National Spanish project TIN2016-75850-R and the support of the RUDN University Program 5-100 (Russian Federation).

References (50)

  • X. Deng et al.

    D-CFPR: D numbers extended consistent fuzzy preference relations

    Knowledge-Based Syst.

    (2015)
  • X. Zhou et al.

    Dependence assessment in human reliability analysis based on D numbers and AHP

    Nucl. Eng. Des.

    (2017)
  • X. Deng et al.

    Supplier selection using AHP methodology extended by D numbers

    Expert Syst. Appl.

    (2014)
  • X. Zhou et al.

    D-DEMATEL: a new method to identify critical success factors in emergency management

    Saf. Sci.

    (2017)
  • Y.-M. Wang et al.

    On the combination and normalization of interval-valued belief structures

    Inf. Sci. (Ny)

    (2007)
  • E.S. Lee et al.

    An interval dempster-shafer approach

    Comput. Math. with Appl.

    (1992)
  • T. Denœux

    Reasoning with imprecise belief structures

    Int. J. Approx. Reason.

    (1999)
  • P. Sevastianov et al.

    A framework for rule-base evidential reasoning in the interval setting applied to diagnosing type 2 diabetes

    Expert Syst. Appl.

    (2012)
  • Y. Song et al.

    Combination of interval-valued belief structures based on intuitionistic fuzzy set

    Knowledge-Based Syst.

    (2014)
  • H.-C. Liu et al.

    Failure mode and effects analysis using D numbers and grey relational projection method

    Expert Syst. Appl.

    (2014)
  • J. Yang et al.

    Risk evaluation in failure mode and effects analysis of aircraft turbine rotor blades using Dempster–Shafer evidence theory under uncertainty

    Eng. Fail. Anal.

    (2011)
  • A. Lindhe et al.

    Fault tree analysis for integrated and probabilistic risk analysis of drinking water systems

    Water Res.

    (2009)
  • Z. Xu et al.

    An overview of probabilistic-based expressions for qualitative decision-making: techniques, comparisons and developments

    Int. J. Mach. Learn. Cybern.

    (2019)
  • G. Zhang et al.

    Generalizing linguistic distributions in hesitant decision context

    Int. J. Comput. Intell. Syst.

    (2017)
  • J. Li et al.

    Multi-criteria outranking methods with hesitant probabilistic fuzzy sets

    Cognit. Comput.

    (2017)
  • Cited by (0)

    View full text