See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/384896740

Generalized picture fuzzy Frank aggregation operators and their applications

Article in Alexandria Engineering Journal · December 2024

DOI: 10.1016/j.aej.2024.09.081

CITATIONS READS

0 46

5 authors, including:

Jiulin Jin Wen Teng

Guiyang University Guizhou University of Finance and Economics
16 PUBLICATIONS 65 CITATIONS 27 PUBLICATIONS 15 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Jiulin Jin on 15 October 2024.

The user has requested enhancement of the downloaded file.

 Alexandria Engineering Journal 109 (2024) 726–739

Contents lists available at ScienceDirect

Alexandria Engineering Journal

journal homepage: www.elsevier.com/locate/aej

Original article

Generalized picture fuzzy Frank aggregation operators and their applications

Jiulin Jin a , Dragan Pamucar b,c,d ,∗, Shangshu Shi e , Hui Zhang f,g , Wen Teng h
a
School of Science, Guiyang University, Guiyang 550005, China
b
Department of Operations Research and Statistics, Faculty of Organizational Sciences, University of Belgrade, 11000 Belgrade, Serbia
c
Department of Industrial Engineering and Management, Yuan Ze University, Taoyuan City, Taiwan, ROC
d
Department of Mechanics and Mathematics, Western Caspian University, Baku, Azerbaijan
e Department of Basic Teaching, Guizhou Vocational College of Agriculture, Guiyang 551400, China
f School of Information, Guizhou University of Finance and Economics, Guiyang 550025, China
g Postdoctoral Scientific Research Station, ShijiHengtong Technology Co., Ltd., Guiyang 550014, China
h
School of Mathematics and Statistics, Guizhou University of Finance and Economics, Guiyang 550025, China

ARTICLE INFO ABSTRACT

Keywords: Picture fuzzy sets with four-dimensional features are widely used in decision-making as a mathematical tool
Frank aggregation operator because they can capture the uncertainty of data. However, the methods and techniques based on matrix
Picture fuzzy tensor theory are difficult to solve the decision problem involving high-dimensional data in a picture fuzzy setting.
High-dimensional data
Therefore, operators that can identify high-dimensional data in a picture fuzzy environment are proposed to
Multi-criteria decision-making
address this challenge. In this paper, firstly, by integrating the Frank operators into the picture fuzzy tensor, the
generalized picture fuzzy Frank weighted arithmetic (GPFFWA) and generalized picture fuzzy Frank weighted
geometric (GPFFWG) operators are defined. Their specific expressions are discussed, and the idempotency,
order-preservation, boundedness, and commutativity of the proposed operators are also given. Then, combining
the GPFFWA and GPFFWG operators, an algorithm is designed to solve the multi-criteria decision-making
problem with high-dimensional data features in the picture fuzzy environment. Finally, a numerical example
and related analysis demonstrate the effectiveness, superiority, and flexibility of the suggested technique. This
work provides new theoretical and methodological support for developing and practicing the decision-making
discipline.

1. Introduction of intuitionistic fuzzy sets [4] and picture fuzzy sets [5] have been
proposed one after another. From the perspective of mathematical
Multi-criteria decision-making (MCDM) is the decision-making pro- expression, picture fuzzy sets are a generalization of fuzzy sets and
cess based on multiple criteria when dealing with complex problems [1, intuitionistic fuzzy sets and have broader applicability. Through the
2]. In real life and business environments, we often encounter situations continuous deepening of MCDM research, scholars have found that
where we need to weigh different factors and objectives. In such the decision-making problems involved in many application fields can
instances, MCDM methods play a crucial role. Typically, the MCDM be expressed as MCDM problems in the picture fuzzy information
process relies on substantial amounts of data and information. During environment. In this environment, information often has more complex
the data collection, variable definition, measurement, and analysis ambiguity and uncertainty. Therefore, the MCDM theory and method
process, people’s cognition, experience, and prejudice can influence the for picture fuzzy information is one of the research hotspots in the field
outcome, resulting in incomplete and uncertain data and information. of fuzzy decision-making [6]. Recently, Simic et al. [7] proposed a
To characterize and analyze the uncertainty of information, Zadeh picture fuzzy combinative distance-based assessment method to solve
[3] first proposed the concept of fuzzy sets, which provides a ba-
the location problem of vehicle shredding facilities. Fan et al. [8]
sis for the research of fuzzy logic and fuzzy reasoning. With the
proposed a decision-making framework for solving the problem of
deepening of research, some scholars have found that it is difficult
green supplier selection in a picture fuzzy environment by integrat-
to accurately reflect the nature of uncertain things only by relying
ing the ‘‘Additive Ratio Assessment’’ method and ‘‘ViseKriterijumska
on fuzzy sets with a membership function. Therefore, the concepts

∗ Corresponding author at: Department of Operations Research and Statistics, Faculty of Organizational Sciences, University of Belgrade, 11000 Belgrade,
Serbia.
E-mail addresses: j.l.jin@hotmail.com (J. Jin), dpamucar@gmail.com (D. Pamucar), 704321545@qq.com (S. Shi), zhanghui@mail.gufe.edu.cn (H. Zhang),
tengwen@mail.gufe.edu.cn (W. Teng).

https://doi.org/10.1016/j.aej.2024.09.081
Received 19 September 2023; Received in revised form 17 August 2024; Accepted 22 September 2024
Available online 29 September 2024
1110-0168/© 2024 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/4.0/).
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

and Frank operator to represent and process complex data. The work
focuses on the basic operation of picture fuzzy tensor and explores
its application potential in decision-making. This will provide new
theoretical and methodological support for developing and practicing
related fields involving decision-making. The key contributions of this
paper are as follows:

• The generalized picture fuzzy Frank weighted arithmetic (GPF-

FWA) and generalized picture fuzzy Frank weighted geomet-
ric (GPFFWG) operators are proposed, and their computational
Fig. 1. The function of fuzzy tensor.
properties are shown.
• A decision-making framework is developed to address the MCDM
problem with high-dimensional data features using the suggested
Optimizacija I Kompromisno Resenje’’ method. Singh and Kumar [9] operators.
proposed a decision-making framework for selecting the best adsorbent • By using the constructed decision algorithm to assess the quality
by combining the proposed picture fuzzy knowledge measure with the of property services, the effectiveness of the proposed decision
classical decision method. Based on a new similarity, Rani et al. [10] algorithm is demonstrated.
proposed the measurement of alternatives and ranking according to • Through sensitivity and comparative analyses, the flexibility and
the compromise solution method for picture fuzzy information. Bani- superiority of the proposed decision algorithm are expressed.
Doumi et al. [11] proposed a hospital recommender based on picture
fuzzy sets by considering patients’ preferences for hospitals. Gündoğdu This paper has the following highlights:
et al. [12] proposed a hybrid decision model to evaluate the quality • The work is developed in a picture fuzzy environment, so our
of public transport services by combining picture fuzzy sets and linear proposed method can also deal with fuzzy information and in-
assignment models. Kara et al. [13] provided a multi-criteria group tuitionistic fuzzy information at the same time, which can better
decision support system in a picture fuzzy environment for effective deal with complex problems and incomplete information in the
recruitment. real world.
A challenging task in the MCDM process is to model high-dimensional • The work can integrate data of different dimensions and be ex-
data accurately since the attribute values of alternatives are often panded and adjusted according to the needs of specific problems.
affected by many factors, such as experts, time, and space in the actual • The Frank 𝑡-(co)norm with a parameter is incorporated into the
decision-making environment [14]. As the complexity of decision- picture fuzzy tensor for the first time to improve the flexibility
making systems increases, the dimension and scale of data are also in- and scalability of the method for dealing with complex MCDM
creasing [15]. However, matrix-based decision-making methods make problems.
it difficult to cope with this situation. Fig. 1 illustrates that fuzzy
tensors, a mathematical tool that extends traditional tensors and fuzzy The structure of this paper is shown below. Section 2 presents the
sets, offers advantages in handling high-dimensional fuzzy information. research related to the topic of the paper. Section 3 recalls the concepts
In recent years, significant progress has been made in the study of fuzzy of picture fuzzy sets, picture fuzzy Frank operators, and picture fuzzy
tensors. For instance, Chen and Lu studied the periodicity [16], con- tensors. In Section 4, we show the core results, that is, two generalized
vergence [17], and the decomposition theorem [18] of fuzzy tensors. picture fuzzy Frank aggregation operators and their computational
Deng et al. [15] studied the basic operations and properties of fuzzy properties. In Section 5, we use the proposed operator as a tool to
tensors. Chen [19,20] studied the decomposition theorem and three- design an algorithm to solve the MCDM problem with high-dimensional
valued cutting tensors of intuitionistic fuzzy tensors. Jin et al. [14] picture fuzzy data features. In Section 6, an illustrative example is given
studied the essential operation and decomposition theorem of picture to show the applicability of the proposed algorithm. The effectiveness,
fuzzy tensors. In addition, fuzzy tensors also have critical applications superiority, and flexibility of the suggested technique are demonstrated
in the fields of fuzzy classification [18] and pattern mining [21]. This by careful analysis. Finally, the conclusion is shown in Section 7.
overview shows that the study of fuzzy tensors can offer solutions to
practical problems. 2. Related work
The fuzzy information aggregation operator is an important concept
in fuzzy logic, which is used to combine the memberships of fuzzy sets In an MCDM, uncertainty is an essential factor [25], possibly due to
into a single fuzzy value. The earliest fuzzy information aggregation the incomplete acquisition of decision information, the fuzzy relation-
operator is the operation rule of fuzzy set proposed by Zadeh [3] ship between attributes, and the unclear evaluation criteria of decision-
in 1965. Subsequently, many scholars have also carried out research makers for alternatives. To effectively deal with the uncertainty in
on fuzzy information aggregation operators and have achieved many MCDM, researchers have established different models in various en-
beautiful results. The operation rules of these operators are based on the vironments, including fuzzy sets [3], intuitionistic fuzzy sets [4], and
𝑡-(co)norm. Among the many 𝑡-(co)norms, the Frank 𝑡-(co)norm [22] picture fuzzy sets [5]. These theories have been successfully used
has two advantages: First, the Frank 𝑡-(co)norm is the only 𝑡-(co)norm to measure uncertainty in data. In recent years, some scholars have
that satisfies the compatibility law [23]; Secondly, the Frank 𝑡-(co)norm devoted themselves to the development of decision theory in fuzzy
contains a parameter, which can be reduced to Lukasieicz 𝑡-(co)norm, environments to solve problems in various fields (e.g., [26–29]). The
Algebraic 𝑡-(co)norm, Einstein 𝑡-(co)norm and Hamacher 𝑡-(co)norm information aggregation operator is an important part of decision the-
under certain parameter conditions. Therefore, the Frank 𝑡-(co)norm ory, which is widely used because of its universality and flexibility. For
is popularly used to create different operators for addressing MCDM instance, the fuzzy Bonferroni operator is used to improve the urban
problems in complex environments [24]. traffic system [30]; an admissible ordered weighted averaging (AOWA)
It is considered that in the picture fuzzy environment, there are operator is used to evaluate sustainable development policies [31];
few methods to solve the MCDM problem with high-dimensional data the intuitionistic fuzzy rough Schweizer–Sklar aggregation operators
features. The purpose of this work is to propose two generalized pic- are used for investment risk management [32]; the intuitionistic fuzzy
ture fuzzy Frank operators to solve the MCDM problem with high- Aczel-Alsina Hamy mean operators are used to assess the construction
dimensional data features by combining the advantages of the tensor materials [33].

727
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Table 1
Characteristics of different operators.
Operator(s) Capability to Whether Capability to
integrate picture it contains integrate high-
fuzzy information parameter(s) dimensional data
The AOWA operator [31] × × ×
The operators in [24,30,32,33] × ✓ ×
The operators in [34,37–39,43] ✓ × ×
The operators in [35,36,40–42,44–49] ✓ ✓ ×
The GIVNFWG operator [50] × × ✓
The GIVIFWA and GIVIFWG operators [51] × × ✓
The GPFWIA and GPFWIG operators [14] ✓ × ✓
The proposed operators in this paper ✓ ✓ ✓

From the perspective of a mathematical expression, a picture fuzzy there are limited information aggregation operators to solve the MCDM
set contains a fuzzy set and an intuitionistic fuzzy set. Therefore, the problem with high-dimensional picture fuzzy data. Therefore, this work
picture fuzzy set is more suitable for the description of complex fuzzy considers the MCDM problem in the picture fuzzy tensor-based ar-
data in the decision-making environment. However, the information chitecture and proposes generalized picture fuzzy Frank operators. As
aggregation operator mentioned in the previous paragraph makes it shown in Table 1, the differences between this work and others are
difficult to solve the MCDM problem under the picture fuzzy envi- shown as follows:
ronment. In response to this challenge, more and more information
(1) The Frank 𝑡-(co)norm is used as the operation rule in this work.
aggregation operators have been proposed and studied. Wang [34]
Because it is a generalized version of many operations and
proposed a series of picture fuzzy weighted averaging operators and
the only 𝑡-(co)norm that satisfies the compatibility law, it has
picture fuzzy weighted geometric operators based on a probability
mathematical rigor and extensibility.
perspective in his doctoral thesis and applied them to MCDM problems.
(2) This work can deal with high-dimensional fuzzy data, while the
Garg [35] constructed some general forms of picture fuzzy information
methods and techniques based on matrix theory in [24,30–49]
aggregation operators and studied their idempotence, monotonicity,
do not have this feature.
boundedness, transformation invariance, and homogeneity. Consider-
(3) This work can deal with picture fuzzy data, while the fuzzy
ing the uncertainty of medical diagnosis and the correlation between
tensor-based operators in [50,51] lack this capability.
symptoms, Zhang et al. [36] proposed the picture fuzzy point Choquet-
(4) The operation rule used in this work contains a parameter, which
integral aggregation operators and successfully applied them to assist
is more suitable for a wider range of situations. Although the
the hierarchical medical system. In addition, some other picture fuzzy
proposed operators in [14,50,51] can handle high-dimensional
information aggregation operators (e.g., the picture fuzzy weighted
fuzzy data, they do not contain parameters and cannot express
interaction geometric operator [37], the picture fuzzy weighted inter-
the preference behavior of decision-makers.
action averaging operator [38], picture fuzzy power Choquet ordered
geometric operators [39], picture fuzzy Choquet mean operators [40], 3. Preliminaries
picture fuzzy Frank aggregation operators [41], picture fuzzy inter-
actional partitioned Heronian mean aggregation operators [42], pic- This section reviews some related notions.
ture fuzzy Aczel–Alsina average aggregation operators [43], picture
fuzzy Schweizer–Sklar aggregation operators [44], the picture fuzzy Definition 3.1 ([5]). Let 𝑈 be a non-empty universal set. Then, a
Einstein hybrid averaging aggregation operator [45], picture fuzzy picture fuzzy set 𝐴 on 𝑈 is defined by 𝐴 = {(𝑧, 𝜇𝐴 (𝑧), 𝜂𝐴 (𝑧), 𝜈𝐴 (𝑧)) ∶ 𝑧 ∈
point operators [46], picture fuzzy interactional Bonferroni mean op- 𝑈 }, where the functions 𝜇𝐴 ∶ 𝑈 → [0, 1], 𝜂𝐴 ∶ 𝑈 → [0, 1] and 𝜈𝐴 ∶ 𝑈 →
erators [47], picture fuzzy Aczel–Alsina power Bonferroni aggrega- [0, 1] are called the positive membership degree, neutral membership
tion operators [48], and picture fuzzy power partitioned Hamy mean degree and negative membership degree of 𝐴, respectively, with the
operators [49]) are proposed to deal with specific problems. condition (∀𝑧 ∈ 𝑈 ) 0 ≤ 𝜇𝐴 (𝑧) + 𝜂𝐴 (𝑧) + 𝜈𝐴 (𝑧) ≤ 1. Especially if 𝑈 has
In complex decision-making, the evaluation value of an alternative only one element, then the picture fuzzy set 𝐴 is regarded as a picture
is usually affected by many factors. Currently, the matrix-based in- fuzzy number. For convenience, a picture fuzzy number 𝜌 is denoted
formation aggregation operator struggles to handle MCDM problems by (𝜇𝜌 , 𝜂𝜌 , 𝜈𝜌 ).
with high-dimensional fuzzy data features. Tensor is a generalized
matrix form that displays a multi-way array term, which is the most Definition 3.2 ([52]). Let P = {(𝜇, 𝜂, 𝜈) ∈ [0, 1]3 ∶ 𝜇 + 𝜂 + 𝜈 ∈ [0, 1]}.
appropriate and practical way to represent high-dimensional data with- Then the partial order ≤P on P is defined as follows: (∀𝜌, 𝜎 ∈ P)
out losing any information [50]. Therefore, scholars have developed 𝜎 ≤P 𝜌 ⟺ (𝜇𝜎 < 𝜇𝜌 and 𝜈𝜎 ≥ 𝜈𝜌 ) or (𝜇𝜎 = 𝜇𝜌 and 𝜈𝜎 > 𝜈𝜌 ) or (𝜇𝜎 =
some information aggregation operators through fuzzy tensors to assist 𝜇𝜌 , 𝜈𝜎 = 𝜈𝜌 and 𝜂𝜎 ≤ 𝜂𝜌 ).
in solving the MCDM problem with high-dimensional data features.
Deng et al. [51] proposed the generalized interval-valued intuitionistic
Remark 1 ([52]). Let 𝜌, 𝜎 ∈ P and let
fuzzy weighted averaging (GIVIFWA) and generalized interval-valued
intuitionistic fuzzy weighted geometric (GIVIFWG) operators and their ⎧
⎪ 𝜌, if 𝜌 ≤P 𝜎,
basic properties, and used them to solve the ranking problem of agricul- ⎪ 𝜎, if 𝜎 ≤P 𝜌,
tural ecoregions in Hubei Province, China. Jin et al. [14] developed the 𝜌 ∧ 𝜎 =⎨
⎪ ( min{𝜇𝜌 , 𝜇𝜎 }, 1 − min{𝜇𝜌 , 𝜇𝜎 }
generalized picture fuzzy interactive weighted averaging (GPFIWA) and ⎪ − max{𝜈𝜌 , 𝜈𝜎 }, max{𝜈𝜌 , 𝜈𝜎 }), otherwise.
generalized picture fuzzy interactive weighted geometric (GPFIWG) ⎩
operators and their basic properties, and applied them to emergency ⎧ 𝜎, if 𝜌 ≤P 𝜎,
decision management. Singh [50] introduced a generalized interval- ⎪
𝜌 ∨ 𝜎 =⎨ 𝜌, if 𝜎 ≤P 𝜌,
valued neutrosophic fuzzy weighted geometric operator and applied ⎪ (max{𝜇𝜌 , 𝜇𝜎 }, 0, min{𝜈𝜌 , 𝜈𝜎 }), otherwise.
it to select the most effective recycled fiber-based paper mill as a 3rd ⎩
Party Reverse Logistics partner for the packaging industry. Denote 1P = (1, 0, 0) and 0P = (0, 0, 1). Then 1P and 0P are the top and
From the list of references shown above, it can be noted that the bottom elements of the poset (P, ≤P ), respectively. In addition, (P, ≤P )
information aggregation operator plays a vital role in MCDM. However, is a complete lattice.

728
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Definition 3.3 ([41]). Let 𝑘 and 𝑟 are two real numbers such that 𝑘 > 0 the set of all positive integers are denoted by R and N+ , respectively.
and 𝑟 > 1. Then Frank operators on P are defined as follows: (∀𝜌, 𝜎 ∈ P) The symbols 𝜇𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜂𝑖1 𝑖2 ⋯𝑖𝑚 and 𝜈𝑖1 𝑖2 ⋯𝑖𝑚 represent 𝜇𝑎𝑖 𝑖 ⋯𝑖 , 𝜂𝑎𝑖 𝑖 ⋯𝑖
1 2 𝑚 1 2 𝑚
{ } and 𝜈𝑎𝑖 𝑖 ⋯𝑖 , respectively.
⎛ (𝑟1−𝜇𝜌 − 1)(𝑟1−𝜇𝜎 − 1) ⎞ 1 2 𝑚
⎜1− log𝑟 1 + 𝑟−1
,⎟
⎜ { } ⎟ 4.1. Generalized picture fuzzy Frank weighted arithmetic operator
⎜ 𝜂 𝜂
(𝑟 𝜌 − 1)(𝑟 𝜎 − 1) ⎟
𝜌 ⊕ 𝜎 = ⎜ log𝑟 1 + , ⎟.
⎜ 𝑟−1 ⎟ Definition 4.1. Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 ),
{ }
⎜ (𝑟𝜈𝜌 − 1)(𝑟𝜈𝜎 − 1) ⎟ and let 𝑊𝑗 = (𝑤𝑗1 , 𝑤𝑗2 , … , 𝑤𝑗𝑛𝑗 )⊤ ∈ R𝑛𝑗 (𝑗 = 2, … , 𝑚). Then, the
⎜ log𝑟 1 + ⎟
⎝ 𝑟−1 ⎠ GPFFWA operator is defined as follows:
{ 𝜇𝜌 𝜇𝜎 − 1)
}
⎛ (𝑟 − 1)(𝑟 ⎞ GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) = [𝑐𝑖1 ]𝑛1 ×1 ,
⎜ log𝑟 1 + 𝑟−1
, ⎟
⎜ { }⎟
⎜ (𝑟 𝜌 − 1)(𝑟 𝜎 − 1) ⎟
1−𝜂 1−𝜂 where
𝜌 ⊗ 𝜎 = ⎜1− log𝑟 1 + ,⎟ . 𝑛2 𝑛𝑚
𝑟−1 ⨁ ⨁
⎜ { }⎟ 𝑐𝑖1 = ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 (4)
⎜ (𝑟1−𝜈𝜌 − 1)(𝑟1−𝜈𝜎 − 1) ⎟
⎜1− log𝑟 1 + ⎟ 𝑖2 =1 𝑖𝑚 =1
⎝ 𝑟−1 ⎠
{ } Now, we give the expression of the GPFFWA operator as follows.
⎛ (𝑟1−𝜇𝜌 − 1)𝑘 ⎞
⎜1− log𝑟 1 + (𝑟 − 1)𝑘−1 ,⎟
⎜ { } ⎟ Theorem 4.2. Let 𝑟 > 1. Then Eq. (4) can be expressed as follows:
⎜ (𝑟𝜂𝜌 − 1)𝑘 ⎟ ∏𝑛2 ∏𝑛 ( 1−𝜇 )𝑤 ⋯𝑤
𝑘𝜌 = ⎜ log𝑟 1 + , ⎟. ⎛ { ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎞
𝑘−1
⎜ (𝑟 − 1) ⎜1− log 1 + 𝑖2 =1 ,⎟
} ⎟
𝑚
{ ⎜ 𝑟 ∑𝑛2 ∑𝑛 ⎟
⎜ (𝑟𝜈𝜌 − 1)𝑘 ⎟ ⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1
⎜ log𝑟 1 + ⎟ ⎜ (𝑟 − 1) 𝑖2 =1 𝑚 2
⎟
⎝ (𝑟 − 1)𝑘−1 ⎠ ⎜ { ∏ ∏ ( ) ⎟
𝑚 }
𝑛2 𝑛 𝑚 𝜂 𝑖1 𝑖2 ⋯𝑖𝑚 𝑤 2𝑖 ⋯𝑤 𝑚𝑖
{ 𝜇𝜌 𝑘
} ⎜ 𝑖2 =1
⋯ 𝑖𝑚 =1
𝑟 − 1 2
⎟
⎛ (𝑟 − 1) ⎞ 𝑐𝑖1 = ⎜ log𝑟 1 + ∑𝑛2 ∑𝑛𝑚 , ⎟. (5)
⎜ log𝑟 1 + (𝑟 − 1)𝑘−1 , ⎟ ⎜ ⋯ 𝑤 2𝑖 ⋯𝑤 𝑚𝑖𝑚
−1 ⎟
⎜ (𝑟 − 1) 𝑖2 =1 𝑖𝑚 =1 2
{ }⎟ ⎜ ∏𝑛2 ∏𝑛 ( 𝜈 )𝑤 ⋯𝑤 ⎟
⎜ (𝑟1−𝜂𝜌 − 1)𝑘 ⎟ ⎜ { ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎟
𝜌𝑘 = ⎜1− log𝑟 1 + ,⎟ . ⎜ log𝑟 1 + 𝑖2 =1 𝑚 ⎟
⎜ (𝑟 − 1)𝑘−1 ⎟ ∑𝑛2 ∑𝑛
{ } ⎜ ⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1 ⎟
⎜ (𝑟1−𝜈𝜌 − 1)𝑘 ⎟ ⎝ (𝑟 − 1) 𝑖2 =1 𝑚 2 ⎠
⎜1− log𝑟 1 + ⎟
⎝ (𝑟 − 1)𝑘−1 ⎠
Proof. Let 𝑀 = {2, … , 𝑚}. Then, we consider the following cases:
Definition 3.4 ([53]). Let 𝜌 ∈ P, and the score and accuracy functions (I) There exists 𝑗(1) ∈ 𝑀 such that 𝑛𝑙 = 1 for each 𝑙 ∈ 𝑀 − {𝑗(1)}.
of 𝜌 are defined by Then
𝑛
⨁
1 ⨁
𝑗(1)
⨁
1
𝑆(𝜌) =𝜇𝜌 − 𝜈𝜌 , (1) 𝑐𝑖1 = ⋯ ⋯ 𝑎𝑖2 ⋯𝑖𝑗(1) ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚𝑖𝑚
𝑖2 =1 𝑖𝑗(1) =1 𝑖𝑚 =1
𝐻(𝜌) =𝜇𝜌 + 𝜂𝜌 + 𝜈𝜌 . (2)
𝑛
⨁
𝑗(1)

Then, for any two picture fuzzy numbers 𝜌 and 𝜎, we have = 𝑎1⋯𝑖𝑗(1) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚1 .
𝑖𝑗(1) =1
(1) if 𝑆(𝜌) > 𝑆(𝜎) then 𝜌 is superior to 𝜎.
For 𝑛𝑗(1) ∈ N+ , we use mathematical induction to prove the following
(2) if 𝑆(𝜌) < 𝑆(𝜎) then 𝜌 is inferior to 𝜎.
equation:
(3) if 𝑆(𝜌) = 𝑆(𝜎) then
𝑛
⨁
𝑗(1)
∙ if 𝐻(𝜌) > 𝐻(𝜎) then 𝜌 is superior to 𝜎. 𝑎1⋯𝑖𝑗(1) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚1
∙ if 𝐻(𝜌) < 𝐻(𝜎) then 𝜌 is inferior to 𝜎. 𝑖𝑗(1) =1
∙ if 𝐻(𝜌) = 𝐻(𝜎) then 𝜌 is similar to 𝜎. { ∏𝑛𝑗(1) ( 1−𝜇1⋯𝑖𝑗(1) ⋯1 )𝑤 ⋯𝑤 ⋯𝑤
⎛ 𝑟 − 1 21 𝑗(1)𝑖𝑗(1) 𝑚1 } ⎞
⎜1− log 1 + 𝑖𝑗(1) =1 ,⎟
Definition 3.5 ([14,54]). Let 𝑛1 , 𝑛2 , …, 𝑛𝑚 are 𝑚 positive integers. Then ⎜ 𝑟 ∑𝑛𝑗(1)
𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎟
⎜ (𝑟 − 1) 𝑗(1)𝑖 =1 𝑗(1) ⎟
an 𝑚-order picture fuzzy tensor  is defined as [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 , ⎜ ∏𝑛𝑗(1) ( 𝜂1⋯𝑖𝑗(1) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 ⎟
⎜ { 𝑟 − 1 𝑗(1) } ⎟
where 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ∈ P for any 1 ≤ 𝑖𝑗 ≤ 𝑛𝑗 (𝑗 = 1, 2, … , 𝑚). In addition, 𝑖𝑗(1) =1
= ⎜ log𝑟 1 + , ⎟. (6)
the symbol 𝑖⃖⃖⃗𝑗 (1 ≤ 𝑗 ≤ 𝑚) is called the 𝑗-th direction of . If  is ⎜ ∑𝑛𝑗(1) ⎟
𝑤 ⋯𝑤𝑗(1)𝑖
𝑖𝑗(1) =1 21
⋯𝑤𝑚1 −1
expanded in the direction 𝑖⃖⃖⃗𝑗 , then  can be decomposed into (𝑚 − 1)- ⎜ (𝑟 − 1) 𝑗(1) ⎟
⎜ ∏𝑛𝑗(1) ( 𝜈1⋯𝑖 ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 ⎟
order picture fuzzy tensors ∶⋯∶1∶⋯∶ , ∶⋯∶2∶⋯∶ , …, ∶⋯∶𝑛𝑗 ∶⋯∶ . Here, ⎜ { 𝑟 𝑗(1) − 1 𝑗(1) } ⎟
𝑖𝑗(1) =1
 = [∶⋯∶1∶⋯∶ , ∶⋯∶2∶⋯∶ , … , ∶⋯∶𝑛𝑗 ∶⋯∶ ]. ⎜ log𝑟 1 + ∑𝑛𝑗(1)
⎟
⎜ 𝑤 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎟
⎝ (𝑟 − 1) 𝑖𝑗(1) =1 21 𝑗(1) ⎠
For convenience, the set of all 𝑚-order picture fuzzy tensors is
denoted by 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 ). When 𝑛𝑗(1) = 1, from Definition 3.3 it follows that
𝑛𝑗(1)
⨁
Definition 3.6 ([14,54]). The partial order ⪯ on 𝐓P (𝑚, 𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ) is 𝑎1⋯𝑖𝑗(1) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚1
defined as follows: (∀ = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ,  = [𝑏𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝑖𝑗(1) =1
𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 )) =𝑎1⋯1 𝑤21 ⋯ 𝑤𝑗(1)1 ⋯ 𝑤𝑚1
{ ( 1−𝜇 )𝑤 ⋯𝑤 ⋯𝑤 }
 ⪯  ⟺ (∀𝑖1 , 𝑖2 , … , 𝑖𝑚 ) 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ≤P 𝑏𝑖1 𝑖2 ⋯𝑖𝑚 . (3) ⎛ 𝑟 1⋯1 − 1 21 𝑗(1)1 𝑚1 ⎞
⎜1− log𝑟 1 + ,⎟
⎜ (𝑟 − 1)𝑤21 ⋯𝑤𝑗(1)1 ⋯𝑤𝑚1 −1 ⎟
4. Generalized picture fuzzy Frank aggregation operators ⎜ { ( ) 𝑤 ⋯𝑤 ⋯𝑤 } ⎟
𝑟𝜂1⋯1 − 1 21 𝑗(1)1 𝑚1
= ⎜ log𝑟 1 + , ⎟.
⎜ (𝑟 − 1) 𝑤21 ⋯𝑤𝑗(1)1 ⋯𝑤𝑚1 −1 ⎟
In this section, we develop two kinds of important aggregation ⎜ ( 𝜈 )𝑤21 ⋯𝑤𝑗(1)1 ⋯𝑤𝑚1 } ⎟
⎜ { ⎟
operators of picture fuzzy tensors and discuss their computational prop- 𝑟 1⋯1 − 1
erties. For the convenience of writing, the set of all real numbers and ⎜ log𝑟 1 + ⎟
⎝ (𝑟 − 1)𝑤21 ⋯𝑤𝑗(1)1 ⋯𝑤𝑚1 −1 ⎠

729
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

{ ∏𝑛𝑗(2) ( 1−𝜇1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1

∏𝑛𝑗(1) )𝑤 ⋯𝑤
− 1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 } ⎞
⋯𝑤 ⋯𝑤
⎛ 𝑟
and so Eq. (6) holds for 𝑛𝑗(1) = 1. ⎜1− log 1 + 𝑖𝑗(1) =1 𝑖𝑗(2) =1
∑𝑛𝑗(1) ∑𝑛𝑗(2) ,⎟
Let us assume that Eq. (6) holds for 𝑛𝑗(1) = 𝑘1 , that is, ⎜ 𝑟
𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
⎜ (𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎟
⎜ ∏𝑛𝑗(1) ∏𝑛𝑗(2) ( 𝜂1⋯𝑖 ⋯𝑖 ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1
{ 𝑟 𝑗(1) 𝑗(2) − 1 𝑗(1) 𝑗(2) } ⎟
𝑘1 ⎜ 𝑖𝑗(1) =1 𝑖𝑗(2) =1 ⎟
⨁ = ⎜ log𝑟 1 + ∑𝑛𝑗(1) ∑𝑛𝑗(2) , ⎟. (7)
𝑎1⋯𝑖𝑗(1) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚1 ⎜ (𝑟 − 1) 𝑖𝑗(1) =1
𝑤 ⋯𝑤𝑗(1)𝑖𝑗(1) ⋯𝑤𝑗(2)𝑖𝑗(2) ⋯𝑤𝑚1 −1
𝑖𝑗(2) =1 21 ⎟
𝑖𝑗(1) =1 ⎜ ∏𝑛𝑗(1) ∏𝑛𝑗(2) ( 𝜈1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1 ⎟
⎜ { 𝑟 − 1 𝑗(1) 𝑗(2) } ⎟
∏𝑘1 ( 1−𝜇1⋯𝑖 ⋯1 )𝑤 ⋯𝑤 ⋯𝑤 ⎜ log 1 + 𝑖𝑗(1) =1 𝑖𝑗(2) =1𝑛 ⎟
⎛ { 𝑟 𝑗(1) − 1 21 𝑗(1)𝑖𝑗(1) 𝑚1 } ⎞ ⎜ 𝑟 ∑ 𝑗(1) ∑𝑛𝑗(2)
𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
𝑖𝑗(1) =1
⎜1− log 1 + ,⎟ ⎝ (𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎠
⎜ 𝑟 ∑𝑘1 ⎟
𝑤 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1
⎜ (𝑟 − 1) 𝑖𝑗(1) =1 21 𝑗(1) ⎟ When 𝑛𝑗(1) = 𝑛𝑗(2) = 1, Eq. (7) is clearly established. Suppose
⎜ ∏𝑘1 ( 𝜂1⋯𝑖 ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 ⎟ that Eq. (7) holds for 𝑛𝑗(1) = 𝑘1 and 𝑛𝑗(2) = 𝑘2 , that is,
⎜ { 𝑟 𝑗(1) −1 𝑗(1) } ⎟
𝑖𝑗(1) =1
= ⎜ log𝑟 1 + ∑ , ⎟. ⨁ ⨁
⎜ 𝑘 1 𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎟ 𝑘1 𝑘2

⎜ (𝑟 − 1) 𝑖𝑗(1) =1 𝑗(1) ⎟ 𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1

⎜ ∏ ( 𝜈 )𝑤 ⋯𝑤 ⋯𝑤 ⎟ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
⎜ { 𝑘1
𝑟 1⋯𝑖𝑗(1) ⋯1
−1 21 𝑗(1)𝑖𝑗(1) 𝑚1 }
⎟ { ∏𝑘1 ∏𝑘2 ( 1−𝜇1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤 ⋯𝑤
− 1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 } ⎞
⋯𝑤 ⋯𝑤
𝑖𝑗(1) =1 ⎛ 𝑟
⎜ log𝑟 1 + ⎟ ⎜1− log𝑟 1 +
𝑖𝑗(1) =1 𝑖𝑗(2) =1
,⎟
⎜ ∑ 𝑘1
⎟ ∑𝑘1 ∑ 𝑘2
𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎜ 𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
⎝ (𝑟 − 1) 𝑖𝑗(1) =1 𝑗(1) ⎠ ⎜ (𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎟
⎜ { ∏𝑘1 ∏𝑘2 ( 𝜂1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1
} ⎟
⎜ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
𝑟 −1 𝑗(1) 𝑗(2)
⎟
Now, for 𝑛𝑗(1) = 𝑘1 + 1, we have = ⎜ log𝑟 1 + ∑𝑘1 ∑𝑘2 , ⎟.
⎜ (𝑟 − 1) 𝑖𝑗(1) =1
𝑤 ⋯𝑤𝑗(1)𝑖𝑗(1) ⋯𝑤𝑗(2)𝑖𝑗(2) ⋯𝑤𝑚1 −1
𝑖𝑗(2) =1 21 ⎟
⎜ ∏𝑘1 ∏𝑘2 ( 𝜈1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1 ⎟
⨁
𝑘1 +1
⎜ { 𝑟 −1 𝑗(1) 𝑗(2) } ⎟
𝑎1⋯𝑖𝑗(1) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚1 ⎜ log 1 + 𝑖𝑗(1) =1 𝑖𝑗(2) =1 ⎟
⎜ 𝑟 ∑𝑘1 ∑𝑘2 ⎟
𝑖𝑗(1) =1 𝑤21 ⋯𝑤𝑗(1)𝑖𝑗(1) ⋯𝑤𝑗(2)𝑖𝑗(2) ⋯𝑤𝑚1 −1
⎝ (𝑟 − 1) 𝑗(1)𝑖 =1 𝑖𝑗(2) =1 ⎠
⨁
𝑘1
⨁
= 𝑎1⋯𝑖𝑗(1) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑚1 𝑎1⋯(𝑘1 +1)⋯1 𝑤21 ⋯ 𝑤𝑗(1)(𝑘1 +1) ⋯ 𝑤𝑚1 Now, we prove that Eq. (7) holds for 𝑛𝑗(𝑝) = 𝑘𝑝 + 1 (𝑝 ∈ {1, 2}). Without
𝑖𝑗(1) =1 loss of generality, let 𝑛𝑗(1) = 𝑘1 + 1. Then
∏𝑘1 ( )𝑤21 ⋯𝑤𝑗(1)𝑖
⎛ { 𝑟
1−𝜇1⋯𝑖
𝑗(1) ⋯1 −1 }⎞ 𝑗(1)
⋯𝑤𝑚1
𝑘1 +1
𝑘2
⎜1− log𝑟 1 +
𝑖𝑗(1) =1
,⎟ ⨁ ⨁
∑𝑘1 𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1
⎜ 𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎟
⎜ (𝑟 − 1) 𝑖𝑗(1) =1 𝑗(1)
⎟ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
⎜ { ∏𝑘1 ( 𝜂1⋯𝑖𝑗(1) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1
} ⎟
⎜ 𝑟 −1 𝑗(1)
⎟⨁ 𝑘1 𝑘2
= ⎜ log𝑟 1 +
𝑖𝑗(1) =1
, ⎟
⨁ ⨁ ⨁
∑𝑘1 = 𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1
⎜ (𝑟 − 1) 𝑖𝑗(1) =1
𝑤 21 ⋯𝑤 𝑗(1)𝑖𝑗(1) ⋯𝑤 𝑚1 −1 ⎟
⎜ ∏𝑘1 ( 𝜈1⋯𝑖𝑗(1) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 ⎟ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
⎜ { 𝑟 − 1 𝑗(1) } ⎟ 𝑘1 +1 𝑘2
⎜ log 1 + 𝑖𝑗(1) =1 ⎟ ⨁ ⨁
⎜ 𝑟 ∑𝑘1 ⎟ 𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1
⎝ (𝑟 − 1)
𝑤 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1
𝑖𝑗(1) =1 21 𝑗(1) ⎠ 𝑖𝑗(1) =𝑘1 +1 𝑖𝑗(2) =1
{ ( 1−𝜇1⋯(𝑘 +1)⋯1 )𝑤21 ⋯𝑤𝑗(1)(𝑘 +1) ⋯𝑤𝑚1 }
⎛ 𝑟 1 −1 1
⎞ 𝑘1
⨁ 𝑘2
⨁ ⨁
⎜1− log𝑟 1 + ,⎟ = 𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1
⎜ (𝑟 − 1)𝑤21 ⋯𝑤𝑗(1)(𝑘1 +1) ⋯𝑤𝑚1 −1 ⎟
⎜ { ( 𝜂1⋯(𝑘 +1)⋯1 )𝑤 ⋯𝑤 ⋯𝑤 } ⎟ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
𝑟 1 − 1 21 𝑗(1)(𝑘1 +1) 𝑚1
⎜ log 1 + , ⎟ 𝑘2
⎜ 𝑟 𝑤21 ⋯𝑤𝑗(1)(𝑘 +1) ⋯𝑤𝑚1 −1 ⎟ ⨁
(𝑟 − 1) 1
𝑎1⋯(𝑘1 +1)⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)(𝑘1 +1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1
⎜ { ( 𝜈1⋯(𝑘 +1)⋯1 )𝑤21 ⋯𝑤𝑗(1)(𝑘 +1) ⋯𝑤𝑚1 } ⎟
⎜ 𝑟 1 −1 1 ⎟ 𝑖𝑗(2) =1
⎜ log𝑟 1 + ⎟
⎝ (𝑟 − 1)𝑤21 ⋯𝑤𝑗(1)(𝑘1 +1) ⋯𝑤𝑚1 −1 ⎠
{ ∏𝑘1 +1 ( 1−𝜇1⋯𝑖𝑗(1) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1
}⎞ { ∏𝑘1 ∏𝑘2 ( 1−𝜇1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤 ⋯𝑤
− 1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 } ⎞
⋯𝑤 ⋯𝑤
⎛ 𝑖𝑗(1) =1
𝑟 − 1 𝑗(1) ⎛ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
𝑟
⎜1− log𝑟 1 + ,⎟ ⎜1− log𝑟 1 + ∑𝑘1 ∑ 𝑘2 ,⎟
⎜
∑𝑘1 +1
𝑤 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎟ ⎜ 𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
(𝑟 − 1) 𝑖𝑗(1) =1 21 𝑗(1) ⎜ (𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎟
⎜ ⎟ ⎜ ∏ ∏ ( ) ⎟
⎜ ∏ ( ) { 𝑘 𝑘
𝑟 1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 − 1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 } ⎟ ⨁
𝜂 𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤
} ⎟
1 2
{ 𝑘 1 +1 𝜂
𝑟 1⋯𝑖𝑗(1) ⋯1 − 1
𝑤21 ⋯𝑤 𝑗(1)𝑖 𝑗(1)
⋯𝑤 𝑚1 ⎜ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
⎜ 𝑖𝑗(1) =1 ⎟ = ⎜ log𝑟 1 + ∑𝑘1 ∑𝑘2 , ⎟
= ⎜ log𝑟 1 + ∑𝑘1 +1 , ⎟. ⎜ 𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
(𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1
⎜ (𝑟 − 1) 𝑖𝑗(1) =1
𝑤 21 ⋯𝑤 𝑗(1)𝑖 𝑗(1)
⋯𝑤 𝑚1 −1 ⎟ ⎜ ∏𝑘1 ∏𝑘2 ( 𝜈1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1 ⎟
⎜ ⎟ ⎜ { } ⎟
∏𝑘1 +1 ( 𝜈1⋯𝑖𝑗(1) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 𝑟 − 1 𝑗(1) 𝑗(2)

⎜ { 𝑟 − 1 𝑗(1) } ⎟ ⎜ log 1 + 𝑖𝑗(1) =1 𝑖𝑗(2) =1 ⎟

⎜ 𝑟 ∑𝑘1 ∑𝑘2 ⎟
⎜ log 1 + 𝑖 𝑗(1) =1 ⎟ ⎝
𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1
⎠
⎜ 𝑟 ∑𝑘1 +1 ⎟ (𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1
⎝ 𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑚1 −1 ⎠ ∏ ( )
(𝑟 − 1) 𝑖𝑗(1) =1 𝑗(1)
⎛ { 𝑘 2
𝑟
1−𝜇 1⋯(𝑘1 +1)⋯𝑖𝑗(2) ⋯1
− 1
𝑤 21 ⋯𝑤 𝑗(1)(𝑘1 +1) ⋯𝑤 𝑗(2)𝑖𝑗(2) ⋯𝑤 𝑚1 }
⎞
𝑖𝑗(2) =1
⎜1− log𝑟 1 + ∑ 𝑘2 ,⎟
⎜ 𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
Hence, Eq. (6) holds. ⎜ (𝑟 − 1) 𝑖𝑗(2) =1 21 𝑗(1)(𝑘1 +1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎟
⎜ ∏ ( ) ⎟
(II) There exist two different elements 𝑗(1), 𝑗(2) ∈ 𝑀 such that 𝑛𝑙 = 1 { 𝑘 2
𝑟 1⋯𝑖𝑗(1) ⋯1 − 1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 }
𝜂 𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤
⎜ 𝑖𝑗(2) =1 ⎟
for each 𝑙 ∈ 𝑀 − {𝑗(1), 𝑗(2)}. Then, by Eq. (4), we have ⎜ log 𝑟 1 + ∑𝑘2 , ⎟
⎜ (𝑟 − 1)
𝑤 ⋯𝑤𝑗(1)(𝑘1 +1) ⋯𝑤𝑗(2)𝑖𝑗(2) ⋯𝑤𝑚1 −1
𝑖𝑗(2) =1 21 ⎟
𝑛 𝑛 ⎜ ∏ ( ) ⎟
⨁
1 ⨁
𝑗(1)
⨁
𝑗(2)
⨁
1 ⎜ { 𝑘 2
𝑟
𝜈1⋯𝑖𝑗(1) ⋯1
− 1
𝑤 21 ⋯𝑤 𝑗(1)𝑖𝑗(1) ⋯𝑤 𝑗(2)𝑖𝑗(2) ⋯𝑤 𝑚1 }
⎟
𝑐𝑖1 = ⋯ ⋯ ⋯ 𝑎𝑖2 ⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⎜ log 1 + 𝑖𝑗(2) =1 ⎟
⎜ 𝑟 ∑𝑘2 ⎟
𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1
𝑖2 =1 𝑖𝑗(1) =1 𝑖𝑗(2) =1 𝑖𝑚 =1 ⎝ (𝑟 − 1) 𝑖𝑗(2) =1 21 𝑗(1)(𝑘1 +1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎠
∏ ∏ ( )
⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚𝑖𝑚 ⎛ { 𝑘1 +1 𝑘2
𝑟
1−𝜇1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1
− 1
𝑤 21 ⋯𝑤 𝑗(1)𝑖𝑗(1) ⋯𝑤 𝑗(2)𝑖𝑗(2) ⋯𝑤𝑚1 }
⎞
𝑖𝑗(1) =1 𝑖𝑗(2) =1
⎜1− log𝑟 1 + ∑𝑘1 +1 ∑𝑘2 ,⎟
𝑛 𝑛 ⎜ ⎟
⨁
𝑗(1)
⨁
𝑗(2)
⎜
𝑤 ⋯𝑤
(𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1
⋯𝑤 ⋯𝑤 −1
⎟
= 𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1 . ⎜ { ∏𝑘1 +1 ∏𝑘2 ( 𝜂1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1
} ⎟
⎜ 𝑖𝑗(1) =1 𝑖𝑗(2) =1
𝑟 −1 𝑗(1) 𝑗(2)
⎟
𝑖𝑗(1) =1 𝑖𝑗(2) =1
= ⎜ log𝑟 1 + ∑𝑘1 +1 ∑𝑘2 , ⎟.
⎜ (𝑟 − 1) 𝑖𝑗(1) =1
𝑤 ⋯𝑤𝑗(1)𝑖𝑗(1) ⋯𝑤𝑗(2)𝑖𝑗(2) ⋯𝑤𝑚1 −1
𝑖𝑗(2) =1 21 ⎟
For 𝑛𝑗(1) , 𝑛𝑗(2) ∈ N+ , we use mathematical induction to prove the ⎜ ∏𝑘1 +1 ∏𝑘2 ( 𝜈1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 )𝑤21 ⋯𝑤𝑗(1)𝑖 ⋯𝑤𝑗(2)𝑖 ⋯𝑤𝑚1 ⎟
⎜ { 𝑟 −1 𝑗(1) 𝑗(2) } ⎟
following equation: ⎜ log 1 + 𝑖𝑗(1) =1 𝑖𝑗(2) =1 ⎟
⎜ 𝑟 ∑𝑘1 +1 ∑𝑘2
𝑤 ⋯𝑤 ⋯𝑤 ⋯𝑤 −1 ⎟
𝑛𝑗(1) 𝑛𝑗(2) ⎝ (𝑟 − 1) 𝑖𝑗(1) =1 𝑖𝑗(2) =1 21 𝑗(1)𝑖𝑗(1) 𝑗(2)𝑖𝑗(2) 𝑚1 ⎠
⨁ ⨁
𝑎1⋯𝑖𝑗(1) ⋯𝑖𝑗(2) ⋯1 𝑤21 ⋯ 𝑤𝑗(1)𝑖𝑗(1) ⋯ 𝑤𝑗(2)𝑖𝑗(2) ⋯ 𝑤𝑚1
𝑖𝑗(1) =1 𝑖𝑗(2) =1
Hence, Eq. (7) holds.

730
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

(III) There exist different elements 𝑗(1), … , 𝑗(𝑝) ∈ 𝑀 such that 𝑛𝑙 = 1 where
for each 𝑙 ∈ 𝑀 − {𝑗(1), … , 𝑗(𝑝)}. Then, by a similar method, we can { 𝑛2 𝑛𝑚 }
⎛ ∏ ∏ ( 1−𝜇𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 ⎞
prove that Eq. (5) holds for any 𝑝 = 3, … , 𝑚 − 2. ⎜ 1− log𝑟 1 + ⋯ 𝑟 1 2 𝑚 −1 2 𝑚 ,⎟
(IV) If 𝑛𝑙 ≥ 2 for any 𝑙 ∈ 𝑀, then we use mathematical induction to ⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟
⎜ { 𝑛2 𝑛𝑚 } ⎟
prove Eq. (5). When 𝑛2 = 𝑛3 = ⋯ = 𝑛𝑚 = 2, we have ⎜ ∏ ∏ ( 𝜂𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 ⎟
𝑐𝑖1 = ⎜ log𝑟 1 + ⋯ 𝑟 12 𝑚 −1 2 𝑚 , ⎟.
⨁
2
⨁
2
⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟
𝑐𝑖 1 = ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 ⎜ { } ⎟
𝑛2 𝑛𝑚
𝑖2 =1 𝑖𝑚 =1 ⎜ ∏ ∏ ( 𝜈𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 ⎟
⎜ log𝑟 1 + ⋯ 𝑟 12 𝑚 −1 2 𝑚
⎟
⨁
1
⨁
2
⨁
2
⨁⨁
2
⨁
2
⨁
2
⎝ 𝑖2 =1 𝑖𝑚 =1 ⎠
= ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 .
𝑖2 =1 𝑖3 =1 𝑖𝑚 =1 𝑖2 =2 𝑖3 =1 𝑖𝑚 =1
Next, we study the basic properties of the GPFFWA operator.
By combining (III), it is easy to verify that the following equation holds:
{ ∏2 ∏2 ( 1−𝜇𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 } Proposition 4.4 (Idempotency). Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1 ×
⎛ 𝑖2 =1 ⋯ 𝑖𝑚 =1 𝑟
1 2 𝑚 −1 2 𝑚
⎞
⎜1− log𝑟 1 + ∑2 ∑2 ,⎟ 𝑛2 × ⋯ × 𝑛𝑚 ) such that 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 = (𝜇, 𝜂, 𝜈) for all 𝑖1 , 𝑖2 , … , 𝑖𝑚 , and let 𝑟 > 1.
⎜ (𝑟 − 1) 𝑖2 =1
⋯ 𝑖 =1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1
𝑚 2 ⎟ Then, for any weight vectors 𝑊2 , 𝑊3 , …, 𝑊𝑚 , we have
⎜ ∏2 ∏2 ( 𝜂𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 } ⎟
⎜ { ⎟
𝑖2 =1 ⋯
1 2 𝑚 −1 𝑚
𝑖𝑚 =1 𝑟
2
𝑐𝑖1 = ⎜ log𝑟 1 + ∑2 ∑2 , ⎟. GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) = [(𝜇, 𝜂, 𝜈)]𝑛1 ×1 . (9)
⎜ ⋯ 𝑖 =1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1 ⎟
⎜ (𝑟 − 1) 𝑖2 =1 𝑚 2
⎟
⎜ { ∏2 ∏2 ( 𝜈𝑖 𝑖 ⋯𝑖𝑚 )𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 } ⎟ Proof. For any 1 ≤ 𝑖1 ≤ 𝑛1 , from Theorem 4.3 it follows that the 𝑖1 -th
⎜ 𝑖2 =1 ⋯ 𝑖𝑚 =1 𝑟 1 2 − 1 2
⎟
⎜ log 𝑟 1 + ∑2 ∑ 2 ⎟ component of the GPFFWA operator is shown as follows:
⎝ (𝑟 − 1) 𝑖2 =1 ⋯ 𝑖𝑚 =1 𝑤2𝑖2 ⋯𝑤𝑚𝑖𝑚 −1 ⎠
GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1
Suppose that Eq. (5) holds for 𝑛2 = 𝑘2 , 𝑛3 = 𝑘3 , ⋯ , 𝑛𝑚 = 𝑘𝑚 , that is,
{ 𝑛2 𝑛𝑚 }
⎛ ∏ ∏ ( 1−𝜇 )𝑤 ⋯𝑤 ⎞
𝑘2 𝑘𝑚
⨁ ⨁ ⎜1− log𝑟 1 + ⋯ 𝑟 − 1 2𝑖2 𝑚𝑖𝑚 ,⎟
⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 ⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟
𝑖2 =1 𝑖𝑚 =1 ⎜ { 𝑛2 𝑛𝑚 } ⎟
∏𝑘2 ∏𝑘 ( 1−𝜇 )𝑤 ⋯𝑤 ⎜ ∏ ∏ ( )𝑤 ⋯𝑤 ⎟
⎛ { ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎞ = ⎜ log𝑟 1 + ⋯ 𝑟𝜂 − 1 2𝑖2 𝑚𝑖𝑚 , ⎟
𝑖2 =1
⎜1− log𝑟 1 + 𝑚
∑𝑘2 ∑𝑘 ,⎟ ⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟
⎜ ⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1 ⎟ ⎜ { 𝑛2 𝑛𝑚 } ⎟
⎜ (𝑟 − 1) 𝑖2 =1 𝑚 2
⎟ ⎜ ∏ ∏ ( 𝜈 )𝑤 ⋯𝑤 ⎟
⎜ ∏𝑘2 ∏𝑘𝑚 ( ) ⎟ ⎜ log𝑟 1 + ⋯ 𝑟 − 1 2𝑖2 𝑚𝑖𝑚 ⎟
{ ⋯ 𝑟
𝜂𝑖 𝑖 ⋯𝑖𝑚
1 2 − 1
𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 }
2
⎝ ⎠
⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟ 𝑖2 =1 𝑖𝑚 =1
= ⎜ log𝑟 1 + , ⎟. (8) { ∑ 𝑛2 ∑ 𝑛𝑚 }
∑𝑘2 ∑𝑘𝑚 ⎛ ( ) ⋯ 𝑖 =1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 ⎞
⎜ (𝑟 − 1) 𝑖2 =1
⋯ 𝑖𝑚 =1
𝑤 2𝑖2
⋯𝑤 𝑚𝑖 𝑚
−1 ⎟ ⎜1− log𝑟 1 + 𝑟
1−𝜇
− 1 𝑖2 =1 𝑚 2 ,⎟
⎜ ∏𝑘2 ∏𝑘𝑚 ( 𝜈𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 } ⎟ ⎜
⎜ { ⎟ { } ⎟
⋯ 𝑖 =1 𝑟 1 2 𝑚 −1 2 𝑚
⎜ ( 𝜂 )∑𝑛𝑖 2=1 ⋯ ∑𝑛𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 ⎟
⎜ log 1 + 𝑖2 =1 𝑚 ⎟ = ⎜ log𝑟 1 + 𝑟 − 1 2 𝑚 2 , ⎟
⎜ 𝑟 ∑𝑘2 ∑ 𝑘 ⎟ ⎜
⎝ 𝑖2 =1
⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1
2 ⎠ { } ⎟
(𝑟 − 1) )∑𝑛2 ⋯ ∑𝑛𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚
𝑚
⎜ ( ⎟
Now, we prove that Eq. (5) holds for 𝑛𝑗 = 𝑘𝑗 + 1 (𝑗 ∈ 𝑀). Since ⎜ log𝑟 1 + 𝑟𝜈 − 1 𝑖2 =1 𝑚 2
⎟
⎝ ⎠
Definition 3.3 and Eq. (8), we have
=(𝜇, 𝜂, 𝜈).
𝑘2 𝑘𝑗 +1 𝑘𝑚
⨁ ⨁ ⨁
⋯ ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 Therefore, Eq. (9) as required.
𝑖2 =1 𝑖𝑗 =1 𝑖𝑚 =1
𝑘2
⨁ ⨁
𝑘
𝑗 𝑘𝑚
⨁ ⨁ Proposition 4.5 (Order-preservation). Let ,  ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 )
= ⋯ ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚 such that  ⪯ . Then, for all 𝑖1 ,
𝑖2 =1 𝑖𝑗 =1 𝑖𝑚 =1

𝑘2 𝑘𝑗 +1 𝑘𝑚 GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 ≤P GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 .

⨁ ⨁ ⨁
⋯ ⋯ 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 𝑤2𝑖2 ⋯ 𝑤𝑚𝑖𝑚
𝑖2 =1 𝑖𝑗 =𝑘𝑗 +1 𝑖𝑚 =1
(10)

⎛ ∏𝑘2 ∏𝑘 ( 1−𝜇 ∏𝑘𝑗 +1 )𝑤 ⋯𝑤

{ ⋯ 𝑖 =1 ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎞ Proof. Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 and  = [𝑏𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 . Then,
⎜ 𝑖2 =1 𝑗 𝑚 ⎟
⎜1− log𝑟 1 + ∑𝑘2 ∑𝑘𝑗 +1 ∑𝑘𝑚
,⎟
by Eq. (3), we have (∀𝑖1 , 𝑖2 , … , 𝑖𝑚 ) 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ≤P 𝑏𝑖1 𝑖2 ⋯𝑖𝑚 . If 𝑤2𝑖2 𝑤3𝑖3 ⋯ 𝑤𝑚𝑖𝑚
⎜ (𝑟 − 1) 𝑖2 =1
⋯ 𝑖𝑗 =1
⋯ 𝑤
𝑖𝑚 =1 2𝑖2
⋯𝑤 𝑚𝑖𝑚 −1 ⎟
⎜ ⎟ = 0 for any 𝑖2 , 𝑖3 , …, 𝑖𝑚 . Then, it is easy to verify that
⎜ ∏ ∏ 𝑘𝑗 +1 ∏ ( 𝜂𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 ⎟
{ 𝑘 2 𝑘 𝑚
⋯ 𝑖 =1 ⋯ 𝑖 =1 𝑟 1 2 𝑚 − 1 2 𝑚 } GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 = GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 =
⎜ 𝑖2 =1 𝑗 𝑚 ⎟
= ⎜ log𝑟 1 + ∑𝑘2 ∑𝑘𝑗 +1 ∑𝑘𝑚
, ⎟. (1, 0, 0) for any 𝑖1 . Otherwise, we consider the following three cases:
⎜ ⋯ ⋯ 𝑤
𝑖𝑚 =1 2𝑖2
⋯𝑤 𝑚𝑖𝑚 −1 ⎟ Case 1. 𝜇𝑎𝑖 𝑖 ⋯𝑖 < 𝜇𝑏𝑖 𝑖 ⋯𝑖 and 𝜈𝑎𝑖 𝑖 ⋯𝑖 ≥ 𝜈𝑏𝑖 𝑖 ⋯𝑖 . Then, we have
⎜ (𝑟 − 1) 𝑖2 =1 𝑖𝑗 =1
⎟
⎜ ∏ ∏ 𝑘 +1 ∏ ( ) ⋯𝑤 ⎟
1 2 𝑚 1 2 𝑚 1 2 𝑚 1 2 𝑚
{ 𝑘 2
⋯ 𝑖 =1 ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎟
𝑗 𝑘 𝜈 𝑤
{ }
⎜ 𝑖2 =1 𝑗 𝑚 ∏𝑛2 𝑛𝑚
∏ ( 1−𝜇𝑎 )𝑤 ⋯𝑤
⎜ log𝑟 1 + ∑ 𝑘2 ∑𝑘𝑗 +1 ∑𝑘 ⎟ log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚
− 1 2𝑖2 𝑚𝑖𝑚
⎜ ⋯ 𝑖 =1 ⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1 ⎟
⎝ (𝑟 − 1) 𝑖2 =1 𝑗 𝑚 2
⎠ 𝑖2 =1 𝑖𝑚 =1
{ 𝑛2 𝑛𝑚 }
Combining Definition 4.1 and Theorem 4.2, it is easy to obtain the ∏ ∏ ( 1−𝜇𝑏 )𝑤 ⋯𝑤
> log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚
− 1 2𝑖2 𝑚𝑖𝑚 ,
following statement. 𝑖2 =1 𝑖𝑚 =1
{ 𝑛2 𝑛𝑚 }
∏ ∏ ( 𝜈𝑎 )𝑤 ⋯𝑤
Theorem 4.3. Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 ), log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚
and let 𝑊𝑗 = (𝑤𝑗1 , 𝑤𝑗2 , … , 𝑤𝑗𝑛𝑗 )⊤ ∈ [0, 1]𝑛𝑗 (𝑗 = 2, … , 𝑚) such that 𝑖2 =1 𝑖𝑚 =1
∑ 𝑛𝑗 { 𝑛2 𝑛𝑚 }
𝑤 = 1. Then, for 𝑟 > 1, the GPFFWA operator is shown as follows: ∏ ∏ ( 𝜈𝑏 )𝑤 ⋯𝑤
𝑖 =1 𝑗𝑖𝑗
𝑗 ≥ log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 .
GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) = [𝑐𝑖1 ]𝑛1 ×1 , 𝑖2 =1 𝑖𝑚 =1

731
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Case 2. 𝜇𝑎𝑖 = 𝜇𝑏𝑖 and 𝜈𝑎𝑖 > 𝜈𝑏𝑖 . Then, we have where
1 𝑖2 ⋯𝑖𝑚 1 𝑖2 ⋯𝑖𝑚 1 𝑖2 ⋯𝑖𝑚 1 𝑖2 ⋯𝑖𝑚
𝑛2 𝑛𝑚
{ 𝑛2 𝑛𝑚 } ⨂ ⨂
∏ ∏ ( 1−𝜇𝑎 )𝑤 ⋯𝑤 𝑑𝑖1 = ⋯
𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚
(𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ) 2 . (11)
log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚
− 1 2𝑖2 𝑚𝑖𝑚
𝑖2 =1 𝑖𝑚 =1
𝑖2 =1 𝑖𝑚 =1
{ 𝑛2 𝑛𝑚 } Similar to the proofs in Section 4.1, we can get the following
∏ ∏ ( 1−𝜇𝑏 )𝑤 ⋯𝑤
= log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚
− 1 2𝑖2 𝑚𝑖𝑚 , statements.
𝑖2 =1 𝑖𝑚 =1
{ 𝑛2 𝑛𝑚 }
∏ ∏ ( 𝜈𝑎 )𝑤 ⋯𝑤 Theorem 4.9. Let 𝑟 > 1. Then Eq. (11) can be expressed as follows:
log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 ∏𝑛2 ∏𝑛 ( 𝜇 )𝑤 ⋯𝑤
𝑖2 =1 𝑖𝑚 =1 ⎛ { ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎞
{ 𝑛2 𝑛𝑚 } ⎜ log𝑟 1 + 𝑖2 =1 𝑚
∑𝑛2 ∑𝑛 , ⎟
∏ ∏( 𝜈𝑏𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 ⎜ ⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1 ⎟
𝑖2 =1
> log𝑟 1 + ⋯ 𝑟 1 2 𝑚 −1 2 . ⎜ (𝑟 − 1) 𝑚 2
⎟
⎜ ∏ ∏ ( )
{ ⋯ 𝑖 𝑚=1 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 } ⎟⎟
𝑖2 =1 𝑖𝑚 =1 𝑛 2 𝑛 1−𝜂 𝑤 ⋯𝑤
⎜ 𝑖2 =1 𝑚
Case 3. 𝜇𝑎𝑖 𝑖 ⋯𝑖 = 𝜇𝑏𝑖 𝑖 ⋯𝑖 , 𝜈𝑎𝑖 𝑖 ⋯𝑖 = 𝜈𝑏𝑖 𝑖 ⋯𝑖 and 𝜂𝑎𝑖 𝑖 ⋯𝑖 ≤ 𝑑𝑖1 = ⎜1− log𝑟 1 + ∑𝑛2 ∑𝑛 ,⎟ .
⋯ 𝑖 𝑚=1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1
𝜂𝑏𝑖
1 2 𝑚
. Then, we have
1 2 𝑚 1 2 𝑚 1 2 𝑚 1 2 𝑚 ⎜ (𝑟 − 1) 𝑖2 =1 𝑚 2 ⎟
𝑖 ⋯𝑖
1 2 𝑚 ⎜ ∏𝑛2 ∏𝑛𝑚 ( 1−𝜈𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 } ⎟
⎜ { ⎟
{ 𝑛2 𝑛𝑚 } ⋯ 𝑖 =1 𝑟 1 2 𝑚 −1 2 𝑚
∏ ∏ ( 1−𝜇𝑎𝑖 𝑖 ⋯𝑖 )𝑤2𝑖 ⋯𝑤𝑚𝑖 ⎜1− log𝑟 1 + 𝑖2 =1 𝑚 ⎟
∑𝑛2 ∑𝑛𝑚
log𝑟 1 + ⋯ 𝑟 1 2 𝑚 −1 2 𝑚
⎜ ⋯ 𝑖 =1 𝑤2𝑖 ⋯𝑤𝑚𝑖𝑚 −1 ⎟
𝑖2 =1 𝑖𝑚 =1 ⎝ (𝑟 − 1) 𝑖2 =1 𝑚 2 ⎠
{ 𝑛2 𝑛𝑚 }
∏ ∏ ( 1−𝜇𝑏 )𝑤 ⋯𝑤
= log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚
− 1 2𝑖2 𝑚𝑖𝑚 , Theorem 4.10. Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 ),
𝑖2 =1 𝑖𝑚 =1 and let 𝑊𝑗 = (𝑤𝑗1 , 𝑤𝑗2 , … , 𝑤𝑗𝑛𝑗 )⊤ ∈ [0, 1]𝑛𝑗 (𝑗 = 2, … , 𝑚) such that
{ 𝑛2 𝑛𝑚 } ∑ 𝑛𝑗
∏ ∏ ( 𝜈𝑎 )𝑤 ⋯𝑤 𝑤 = 1. Then, for 𝑟 > 1, the GPFFWG operator is shown as follows:
log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 𝑖 =1 𝑗𝑖𝑗
𝑗

𝑖2 =1 𝑖𝑚 =1 GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) = [𝑑𝑖1 ]𝑛1 ×1 ,

{ 𝑛2 𝑛𝑚 }
∏ ∏ ( 𝜈𝑏 )𝑤 ⋯𝑤
= log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 , where
{ 𝑛2 𝑛𝑚 }
𝑖2 =1 𝑖𝑚 =1
⎛ ∏ ∏ ( 𝜇𝑖 𝑖 ⋯𝑖 )𝑤 ⋯𝑤 ⎞
{ 𝑛2 𝑛𝑚 } ⎜ log𝑟 1 + ⋯ 𝑟 1 2 𝑚 − 1 2𝑖2 𝑚𝑖𝑚 , ⎟
∏ ∏ ( 𝜂𝑎 )𝑤 ⋯𝑤
log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 ⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟
⎜ { 𝑛2 𝑛𝑚 }⎟
𝑖2 =1 𝑖𝑚 =1
⎜ ∏ ∏ ( ) 𝑤 ⋯𝑤 ⎟
{ 𝑛2 𝑛𝑚 } 𝑑𝑖1 = ⎜1− log𝑟 1 + ⋯
1−𝜂
𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 ,⎟ .
∏ ∏ ( 𝜂𝑏 )𝑤 ⋯𝑤
≤ log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 . ⎜ 𝑖2 =1 𝑖𝑚 =1 ⎟
⎜ { 𝑛2 𝑛𝑚 }⎟
𝑖2 =1 𝑖𝑚 =1
⎜ ∏ ∏ ( 1−𝜈 )𝑤 ⋯𝑤 ⎟
For both cases, from Definition 3.2 it follows that Eq. (10) holds. ⎜1− log𝑟 1 + ⋯ 𝑟 𝑖1 𝑖2 ⋯𝑖𝑚 − 1 2𝑖2 𝑚𝑖𝑚 ⎟
⎝ 𝑖2 =1 𝑖𝑚 =1 ⎠

Proposition 4.6 (Boundedness). Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1 ×
Proposition 4.11 (Idempotency). Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1
𝑛2 × ⋯ × 𝑛𝑚 ), and let
⋀ ×𝑛2 ×⋯×𝑛𝑚 ) such that 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 = (𝜇, 𝜂, 𝜈) for all 𝑖1 , 𝑖2 , … , 𝑖𝑚 , and let 𝑟 > 1.
𝛼− = 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 , Then, for any weight vectors 𝑊2 , 𝑊3 , …, 𝑊𝑚 , we have
𝑖1 ,𝑖2 ,…,𝑖𝑚
⋁ GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) = [(𝜇, 𝜂, 𝜈)]𝑛1 ×1 .
𝛼+ = 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 .
𝑖1 ,𝑖2 ,…,𝑖𝑚

Then, for all 𝑖1 , 𝛼 − ≤P GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 ≤P 𝛼 + . Proposition 4.12 (Order-preservation). Let ,  ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 )
such that  ⪯ . Then, for all 𝑖1 ,

Proof. Since Remark 1, for any 𝑖1 , 𝑖2 , … , 𝑖𝑚 , GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 ≤P GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 .
⋀ ⋁
𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ≤P 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ≤P 𝑎𝑖1 𝑖2 ⋯𝑖𝑚
𝑖1 ,𝑖2 ,…,𝑖𝑚 𝑖1 ,𝑖2 ,…,𝑖𝑚 Proposition 4.13 (Boundedness). Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1
× 𝑛2 × ⋯ × 𝑛𝑚 ), and let
and so 𝛼−
≤P 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ≤P Here, 𝛼+ .and [𝛼 − ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 [𝛼 + ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚
⋀
are denoted by − and + , respectively. Then, by Remark 1 and Eq. (3), 𝛼− = 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ,
we have − , + ∈ 𝐓P (𝑚, 𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ) and − ⪯  ⪯ + . Thus, from 𝑖1 ,𝑖2 ,…,𝑖𝑚
⋁
Propositions 4.4 and 4.5 it follows that 𝛼 − = GPFFWA(− ◦𝑊2 ◦𝑊3 ◦ ⋯ 𝛼+ = 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 .
◦𝑊𝑚 )𝑖1 ≤P GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 ≤P GPFFWA(+ ◦𝑊2 ◦𝑊3 𝑖1 ,𝑖2 ,…,𝑖𝑚
◦ ⋯ ◦𝑊𝑚 )𝑖1 = 𝛼 + for any 𝑖1 .
Then, for all 𝑖1 , 𝛼 − ≤P GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )𝑖1 ≤P 𝛼 + .

Proposition 4.7 (Commutativity). Let  ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 ). Then,

Proposition 4.14 (Commutativity). Let  ∈ 𝐓P (𝑚, 𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ). Then,
for any permutation 𝜎 on {2, 3, … , 𝑚},
for any permutation 𝜎 on {2, 3, … , 𝑚},
GPFFWA(◦𝑊𝜎(2) ◦𝑊𝜎(𝑚) ◦ ⋯ ◦𝑊𝜎(𝑚) ) = GPFFWA(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ).
GPFFWG(◦𝑊𝜎(2) ◦𝑊𝜎(𝑚) ◦ ⋯ ◦𝑊𝜎(𝑚) ) = GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ).
Proof. It can be seen from Theorem 4.2 that the conclusion is obvious.
5. The MCDM method using the proposed operators
4.2. Generalized picture fuzzy Frank weighted geometric operator
In this section, we utilize the suggested operators to address the
intricate MCDM problem. An MCDM problem with high-dimensional
Definition 4.8. Let  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 ∈ 𝐓P (𝑚, 𝑛1 × 𝑛2 × ⋯ × 𝑛𝑚 ),
and let 𝑊𝑗 = (𝑤𝑗1 , 𝑤𝑗2 , … , 𝑤𝑗𝑛𝑗 )⊤ ∈ R𝑛𝑗 (𝑗 = 2, … , 𝑚). Then the picture fuzzy data features contains the following five elements [14]:
GPFFWG operator is defined as follows: (1) The alternative set 𝑍 = {𝑧1 , 𝑧2 , … , 𝑧𝑛1 }, where 2 ≤ 𝑛1 < +∞;
GPFFWG(◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) = [𝑑𝑖1 ]𝑛1 ×1 , (2) The criterion set 2 = {𝑓21 , 𝑓22 , … , 𝑓2𝑛2 }, where 2 ≤ 𝑛2 < +∞;

732
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Fig. 2. The process of the proposed method.

(3) Sets of multiple factors 3 = {𝑓31 , 𝑓32 , … , 𝑓3𝑛3 }, 4 = picture fuzzy evaluation information is shown in Tables A.1–A.12 (see
{𝑓41 , 𝑓42 , … , 𝑓4𝑛4 }, ⋯, 𝑚 = {𝑓𝑚1 , 𝑓𝑚2 , … , 𝑓𝑚𝑛𝑚 }. Appendix).
(4) The weight information 𝑊2 = {𝑤21 , 𝑤22 , … , 𝑤2𝑛2 }, 𝑊3 =
{𝑤31 , 𝑤32 , … , 𝑤3𝑛3 }, ⋯, 𝑊𝑚 = {𝑤𝑚1 , 𝑤𝑚2 , … , 𝑤𝑚𝑛𝑚 }, where 𝑊𝑗 ∈ 6.1. Implementation of the proposed approach
∑𝑛
[0, 1]𝑛𝑗 is the weight vector of 𝑗 such that 𝑖 𝑗=1 𝑤𝑗𝑖𝑗 = 1 for
𝑗
1 ≤ 𝑗 ≤ 𝑚. Now, we use the proposed approach to deal with the above problem.
(5) The evaluation information 𝐴 = {𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ∶ (∀1 ≤ 𝑗 ≤ Step 1. According to Tables A.1–A.12, we can get a picture fuzzy
𝑚) 1 ≤ 𝑖𝑗 ≤ 𝑛𝑗 }, where 𝑎𝑖1 𝑖2 ⋯𝑖𝑚 represents the evaluation value tensor  = [𝑎𝑖1 𝑖2 𝑖3 𝑖4 ]5×5×3×4 , where 𝑎𝑖1 𝑖2 𝑖3 𝑖4 denotes the element in the
(𝜇𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜂𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜈𝑖1 𝑖2 ⋯𝑖𝑚 ) of the alternative 𝑧𝑖1 under 𝑓2𝑖2 , 𝑓3𝑖3 , 𝑖1 -th row and 𝑖2 -th column of matrix ∶∶𝑖3 𝑖4 .
⋯, 𝑓𝑚𝑖𝑚 . Step 2. Obviously, 𝑓21 , 𝑓22 , 𝑓23 , 𝑓24 and 𝑓25 are benefit type criteria.
Now, we use the proposed operator to determine the ranking results Then, by Eq. (12), we have ∗ = .
of all alternatives, as shown in Fig. 2. The complete process of the Step 3. Let 𝑟 = 2. Then, by Theorems 4.3 and 4.10, we utilize the
proposed method is listed as follows. GPFFWA and GPFFWG operators as follows:
Step 1. According to the evaluation information, a picture fuzzy GPFFWA(∗ ◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )
tensor  = [𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 is established.
⎡⎛ { } ⎞⎤
Step 2. By considering two types of criteria, we standardize  = ∏ 5 ∏ 3 ∏ 4
( 1−𝜇𝑖 𝑖 𝑖 𝑖 )𝑤2𝑖 𝑤3𝑖 𝑤4𝑖
⎢⎜1− log𝑟 1 + 𝑟 1 2 3 4 −1 2 3 4 ,⎟⎥
[𝑎𝑖1 𝑖2 ⋯𝑖𝑚 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 to ∗ = [𝑎∗𝑖 𝑖 ⋯𝑖 ]𝑛1 ×𝑛2 ×⋯×𝑛𝑚 , where ⎢⎜ 𝑖2 =1 𝑖3 =1 𝑖4 =1
⎟⎥
1 2 𝑚
⎢⎜ ⎟⎥
{ ⎢⎜ { } ⎟⎥
∏ 5 ∏ 3 ∏ 4
( 𝜂𝑖 𝑖 𝑖 𝑖 )𝑤2𝑖 𝑤3𝑖 𝑤4𝑖
𝑎∗𝑖 𝑖 =
(𝜇𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜂𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜈𝑖1 𝑖2 ⋯𝑖𝑚 ) if 𝑓𝑖2 is a benefit type criterion, = ⎢⎜ log𝑟 1 + 𝑟 1234 −1 2 3 4 , ⎟⎥
1 2 ⋯𝑖𝑚 (𝜈𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜂𝑖1 𝑖2 ⋯𝑖𝑚 , 𝜇𝑖1 𝑖2 ⋯𝑖𝑚 ) if 𝑓𝑖2 is a cost type criterion. ⎢⎜ 𝑖2 =1 𝑖3 =1 𝑖4 =1
⎟⎥
⎢⎜ ⎟⎥
⎢⎜ { } ⎟⎥
(12) ∏ ∏ ∏( 𝜈
5 3 4
)
⎢⎜ log 1 + 𝑤 𝑤 𝑤
𝑟 𝑖1 𝑖2 𝑖3 𝑖4 − 1 2𝑖2 3𝑖3 4𝑖4 ⎟⎥
⎢⎜ 𝑟 ⎟⎥
Step 3. According to Theorem 4.3 (or Theorem 4.10), we utilize the ⎣⎝ 𝑖2 =1 𝑖3 =1 𝑖4 =1 ⎠⎦5×1
GPFFWA (or GPFFWG) operator: (∀1 ≤ 𝑖1 ≤ 𝑛1 ) ⎛(0.5212, 0.1963, 0.1509)⎞
∗
⎜ ⎟
𝑥𝑖1 = GPFFWA( ◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 ) ⎜(0.4939, 0.1888, 0.2150)⎟
⎜ ⎟
(or 𝑥𝑖1 = GPFFWG(∗ ◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )) = ⎜(0.4870, 0.1651, 0.2176)⎟ ,
⎜ ⎟
Step 4. For each 𝑖1 , the score of 𝑥𝑖1 is calculated using Eq. (1). ⎜(0.5789, 0.1651, 0.1233)⎟
⎜(0.6743, 0.1437, 0.1001)⎟
Step 5. Determine the ranking result of the alternatives using the ⎝ ⎠
following rule:

𝑆(𝑥𝑖 ) > 𝑆(𝑥𝑗 ) ⟹ 𝑧𝑖 ≫ 𝑧𝑗 , (13) GPFFWG(∗ ◦𝑊2 ◦𝑊3 ◦ ⋯ ◦𝑊𝑚 )

⎡⎛ { } ⎞⎤
∏ 5 ∏ 3 ∏ 4
( 𝜇𝑖 𝑖 𝑖 𝑖 )𝑤 𝑤 𝑤
where the symbol ≫ means that ‘‘superior than to’’. ⎢⎜ log𝑟 1 + 𝑟 1 2 3 4 − 1 2𝑖2 3𝑖3 4𝑖4 , ⎟⎥
⎢⎜ 𝑖2 =1 𝑖3 =1 𝑖4 =1
⎟⎥
⎢⎜ ⎟⎥
6. An illustrative example ⎢⎜ { } ⎟⎥
∏ ∏ ∏ ( 1−𝜂
5 3 4
)𝑤2𝑖 𝑤3𝑖 𝑤4𝑖
= ⎢⎜1− log𝑟 1 + 𝑖 𝑖
𝑟 1234 −1 𝑖 𝑖 2 3 4 ,⎟⎥
With the continuous intensification of urbanization, people have ⎢⎜ 𝑖2 =1 𝑖3 =1 𝑖4 =1
⎟⎥
⎢⎜ ⎟⎥
put forward higher requirements for living standards. The owners’ ⎢⎜ { } ⎟⎥
∏ 5 ∏ 3 ∏ 4
( )
committee of a community measures the service level of five local ⎢⎜1− log 1 + 𝑟 𝑖1 𝑖2 𝑖3 𝑖4 − 1 2𝑖2 3𝑖3 4𝑖4 ⎟⎥
1−𝜈 𝑤 𝑤 𝑤
⎢⎜ 𝑟 ⎟⎥
property management teams 𝑧1 , 𝑧2 , 𝑧3 , 𝑧4 and 𝑧5 under five criteria ⎣⎝ 𝑖2 =1 𝑖3 =1 𝑖4 =1 ⎠⎦5×1
(i.e., basic equipment 𝑓21 , security capability 𝑓22 , service capability 𝑓23 , ⎛(0.5425, 0.1650, 0.1614)⎞
environmental health 𝑓24 and cultural entertainment 𝑓25 , as detailed ⎜ ⎟
⎜(0.5180, 0.1697, 0.2056)⎟
in [54]) to select the best property management team. It is noted that ⎜ ⎟
the evaluation value of each property management team is affected by = ⎜(0.5003, 0.1430, 0.2108)⎟ .
factors such as time and region. To improve the quality of decision- ⎜ ⎟
⎜(0.5823, 0.1534, 0.1424)⎟
making, the owners’ committee considers two factors: time (i.e., 𝑓31 , 𝑓32 ⎜(0.6752, 0.1381, 0.1053)⎟
⎝ ⎠
and 𝑓33 in the past three years) and region (i.e., 𝑓41 , 𝑓42 , 𝑓43 and
𝑓44 ). This section uses the information from the cases in [54] for Step 4. For the data of the previous step, we can calculate the score
ease of comparison. Specifically, the weight vectors of the criterion of each alternative using Eq. (1), and the results are shown in Table 2.
set 2 = {𝑓21 , 𝑓22 , 𝑓23 , 𝑓24 , 𝑓25 }, time set 3 = {𝑓31 , 𝑓32 , 𝑓33 } and Step 5. By the rule (13), the ranking results are shown in Ta-
region set 4 = {𝑓41 , 𝑓42 , 𝑓43 , 𝑓44 } are 𝑊2 = (0.15, 0.3, 0.2, 0.15, 0.2)⊤ , ble 2. Therefore, through the calculation of the GPFFWA and GPFFWG
𝑊3 = (0.2, 0.3, 0.5)⊤ and 𝑊4 = (0.15, 0.3, 0.25, 0.3)⊤ , respectively. The operators, 𝑧5 is confirmed as the optimal alternative.

733
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Table 2
Outcomes of the numerical example.
Operators 𝑆(𝑧1 ) 𝑆(𝑧2 ) 𝑆(𝑧3 ) 𝑆(𝑧4 ) 𝑆(𝑧5 ) Ranking results
The GPFFWA operator 0.3703 0.2789 0.2694 0.4556 0.5742 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3
The GPFFWG operator 0.3811 0.3124 0.2895 0.4399 0.5699 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3

Fig. 3. Outcome of sensitivity analysis.

6.2. Sensitivity analysis new overall ranking of the alternatives should be identical to the
original overall ranking of the un-decomposed problem.
To demonstrate the robustness of the proposed operators, this part
conducts a sensitivity analysis by using the numerical example men- Now, we follow the test criteria to demonstrate the validity of the
tioned above. Specifically, when 𝑟 takes 199 integers in [2,200], the proposed approach.
scores of the five alternatives are calculated by the method in Section 5.
Validity assessment using criterion 1. According to the idea of
The results to improve visibility are shown in Fig. 3. Through Fig. 3,
criterion 1, the non-optimal alternative is replaced by a worse alter-
we have the following findings:
native. If the picture fuzzy tensor formed by the original data and the
(1) The change in the value of 𝑟 affects the score of each alternative. picture fuzzy tensor formed by the modified data are recorded as  and
When the value of the parameter 𝑟 is small, the scores of the  ′ , respectively, then  ⪯  ′ . Combining the order-preserving property
alternatives determined by the GPFFWA (GPFFWG) operator of the proposed GPFFWA and GPFFWG operators (see Propositions 4.5
increase (decrease) as 𝑟 increases. However, when the value of 𝑟 and 4.12) and Definition 3.4, it is easy to see that the proposed method
reaches a certain level, the scores of the alternatives determined satisfies criterion 1. For example, in each decision matrix, the non-
by the GPFFWA and GPFFWG operators tend to stabilize with optimal alternative 𝑧3 is replaced by the worse alternative 𝑧′3 . The
further increases in 𝑟. recommended values are shown in Table 3. Then, the proposed method
(2) There is no change in the ranking results of the five alternatives, is applied to the modified data to calculate the scores of various
that is, 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3 . Therefore, the optimal alternatives. The ranking results are shown in Table 4 (take 𝑟 = 2).
alternative in each case is 𝑧5 . Note from the output results that 𝑧5 is still the best alternative.
Validity assessment using criteria 2 and 3. We decompose the
given MCDM problem into the following five sub-problems:
6.3. Validity test

(P1) The set of alternatives is {𝑧2 , 𝑧3 , 𝑧4 , 𝑧5 }.

For a given decision problem, different answers may be generated
(P2) The set of alternatives is {𝑧1 , 𝑧3 , 𝑧4 , 𝑧5 }.
if different methods are selected. Clearly, testing the relative perfor-
(P3) The set of alternatives is {𝑧1 , 𝑧2 , 𝑧4 , 𝑧5 }.
mance of these methods is particularly important. Therefore, Wang and
(P4) The set of alternatives is {𝑧1 , 𝑧2 , 𝑧3 , 𝑧5 }.
Triantaphyllou [55] provided the following three test criteria:
(P5) The set of alternatives is {𝑧1 , 𝑧2 , 𝑧3 , 𝑧4 }.
∙ Test criterion 1. An effective MCDM method should remain the
indication of the best alternative when a non-optimal alternative Now, we apply the proposed strategy to solve these sub-problems, and
is replaced by an even worse alternative. the alternatives are sorted as 𝑧5 ≫ 𝑧4 ≫ 𝑧2 ≫ 𝑧3 , 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧3 ,
∙ Test criterion 2. The ranking of alternatives determined by an 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 , 𝑧5 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3 , and 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3 .
effective MCDM method should meet the transitive property. By combining the ranking results of these sub-problems, we obtain the
∙ Test criterion 3. For the same decision problem and when using final ranking sequence, which is 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3 , equivalent
the same MCDM method, after combining the rankings of the to the answer of the given original problem. This shows a transitive
smaller problems that an MCDM problem is decomposed into, the property. Therefore, the proposed technology meets criteria 2 and 3.

734
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Table 3
The information of 𝑧′3 in each picture fuzzy decision matrix.
𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
∶∶11 (0.15, 0.12, 0.61) (0.14, 0.00, 0.35) (0.15, 0.15, 0.40) (0.43, 0.13, 0.25) (0.60, 0.16, 0.22)
∶∶12 (0.34, 0.02, 0.44) (0.55, 0.13, 0.25) (0.45, 0.05, 0.30) (0.62, 0.03, 0.14) (0.60, 0.06, 0.12)
∶∶13 (0.15, 0.14, 0.44) (0.54, 0.13, 0.15) (0.35, 0.25, 0.30) (0.32, 0.00, 0.54) (0.55, 0.16, 0.22)
∶∶14 (0.55, 0.00, 0.30) (0.14, 0.13, 0.45) (0.30, 0.25, 0.30) (0.62, 0.13, 0.24) (0.72, 0.16, 0.01)
∶∶21 (0.75, 0.11, 0.05) (0.74, 0.00, 0.05) (0.38, 0.00, 0.10) (0.35, 0.03, 0.24) (0.60, 0.05, 0.25)
∶∶22 (0.24, 0.12, 0.31) (0.14, 0.13, 0.56) (0.26, 0.34, 0.40) (0.62, 0.13, 0.20) (0.60, 0.10, 0.25)
∶∶23 (0.35, 0.14, 0.41) (0.24, 0.13, 0.35) (0.32, 0.23, 0.10) (0.33, 0.13, 0.25) (0.61, 0.08, 0.14)
∶∶24 (0.15, 0.12, 0.41) (0.24, 0.23, 0.35) (0.45, 0.25, 0.10) (0.32, 0.13, 0.24) (0.60, 0.16, 0.18)
∶∶31 (0.35, 0.12, 0.11) (0.54, 0.03, 0.40) (0.35, 0.25, 0.30) (0.32, 0.13, 0.43) (0.80, 0.16, 0.03)
∶∶32 (0.75, 0.12, 0.01) (0.30, 0.23, 0.35) (0.35, 0.25, 0.10) (0.22, 0.13, 0.24) (0.62, 0.16, 0.22)
∶∶33 (0.65, 0.00, 0.11) (0.24, 0.23, 0.36) (0.35, 0.25, 0.30) (0.52, 0.13, 0.12) (0.56, 0.13, 0.28)
∶∶34 (0.35, 0.12, 0.31) (0.84, 0.00, 0.05) (0.15, 0.25, 0.30) (0.42, 0.13, 0.24) (0.44, 0.16, 0.22)

Table 4
Validity assessment using criterion 1.
Operators 𝑆(𝑧1 ) 𝑆(𝑧2 ) 𝑆(𝑧′3 ) 𝑆(𝑧4 ) 𝑆(𝑧5 ) Ranking results
The GPFFWA operator 0.3703 0.2789 0.1798 0.4556 0.5742 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧′3
The GPFFWG operator 0.3811 0.3124 0.1730 0.4399 0.5699 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧′3

Table 5 different dimensions. In general, our operators are more effective

The ranking results compared with the method in [14].
and can deal with more complex decision-making problems.
Approaches Ranking result The optimal
(3) Compare our technology with the operators suggested in [14,50,
alternative
51]. The operators in [14,50,51] can identify high-dimensional
The operator-based TOPSIS method [14] 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧3 ≫ 𝑧2 𝑧5
data. However, these operators have one thing in common: they
The proposed GPFFWA operator 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3 𝑧5
The proposed GPFFWG operator 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧2 ≫ 𝑧3 𝑧5 cannot satisfy the preferences of decision-makers. Our operators
have a parameter that allows decision-makers to choose the
appropriate value to express their preferences according to their
needs. Therefore, our operators are more flexible than other
6.4. Comparative analysis ones [14,50,51].

As previously described, methods for dealing with MCDM problems Through comparative analysis, the proposed operators are superior
with high-dimensional data features are scarce. The operator-based to other operators because they can not only effectively deal with
TOPSIS method [14] combines the GPFWIA and GPFWIG operators high-dimensional picture fuzzy data but also contain a parameter that
and the TOPSIS method to deal with complex MCDM problems. This satisfies the preferences of decision-makers.
section uses the operator-based TOPSIS method to participate in the
comparison. Table 5 is obtained by using the above numerical example. 7. Conclusion
It can be found from Table 5 that the ranking result obtained by the
operator-based TOPSIS method is 𝑧5 ≫ 𝑧4 ≫ 𝑧1 ≫ 𝑧3 ≫ 𝑧2 , which In practical decision-making, we frequently encounter numerous
is slightly different from the results obtained by the proposed two criteria and complex, high-dimensional data, which can make the
operators. However, the decision results are consistent. Therefore, the decision-making process intricate and challenging. In this paper, we
proposed GPFFWA and GPFFWG operators provide another technique introduce the Frank operators into the picture fuzzy tensor, propose the
for determining the best choice in a decision support system. GPFFWA and GPFFWG operators, and explore their properties in detail.
In addition, we compare and evaluate the proposed technology with In addition, we apply the proposed operators to the MCDM problem
the early works in Table 1. Our key findings are as follows: with high-dimensional data features in the picture fuzzy environment
and conduct decision analysis and ranking based on this. Finally, the
(1) Compare our technology with the operators suggested in [30– feasibility and effectiveness of the proposed method are verified by a
33]. Some information aggregation operators are proposed to numerical example and its discussions.
solve the problems in the decision system. However, the oper- The results of this paper will have a positive impact on academic
ators in [30,31] and the operators in [32,33] are only suitable research and practical application in the field of fuzzy decision-making
for aggregating fuzzy information and intuitionistic fuzzy infor- and information integration. However, this work also has some limita-
mation, respectively, and are not suitable for processing picture tions:
fuzzy information. However, our operators can successfully deal
with picture fuzzy information. Note that picture fuzzy sets are • Since the proposed operators involve the processing of high-
direct generalizations of fuzzy sets and intuitionistic fuzzy sets. dimensional data, their computational complexity is high.
Therefore, our operators are more powerful and more applicable • The proposed operators need to select and adjust a parameter,
than other ones [30–33]. and different parameter selections may lead to different results.
(2) Compare our technology with the operators suggested in [34– However, the process of parameter selection is relatively difficult.
49]. The operators in [34–49] are basic operators for aggregating • Weight is a quantitative concept used to measure the relative
picture fuzzy data, and they have important applications in daily importance of a factor in the overall evaluation. In the high-
life. However, these operators cannot identify high-dimensional dimensional data environment, our MCDM technology does not
data. Our operators can identify uncertain information across establish a method to obtain the weight of factors.

735
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Table A.1
Picture fuzzy decision matrix ∶∶11 under the conditions of time 𝑓31 and area 𝑓41 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.61, 0.21, 0.15) (0.35, 0.20, 0.30) (0.34, 0.22, 0.30) (0.40, 0.16, 0.32) (0.55, 0.25, 0.12)
𝑧2 (0.65, 0.10, 0.15) (0.33, 0.20, 0.41) (0.45, 0.22, 0.23) (0.25, 0.35, 0.30) (0.64, 0.16, 0.20)
𝑧3 (0.25, 0.12, 0.41) (0.24, 0.23, 0.35) (0.35, 0.25, 0.40) (0.52, 0.13, 0.24) (0.60, 0.16, 0.22)
𝑧4 (0.41, 0.13, 0.24) (0.30, 0.20, 0.40) (0.52, 0.13, 0.24) (0.65, 0.23, 0.03) (0.55, 0.23, 0.03)
𝑧5 (0.71, 0.13, 0.04) (0.50, 0.20, 0.30) (0.62, 0.13, 0.14) (0.65, 0.23, 0.03) (0.45, 0.23, 0.13)

Table A.2
Picture fuzzy decision matrix ∶∶12 under the conditions of time 𝑓31 and area 𝑓42 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.61, 0.11, 0.25) (0.31, 0.20, 0.40) (0.44, 0.22, 0.30) (0.50, 0.16, 0.32) (0.45, 0.25, 0.20)
𝑧2 (0.55, 0.15, 0.15) (0.41, 0.20, 0.31) (0.35, 0.22, 0.43) (0.35, 0.35, 0.30) (0.54, 0.16, 0.20)
𝑧3 (0.44, 0.12, 0.41) (0.55, 0.13, 0.25) (0.55, 0.05, 0.20) (0.62, 0.13, 0.14) (0.70, 0.16, 0.12)
𝑧4 (0.55, 0.13, 0.24) (0.70, 0.20, 0.10) (0.62, 0.13, 0.24) (0.55, 0.23, 0.13) (0.65, 0.23, 0.08)
𝑧5 (0.81, 0.13, 0.04) (0.40, 0.30, 0.30) (0.71, 0.13, 0.14) (0.45, 0.23, 0.32) (0.55, 0.23, 0.13)

Table A.3
Picture fuzzy decision matrix ∶∶13 under the conditions of time 𝑓31 and area 𝑓43 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.62, 0.11, 0.10) (0.55, 0.10, 0.30) (0.34, 0.32, 0.10) (0.50, 0.16, 0.32) (0.65, 0.25, 0.05)
𝑧2 (0.75, 0.10, 0.15) (0.73, 0.10, 0.11) (0.65, 0.22, 0.13) (0.43, 0.35, 0.10) (0.74, 0.21, 0.05)
𝑧3 (0.35, 0.14, 0.41) (0.54, 0.23, 0.15) (0.65, 0.25, 0.10) (0.52, 0.13, 0.24) (0.55, 0.16, 0.22)
𝑧4 (0.71, 0.13, 0.14) (0.40, 0.20, 0.30) (0.52, 0.13, 0.14) (0.65, 0.23, 0.10) (0.75, 0.13, 0.03)
𝑧5 (0.51, 0.13, 0.04) (0.70, 0.11, 0.12) (0.72, 0.13, 0.14) (0.45, 0.23, 0.23) (0.55, 0.22, 0.13)

Table A.4
Picture fuzzy decision matrix ∶∶14 under the conditions of time 𝑓31 and area 𝑓44 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.50, 0.17, 0.15) (0.35, 0.25, 0.30) (0.36, 0.22, 0.26) (0.45, 0.16, 0.32) (0.65, 0.15, 0.12)
𝑧2 (0.75, 0.10, 0.15) (0.53, 0.20, 0.11) (0.46, 0.22, 0.13) (0.55, 0.35, 0.10) (0.40, 0.11, 0.40)
𝑧3 (0.55, 0.12, 0.20) (0.24, 0.23, 0.45) (0.45, 0.25, 0.20) (0.62, 0.13, 0.24) (0.78, 0.16, 0.01)
𝑧4 (0.41, 0.13, 0.24) (0.30, 0.30, 0.40) (0.72, 0.03, 0.24) (0.65, 0.23, 0.03) (0.45, 0.23, 0.03)
𝑧5 (0.51, 0.13, 0.11) (0.70, 0.20, 0.02) (0.62, 0.13, 0.14) (0.66, 0.23, 0.10) (0.55, 0.21, 0.13)

Table A.5
Picture fuzzy decision matrix ∶∶21 under the conditions of time 𝑓32 and area 𝑓41 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.44, 0.11, 0.15) (0.46, 0.20, 0.30) (0.44, 0.22, 0.10) (0.50, 0.13, 0.32) (0.55, 0.20, 0.12)
𝑧2 (0.45, 0.05, 0.15) (0.33, 0.10, 0.31) (0.45, 0.22, 0.25) (0.35, 0.35, 0.30) (0.74, 0.06, 0.20)
𝑧3 (0.85, 0.11, 0.01) (0.74, 0.03, 0.05) (0.38, 0.25, 0.10) (0.55, 0.03, 0.24) (0.60, 0.05, 0.22)
𝑧4 (0.51, 0.13, 0.24) (0.40, 0.15, 0.40) (0.72, 0.13, 0.14) (0.55, 0.23, 0.03) (0.58, 0.23, 0.17)
𝑧5 (0.51, 0.13, 0.14) (0.45, 0.20, 0.30) (0.58, 0.16, 0.14) (0.67, 0.23, 0.05) (0.47, 0.13, 0.15)

In the future, we will use efficient and highly interpretable algo- Declaration of competing interest
rithms to optimize the information aggregation method further based
on picture fuzzy tensors to overcome the limitations mentioned The authors declare that they have no known competing finan-
above. cial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgments
CRediT authorship contribution statement

This work was supported by the Science and Technology Program of

Guizhou Province (QKHJC[2024]QN081; QKHZC[2023]372), the Nat-
Jiulin Jin: Writing – review & editing, Writing – original draft, Vi-
ural Science Research Foundation of Education Department of Guizhou
sualization, Validation, Supervision, Methodology. Dragan Pamucar:
Province (QJJ[2024]190; QJJ[2023]063), and the Doctoral Research
Writing – review & editing. Shangshu Shi: Writing – review & editing,
Start-Up Foundation of Guiyang University (GYU-KY-2024).
Writing – original draft, Visualization, Validation, Supervision, Method-
ology. Hui Zhang: Writing – review & editing, Writing – original draft,
Visualization, Validation, Supervision, Software, Methodology, Data Appendix
curation. Wen Teng: Writing – review & editing, Writing – original
draft, Visualization, Validation, Supervision, Methodology. See Tables A.1–A.12.

736
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Table A.6
Picture fuzzy decision matrix ∶∶22 under the conditions of time 𝑓32 and area 𝑓42 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.70, 0.15, 0.15) (0.35, 0.30, 0.30) (0.44, 0.21, 0.30) (0.50, 0.15, 0.32) (0.55, 0.15, 0.10)
𝑧2 (0.75, 0.10, 0.15) (0.53, 0.20, 0.11) (0.46, 0.22, 0.21) (0.24, 0.35, 0.31) (0.60, 0.16, 0.20)
𝑧3 (0.24, 0.12, 0.21) (0.24, 0.13, 0.36) (0.36, 0.24, 0.40) (0.62, 0.13, 0.20) (0.60, 0.10, 0.25)
𝑧4 (0.41, 0.03, 0.24) (0.20, 0.40, 0.40) (0.42, 0.13, 0.24) (0.68, 0.23, 0.03) (0.56, 0.13, 0.03)
𝑧5 (0.71, 0.14, 0.04) (0.55, 0.10, 0.30) (0.60, 0.12, 0.14) (0.75, 0.13, 0.03) (0.35, 0.43, 0.11)

Table A.7
Picture fuzzy decision matrix ∶∶23 under the conditions of time 𝑓32 and area 𝑓43 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.47, 0.11, 0.15) (0.32, 0.20, 0.30) (0.64, 0.22, 0.11) (0.30, 0.15, 0.32) (0.45, 0.26, 0.12)
𝑧2 (0.35, 0.12, 0.15) (0.38, 0.27, 0.31) (0.35, 0.32, 0.23) (0.65, 0.15, 0.05) (0.52, 0.14, 0.20)
𝑧3 (0.45, 0.14, 0.31) (0.54, 0.13, 0.25) (0.35, 0.23, 0.10) (0.43, 0.13, 0.25) (0.64, 0.08, 0.14)
𝑧4 (0.51, 0.11, 0.17) (0.31, 0.20, 0.24) (0.62, 0.14, 0.24) (0.56, 0.24, 0.13) (0.61, 0.23, 0.11)
𝑧5 (0.51, 0.13, 0.17) (0.46, 0.12, 0.30) (0.41, 0.21, 0.14) (0.74, 0.14, 0.11) (0.53, 0.23, 0.17)

Table A.8
Picture fuzzy decision matrix ∶∶24 under the conditions of time 𝑓32 and area 𝑓44 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.66, 0.05, 0.15) (0.35, 0.14, 0.30) (0.24, 0.22, 0.31) (0.62, 0.16, 0.12) (0.44, 0.25, 0.13)
𝑧2 (0.65, 0.17, 0.15) (0.35, 0.20, 0.41) (0.45, 0.24, 0.23) (0.25, 0.35, 0.30) (0.62, 0.16, 0.20)
𝑧3 (0.25, 0.12, 0.41) (0.24, 0.23, 0.35) (0.65, 0.25, 0.10) (0.52, 0.13, 0.24) (0.60, 0.16, 0.08)
𝑧4 (0.41, 0.13, 0.24) (0.70, 0.20, 0.10) (0.52, 0.13, 0.24) (0.65, 0.23, 0.03) (0.65, 0.23, 0.02)
𝑧5 (0.71, 0.13, 0.04) (0.70, 0.20, 0.10) (0.72, 0.12, 0.14) (0.65, 0.11, 0.03) (0.55, 0.23, 0.17)

Table A.9
Picture fuzzy decision matrix ∶∶31 under the conditions of time 𝑓33 and area 𝑓41 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.80, 0.01, 0.15) (0.75, 0.20, 0.05) (0.64, 0.22, 0.10) (0.70, 0.16, 0.02) (0.55, 0.15, 0.12)
𝑧2 (0.85, 0.10, 0.05) (0.33, 0.20, 0.41) (0.65, 0.02, 0.23) (0.65, 0.05, 0.30) (0.74, 0.16, 0.05)
𝑧3 (0.55, 0.12, 0.11) (0.54, 0.03, 0.35) (0.35, 0.25, 0.30) (0.32, 0.13, 0.43) (0.80, 0.16, 0.03)
𝑧4 (0.41, 0.13, 0.24) (0.30, 0.20, 0.40) (0.72, 0.13, 0.14) (0.45, 0.23, 0.03) (0.85, 0.03, 0.03)
𝑧5 (0.91, 0.03, 0.04) (0.78, 0.20, 0.01) (0.62, 0.13, 0.14) (0.95, 0.02, 0.03) (0.78, 0.03, 0.13)

Table A.10
Picture fuzzy decision matrix ∶∶32 under the conditions of time 𝑓33 and area 𝑓42 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.68, 0.11, 0.15) (0.40, 0.30, 0.30) (0.48, 0.22, 0.30) (0.52, 0.16, 0.32) (0.63, 0.25, 0.12)
𝑧2 (0.75, 0.10, 0.15) (0.49, 0.10, 0.41) (0.55, 0.22, 0.23) (0.45, 0.15, 0.30) (0.64, 0.16, 0.20)
𝑧3 (0.85, 0.12, 0.01) (0.32, 0.23, 0.35) (0.45, 0.25, 0.10) (0.52, 0.13, 0.24) (0.62, 0.16, 0.12)
𝑧4 (0.41, 0.13, 0.24) (0.67, 0.20, 0.11) (0.52, 0.13, 0.24) (0.63, 0.23, 0.03) (0.75, 0.13, 0.03)
𝑧5 (0.81, 0.13, 0.04) (0.70, 0.20, 0.10) (0.89, 0.03, 0.04) (0.95, 0.01, 0.03) (0.96, 0.01, 0.01)

Table A.11
Picture fuzzy decision matrix ∶∶33 under the conditions of time 𝑓33 and area 𝑓43 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.70, 0.11, 0.03) (0.35, 0.09, 0.21) (0.34, 0.09, 0.21) (0.60, 0.16, 0.17) (0.55, 0.25, 0.02)
𝑧2 (0.65, 0.10, 0.13) (0.33, 0.20, 0.27) (0.47, 0.22, 0.17) (0.35, 0.35, 0.20) (0.32, 0.16, 0.04)
𝑧3 (0.65, 0.12, 0.11) (0.24, 0.23, 0.26) (0.35, 0.25, 0.09) (0.52, 0.13, 0.12) (0.56, 0.13, 0.28)
𝑧4 (0.91, 0.01, 0.04) (0.70, 0.20, 0.03) (0.62, 0.13, 0.24) (0.65, 0.11, 0.03) (0.85, 0.04, 0.03)
𝑧5 (0.78, 0.13, 0.04) (0.80, 0.10, 0.03) (0.72, 0.13, 0.14) (0.66, 0.23, 0.03) (0.77, 0.13, 0.03)

737
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

Table A.12
Picture fuzzy decision matrix ∶∶34 under the conditions of time 𝑓33 and area 𝑓44 .
𝑍 ⧵ 2 𝑓21 𝑓22 𝑓23 𝑓24 𝑓25
𝑧1 (0.88, 0.11, 0.01) (0.75, 0.10, 0.03) (0.54, 0.22, 0.10) (0.77, 0.16, 0.02) (0.75, 0.23, 0.02)
𝑧2 (0.55, 0.10, 0.15) (0.33, 0.10, 0.51) (0.75, 0.02, 0.13) (0.55, 0.35, 0.08) (0.74, 0.16, 0.05)
𝑧3 (0.55, 0.12, 0.31) (0.84, 0.03, 0.05) (0.35, 0.25, 0.30) (0.42, 0.13, 0.24) (0.44, 0.16, 0.22)
𝑧4 (0.71, 0.13, 0.04) (0.30, 0.20, 0.20) (0.82, 0.13, 0.05) (0.65, 0.23, 0.05) (0.85, 0.01, 0.03)
𝑧5 (0.82, 0.13, 0.04) (0.90, 0.02, 0.03) (0.62, 0.16, 0.15) (0.95, 0.01, 0.03) (0.45, 0.23, 0.13)

References [28] M.Z. Nezhad, J. Nazarian-Jashnabadi, J. Rezazadeh, et al., Assessing dimensions

influencing IoT implementation readiness in industries: A fuzzy DEMATEL and
[1] S.K. Sahoo, B.B. Choudhury, P.R. Dhal, A bibliometric analysis of material fuzzy AHP analysis, J. Soft Comput. Decis. Anal. 1 (1) (2023) 102–123.
selection using MCDM methods: Trends and insights, Spectr. Mech. Eng. Oper. [29] J.L. Jin, H. Garg, Intuitionistic fuzzy three-way ranking-based TOPSIS approach
Res. 1 (1) (2024) 189–205. with a novel entropy measure and its application to medical treatment selection,
[2] U. Elraaid, I. Badi, M.B. Bouraima, Identifying and addressing obstacles to project Adv. Eng. Softw. 180 (2023) 103459.
management office success in construction projects: An AHP approach, Spectr. [30] S. Moslem, Z. Stevic, I. Tanackov, et al., Sustainable development solutions of
Decis. Mak. Appl. 1 (1) (2024) 33–45. public transportation: An integrated IMF SWARA and Fuzzy Bonferroni operator,
[3] L.A. Zadeh, Fuzzy sets, Inf. Control 8 (3) (1965) 338–353. Sustainable Cities Soc. 93 (2023) 104530.
[4] K. Atanassov, Intuitionistic fuzzy sets, Fuzzy Sets Syst. 20 (1) (1986) 87–96. [31] D. Garcia-Zamora, A. Cruz, F. Neres, et al., Admissible OWA operators for fuzzy
[5] B.C. Cuong, Picture fuzzy sets, J. Comput. Sci. Cybern. 30 (4) (2014) 409–420. numbers, Fuzzy Sets Syst. 480 (2024) 108863.
[6] J.J. Peng, X.G. Chen, X.K. Wang, et al., Picture fuzzy decision-making theories [32] H. Garg, A. Hussain, K. Ullah, Multi-attribute group decision-making algorithm
and methodologies: A systematic review, Int. J. Syst. Sci. 54 (13) (2023) based on intuitionistic fuzzy rough Schweizer-Sklar aggregation operators, Soft
2663–2675. Comput. (2023) http://dx.doi.org/10.1007/s00500-023-09424-x.
[7] V. Simic, S. Karagoz, M. Deveci, et al., Picture fuzzy extension of the CODAS [33] A. Hussain, H. Wang, K. Ullah, et al., Novel intuitionistic fuzzy Aczel Alsina
method for multi-criteria vehicle shredding facility location, Expert Syst. Appl. Hamy mean operators and their applications in the assessment of construction
175 (2021) 114644. material, Complex Intell. Syst. 10 (2024) 1061–1086.
[8] J. Fan, D. Han, M. Wu, Picture fuzzy Additive Ratio Assessment Method (ARAS) [34] C.Y. Wang, Hesitant Fuzzy Set and Picture Fuzzy Set with their Application
and VIseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR) method for Research, (Ph.D. thesis), Hunan University, Changsha, 2015.
multi-attribute decision problem and their application, Complex Intell. Syst. 9 [35] H. Garg, Some picture fuzzy aggregation operators and their applications to
(2023) 5345–5357. multi-criteria decision-making, Arab. J. Sci. Eng. 42 (2017) 5275–5290.
[9] A. Singh, S. Kumar, Picture fuzzy VIKOR-TOPSIS approach based on knowledge [36] R.T. Zhang, Y.P. Xing, J. Wang, et al., A novel multiattribute decision-making
and accuracy measures for suitable adsorbent decision making, Appl. Soft method based on Point-Choquet aggregation operators and its application in
Comput. 147 (2023) 110807. supporting the hierarchical medical treatment system in China, Int. J. Environ.
[10] P. Rani, S.M. Chen, A.R. Mishra, Multi-attribute decision-making based on Res. Public Health 18 (5) (2018) 1718.
similarity measure between picture fuzzy sets and the MARCOS method, Inform. [37] Y. Ju, D. Ju, E.S. Gonzalez, et al., Study of site selection of electric vehicle charg-
Sci. 658 (2024) 119990. ing station based on extended GRP method under picture fuzzy environment,
[11] M. Bani-Doumi, J. Serrano-Guerrero, F. Chiclana, et al., A picture fuzzy set multi Comput. Ind. Eng. 135 (2019) 1271–1285.
criteria decision-making approach to customize hospital recommendations based [38] X. Li, Y.B. Ju, D.W. Ju, et al., Multi-attribute group decision making method
on patient feedback, Appl. Soft Comput. 153 (2024) 111331. based on EDAS under picture fuzzy environment, IEEE Access 7 (2019)
[12] F.K. Gündoğdu, S. Duleba, S. Moslem, et al., Evaluating public transport service 141179-141192.
quality using picture fuzzy analytic hierarchy process and linear assignment [39] C. Tian, J.J. Peng, S. Zhang, et al., Weighted picture fuzzy aggregation operators
model, Appl. Soft Comput. 100 (2021) 106920. and their applications to multi-criteria decision-making problems, Comput. Ind.
[13] K. Kara, G.C. Yalçın, E.G. Kaygısız, et al., A picture fuzzy CIMAS-ARTASI model Eng. 137 (2019) 106037.
for website performance analysis in human resource management, Appl. Soft
[40] A. Singh, S. Kumar, Picture fuzzy Choquet integral-based VIKOR for multicriteria
Comput. 162 (2024) 111826.
group decision-making problems, Granul. Comput. 6 (2021) 587–601.
[14] J.L. Jin, F.Y. Zhu, T.J. You, Picture fuzzy tensor and its application in
[41] M.R. Seikh, U. Mandal, Some picture fuzzy aggregation operators based on
multi-attribute decision making, J. Intell. Fuzzy Systems 40 (6) (2021)
Frank 𝑡-norm and 𝑡-conorm: Application to MADM process, Informatica 45 (2021)
11995–12009.
447–461.
[15] S.Y. Deng, J.Z. Liu, X.F. Wang, The properties of fuzzy tensor and its application
[42] M.W. Lin, X.M. Li, R.Q. Chen, et al., Picture fuzzy interactional partitioned
in multiple attribute group decision making, IEEE Trans. Fuzzy Syst. 27 (3)
Heronian mean aggregation operators: An application to MADM process, Artif.
(2019) 589–597.
Intell. Rev. 55 (2022) 1171–1208.
[16] L. Chen, L.Z. Lu, Algorithms for finding oscillation period of fuzzy tensors,
[43] T. Senapati, Approaches to multi-attribute decision-making based on picture
Frontiers Artificial Intelligence Appl. 293 (2014) 51–57.
fuzzy Aczel-Alsina average aggregation operators, Comput. Appl. Math. 41 (2022)
[17] L. Chen, L.Z. Lu, Convergence algorithms of fuzzy tensors, J. Intell. Fuzzy Systems
40.
34 (3) (2018) 1769–1775.
[44] C. Tian, J.J. Peng, Z.Q. Zhang, et al., An extended picture fuzzy MULTIMOORA
[18] L. Chen, L.Z. Lu, Decomposition theorem of fuzzy tensors and its applications,
method based on Schweizer-Sklar aggregation operators, Soft Comput. 26 (2022)
J. Intell. Fuzzy Systems 36 (1) (2019) 575–581.
[19] L. Chen, Decomposition theorem of intuitionistic fuzzy tensors, Comput. Appl. 3435–3454.
Math. 39 (2020) 18. [45] S. Ashraf, N. Rehman, S. Abdullah, et al., Decision support model for the patient
[20] L. Chen, Three-value cutting tensors of intuitionistic fuzzy tensors, Soft Comput. admission scheduling problem based on picture fuzzy aggregation information
24 (2020) 18953–18958. and TOPSIS methodology, Math. Biosci. Eng. 19 (3) (2022) 3147–3176.
[21] N. Nadisic, A. Coussat, L. Cerf, Mining skypatterns in fuzzy tensors, Data Min. [46] S. Kumar, H. Garg, Some novel point operators and multiple rounds voting
Knowl. Discov. 33 (2019) 1298–1322. process based decision-making algorithm under picture fuzzy set environment,
[22] M.J. Frank, On the simultaneous associativity of 𝐹 (𝑥, 𝑦) and 𝑥 + 𝑦 − 𝐹 (𝑥, 𝑦), Adv. Eng. Softw. 174 (2022) 103274.
Aequationes Math. 19 (1979) 194–226. [47] L. Liu, X. Wu, G. Chen, Picture fuzzy interactional Bonferroni mean operators via
[23] W.S. Wang, H.C. He, Research on flexible probability logic operator based on strict triangular norms and applications to multicriteria decision making, IEEE
Frank 𝑇 ∕𝑆 norms, Acta Electron. Sin. 37 (5) (2009) 1141–1145. Trans. Fuzzy Syst. 31 (8) (2023) 2632–2644.
[24] W.H. Xu, Z. Yao, J. Wang, et al., A novel group decision-making method for [48] L. Ma, K. Jabeen, W. Karamti, et al., Aczel-Alsina power Bonferroni aggregation
interval-valued 𝑞-rung dual hesitant fuzzy information using extended power operators for picture fuzzy information and decision analysis, Complex Intell.
average operator and Frank operations, Artif. Intell. Rev. 57 (2024) 43. Syst. 10 (2024) 3329–3352.
[25] H. Garg, A new exponential-logarithm-based single-valued neutrosophic set and [49] T. Punetha, Komal, D. Pamucar, Novel picture fuzzy power partitioned hamy
their applications, Expert Syst. Appl. 238 (2024) 121854. mean operators with Dempster-Shafer theory and their applications in MCDM,
[26] S. Biswas, D. Bozanic, D. Pamucar, et al., A spherical fuzzy based decision making Artif. Intell. Rev. 57 (2024) 145.
framework with Einstein aggregation for comparing preparedness of SMES in [50] M. Singh, M. Pant, L.P. Kong, et al., A PCA-based fuzzy tensor evaluation model
quality 4.0, Facta Univ. Ser. Mech. Eng. 21 (3) (2023) 453–478. for multiple-criteria group decision making, Appl. Soft Comput. 132 (2023)
[27] A.R. Mishra, P. Rani, F. Cavallaro, et al., Assessment of sustainable wastew- 109753.
ater treatment technologies using interval-valued intuitionistic fuzzy distance [51] S.Y. Deng, J.Z. Liu, J.T. Tan, et al., A novel method based on fuzzy ten-
measure-based MAIRCA method, Facta Univ. Ser. Mech. Eng. 21 (3) (2023) sor technique for interval-valued intuitionistic fuzzy decision-making with
359–386. high-dimension data, Int. J. Comput. Intell. Syst. 12 (2019) 580–596.

738
J. Jin et al. Alexandria Engineering Journal 109 (2024) 726–739

[52] B.C. Cuong, V.H. Pham, Some fuzzy logic operators for picture fuzzy sets, in: [54] J.L. Jin, Research on Picture Fuzzy Information Theory and Its Application
2015 Seventh International Conference on Knowledge and Systems Engineering, in Multi-Criteria Decision-Making, (Ph.D. thesis), Guizhou Normal University,
KSE, IEEE, 2015, pp. 132–137. Guiyang, 2022.
[53] G.W. Wei, Picture fuzzy aggregation operators and their application to multiple [55] X.T. Wang, E. Triantaphyllou, Ranking irregularities when evaluating alternatives
attribute decision making, J. Intell. Fuzzy Systems 33 (2017) 713–724. by using some ELECTRE methods, Omega 36 (2008) 45–63.

739

View publication stats