Professional Documents
Culture Documents
https://doi.org/10.1016/j.engappai.2023.106255
Received 9 December 2022; Received in revised form 20 March 2023; Accepted 30 March 2023
Available online 6 April 2023
0952-1976/© 2023 Elsevier Ltd. All rights reserved.
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
another approach to study randomness in experts’ evaluations based scale QFD and extended TODIM (Liu et al., 2022), extended MARCOS
on fuzzy sets and probability theory. for decision analysis under 2tL q-rung picture fuzzy (Akram et al.,
Due to the objectivity of rough numbers in information assessment, 2023c), decision making based on CRITIC-EDAS under Pythagorean
most of the other models such as fuzzy numbers, interval numbers, 2tL fuzzy information (Akram et al., 2023d), decision making based 2tL
model and CM theory are integrated with rough numbers to handle Fermatean fuzzy sets (Akram et al., 2023e), the perception of complex
information aggregation in DM problems. Li et al. (2019b,a) converted fuzzy sets under rough approximations to introduce the novel concept
rough numbers into clouds as well as interval clouds and integrated of rough complex fuzzy sets (Sarwar et al., 2023a), risk assessment
with TOPSIS method to study HCWM in different real-world applica- based on the novel notion of rough fuzzy integrated clouds (Sarwar
tions. Jianxing et al. (2021) integrated CM theory with VIKOR approach et al., 2023b), decision making based on rough approximations and
to minimize the linguistic uncertainties during the evaluation process cloud theory (Sarwar, 2020, 2022).
using both randomness and fuzziness, and analysed risk assessment
Knowledge gaps and motives of the proposed study: The existing
problem of submarine pipelines. Gong et al. (2021) proposed a CM
researches in literature based on various hybrid mathematical models
theory based multi-objective portfolio selection approach with vari-
including integration of fuzziness and rough approximations, fuzziness
able risk appetite, which consolidates objectives of skewness, variance,
and 2tL information, roughness and cloud models and fuzziness and
mean, and liquidity restrained by several realistic constraints.
cloud models to handle different types of uncertainties. However, these
Zhu et al. (2020) introduced FR (fuzzy rough) numbers and pro-
integrated cannot be utilized to handle multi-granular information with
posed and an integrated FR-AHP-TOPSIS method for design concept
roughness in the presence of randomness. To overcome the shortcom-
analysis. Chen and Ming (2020) proposed RF (rough fuzzy) numbers
ings in existing fuzzy based and rough set theory based methods, this
and integrated RF numbers with BWM and DEA (data envelopment
research article proposes a hybrid CM theory, 2tL setting and rough
analysis) methods for smart product service module. Huang et al.
approximations based HCWM method. The novel model proposed is
(2021a,b) and Xiao et al. (2021) introduced three different powerful
known as dual 2-tuple linguistic rough number (D2tLRN) clouds or dual
models IRI (interval rough integrated) clouds, dual IRI clouds and CR
2-tuple linguistic rough integrated (D2tLRN) clouds, and integrated
(cloud rough) numbers for candidate design schemes, remanufacturing
with TOPSIS method for the selection of medical disposal technologies.
system tools and assessment method of engineering characteristics. Rao
This method deals with randomness, uncertainty and multi-granularity
et al. (2022) used dual Z-numbers and converted then into dual clouds
simultaneously for assessment aggregation in HCWM problems. The
for sustainable suppliers selection using grey relational degree and
main characteristics of proposed method are as follows:
weighted averaging operators. Huang et al. (2022) introduced Z-CR
numbers and integrated with BWM (best–worst method) and MABAC 1. Most of the existing HCWM approaches based on fuzzy numbers
(multi-attributive border approximation area comparison) for design are used to deal with vagueness and uncertain evaluations of ex-
concept evaluation. Wu et al. (2017) studied CM theory with 2tL setting perts. But, in these methods, much of the vagueness such that the
for the assessment of wind farm location selection. Wan et al. (2017) selection of pre-defined functions and the rational bias between
discussed the importance of 2tL information in supplier selection us- experts is ignored. Rough number is an objective mathematical
ing ANP (analytic network process) and ELECTRE-II (Elimination and model which is computed on the given assessment data-set with-
Choice Translating Reality II). out any additional information, suppositions and pre-defined
The researchers of different domains are actively working in such parameters. Due to objective benefits of rough numbers, upper
DM problems, for instance, decision analysis based on 2tL m-polar and lower approximations of fuzzy numbers are computed to
fuzzy sets (Akram et al., 2023b), spherical fuzzy sets (Akram et al., determine RF numbers for aggregating experts’ evaluations in
2022b), overcoming biomass energy barriers (Irfan et al., 2022), green DM problems.
low-carbon port evaluation based on weighted distances and TOPSIS 2. Rough numbers have been successfully implemented to manip-
method (Yang et al., 2022), 2tL t-spherical fuzzy model (Akram et al., ulate and study interpersonal uncertainty due to its strength for
2022a), cubic Fermatean fuzzy sets based MARCOS approach (Rong information aggregation in experts’ evaluations into intervals.
et al., 2022), Fermatean hesitant fuzzy sets based VIKOR method (Mishra However, it cannot deal with multi-granularity in experts’ judge-
et al., 2022), automatic manufacturing process based rough ELECTRE-II ments, while the initial data is mostly given by a team of DMs
approach (Sarwar et al., 2021), decision making based on 2-tuple spher- who use linguistic sets for the evaluation. To manipulate the
ical fuzzy linguistic sets (Khan et al., 2023) and Hamacher aggregation concepts of multi-granularity and interpersonal uncertainties,
operators integrated with 2-tuple linguistic Fermatean fuzzy informa- simultaneously, 2tL variables are integrated with upper and
tion (Akram et al., 2022c), integrated SWARA-WASPAS based on spher- lower approximation limits to compute a hybrid technique called
ical fuzzy sets for site selection of medical wastage (Ghoushchi et al., 2-tuple linguistic rough (2tLR) numbers.
2021), spherical fuzzy BCM for waste management problems (Haseli 3. Due to different experience and knowledge of DMrs, the random-
and Ghoushchi, 2022), extended MACBETH method based on rough ness in evaluations or judgements of experts is not considered
numbers for traffic management (Gokasar et al., 2023), interval 2- in fuzzy based and rough number based HCWM approaches.
tuple linguistic information (Wang et al., 2015), LAAW (Logarithmic Cloud theory based on probability theory and fuzzy sets is an
Additive Assessment of the Weight coefficients) and RAFSI (Ranking AI approach to reflect both randomness and fuzziness in hu-
of Alternatives through Functional mapping of criterion subintervals man knowledge concepts. It handles the issue that degrees of
into a Single Interval) model based on rough number for site selec- membership in fuzzy sets are accurate by allowing a stochastic
tion problems (Deveci et al., 2022b), rough macbeth and rafsi based disturbance of membership values. In the proposed method, 2tLR
decision support system for flight base selection (Akyurt et al., 2021), numbers are converted into 2tL clouds and the new hybrid model
ordinal priority approach for sustainable development (Deveci et al., is known as dual 2-tuple linguistic rough number (D2tLRN) clouds.
2022a), WASPAS method based on 2tL Fermatean fuzzy information 4. The disposal of medical wastes (causing environmental distur-
for solid waste disposal management (Akram et al., 2023a), extended bance and pollution) has always been an active domain of
PROMETHEE approach based on 2tL m-polar fuzzy information in research using different mathematical models. Different types
biosciences (Akram et al., 2022d), generalized divergence based de- of disposal alternatives including Landfill, microwaves, steam
cision making (Xiao et al., 2022b), hybrid models based on type-2 sterilization and incineration are used to dispose the medical
soft information (Sarwar and Akram, 2023), generalized evidential waste. One disposal technology cannot be used in each area, lo-
divergence measure (Xiao, 2022a), generalized quantum evidence the- cality and environment. Therefore, different HCWM approaches
ory (Xiao, 2022c), integrated behavioural decision analysis based large have been proposed to select a suitable treatment technology.
2
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
3
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
1 1
2tL variable. There exists a reverse function 𝛥−1 ∶ 𝑋 ×[− 2𝑚 , 2𝑚
) → [0, 1] 2.3. Cloud model theory
used to covert a 2tL term (𝑥𝑗 , 𝜉) into an equivalent numerical value 𝜂
Cloud theory is a machine learning approach that can analyse
as defined in Eq. (2).
uncertainties, that is, fuzziness and randomness, in human knowledge.
𝑗 This concept was first initiated by Li et al. (1995a) based on probability
𝛥−1 (𝑥𝑗 , 𝜉) = 𝜉 + = 𝜂. (2)
𝑚 theory and fuzzy set theory. In a fuzzy set, each object is given a
membership value to represent its degree of uncertainty. However, it is
Example 2. Consider a linguistic term set 𝑋 = {𝑥𝑜 , 𝑥1 , 𝑥2 , 𝑥3 , 𝑥4 , 𝑥5 , 𝑥6 }, against the spirit of fuzzy set theory, where the uncertainty of an object
1 1
then clearly 𝑚 = 6 and 𝜉 ∈ [− 12 , 12 ). By Eq. (2), 𝛥−1 (𝑥𝑜 , 0) = 0 + 06 , in a given fuzzy set is precise and certain. CM theory deals with this
𝛥 (𝑥3 , 0.03) = 0.405, 𝛥 (𝑥1 , 0.02) = 0.145, 𝛥−1 (𝑥5 , −0.01) = 0.615 and
−1 −1
issue allowing a stochastic disturbance of membership degree around
𝛥−1 (𝑥2 , 0.04) = 0.290. a determined central value (Li et al., 2009).
Similarly, if the outcome of aggregation operation is 𝜂 = 0.175, the Assume that 𝑄 is a qualitative concept defined over the universe 𝑌
representation of this information as 2tL term is 𝛥(0.175) = (𝑥1 , 0.05) and 𝑧 ∈ 𝑌 . 𝜇𝑄 (𝑧) ∈ [0, 1] is the membership degree of 𝑧 which is a
and 𝛥(0.480) = (𝑥4 , −0.02). random number rather than a fixed number and follows a probability
distribution. The distribution on 𝑧 in 𝑌 is called a cloud and 𝑧 is known
as cloud drop. It is clear that the hybrid RF set theory can deal with
Definition 3 (Herrera and Martínez, 2000). Let (𝑥𝑎 , 𝜉1 ) and (𝑥𝑏 , 𝜉2 ) be
uncertainty and ambiguities in experts’ judgements but it cannot handle
2tL terms, then
the randomness in fuzzy values and evaluations of DMrs. CM theory can
1. If 𝑎 < 𝑏, then (𝑥𝑎 , 𝜉1 ) < (𝑥𝑏 , 𝜉2 ); efficiently deal with the randomness in judgement intervals. Generally,
2. If 𝑎 > 𝑏, then (𝑥𝑎 , 𝜉1 ) > (𝑥𝑏 , 𝜉2 ); a cloud is represented as (𝐸𝑥, 𝐸𝑛, 𝐻𝐸𝑛), where
3. If 𝑎 = 𝑏, then (1) The expectation 𝐸𝑥 represents the qualitative concept of the
element in the domain.
(a) if 𝜉1 = 𝜉2 , then (𝑥𝑎 , 𝜉1 ) = (𝑥𝑏 , 𝜉2 ); (2) The entropy 𝐸𝑛 describes the randomness of the concept. It
(b) if 𝜉1 < 𝜉2 , then (𝑥𝑎 , 𝜉1 ) < (𝑥𝑏 , 𝜉2 ); measures the range and dispersion extent accepted by that qualitative
(c) if 𝜉1 > 𝜉2 , then (𝑥𝑎 , 𝜉1 ) > (𝑥𝑏 , 𝜉2 ). concept.
(3) The hyper entropy 𝐻𝐸𝑛, entropy of 𝐸𝑛, shows the uncertainty
of the membership degrees.
2.2. Rough numbers
3. The proposed dual 2-tuple linguistic rough number clouds
1 ∑
𝜉𝑗 = 𝑢 𝜉 (5) Definition 4. Let 𝛺 be the universe containing all 2tL assessment
𝑛
𝜉∈𝑎𝑝𝑟(𝜉𝑗 ) values given by DMrs. The assessment values can be classified into
1 ∑ 𝑚 classes 𝛿1 , 𝛿2 , … , 𝛿𝑚 in ascending order 𝛿1 ≤ 𝛿2 ≤ ⋯ ≤ 𝛿𝑚 , where
𝜉 = 𝜉 (6)
𝑗 𝑛𝑙 𝛿𝑗 = (𝑥𝑗 , 𝜉𝑗 ) is the 𝑗th assessment value from DMrs, 𝑥𝑗 ∈ 𝑋 describes
𝜉∈𝑎𝑝𝑟(𝜉𝑗 )
the central value of the 𝑗th linguistic term and 𝜉𝑗 ∈ [−0.5, 0.5) represents
where 𝑛𝑘𝑈 𝑙 and 𝑛𝑢 denotes the number of elements in the upper the deviation of the 𝑗th linguistic term from the central value. The
approximations 𝑎𝑝𝑟(𝜉𝑗 ) and 𝑛𝑙 denotes the number of elements in the upper and lower approximations of 𝛿𝑗 = (𝑥𝑗 , 𝜉𝑗 ) can be computed using
lower approximation 𝑎𝑝𝑟(𝜉𝑗 ). The rough number of value 𝜉𝑗 is written Eqs. (7) and (8).
{ }
as 𝑅𝑁(𝜉𝑗 ) = [𝜉 , 𝜉 𝑗 ]. 𝑎𝑝𝑟(𝑥𝑗 , 𝜉𝑗 ) = ∪ (𝑥, 𝜉) ∣ (𝑥, 𝜉) ≥ (𝑥𝑗 , 𝜉𝑗 ), (𝑥, 𝜉) ∈ 𝛺 (7)
𝑗
4
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
{ }
𝑎𝑝𝑟(𝑥𝑗 , 𝜉𝑗 ) = ∪ (𝑥, 𝜉) ∣ (𝑥, 𝜉) ≤ (𝑥𝑗 , 𝜉𝑗 ), (𝑥, 𝜉) ∈ 𝛺 (8) Table 3
Algebraic operations on 2tL cloud model.
The upper and lower limits 𝛿 𝑗 = (𝑥𝑗 , 𝜉𝑗 ) and 𝛿 𝑗 = (𝑥𝑗 , 𝜉𝑗 ) of 𝛿𝑗 = (𝑥𝑗 , 𝜉𝑗 ) Operation Expectation Entropy Hyper entropy
( ) √ √
are obtained in Eqs. (9) and (10). 𝐶1 + 𝐶2 𝛥 𝛥−1 𝛼1 + 𝛥−1 𝛼2 𝐸𝑛21 + 𝐸𝑛22 𝐻𝐸𝑛21 + 𝐻𝐸𝑛22
( )
⎛ ⎞ 𝐶1 × 𝐶2 𝛥 𝛥−1 𝛼1 × 𝛥−1 𝛼2 |(𝛥−1 𝛼1 )(𝛥−1 𝛼2 )| |(𝛥−1 𝛼1 )(𝛥−1 𝛼2 )|
1 ∑ √( √(
𝛿𝑗 = 𝛥 ⎜ 𝑢 𝛥−1 (𝑥, 𝜉)⎟ (9) )2 ( )2 ) 2 ( )2
⎜𝑛 ⎟ ×
𝐸𝑛1 𝐸𝑛
+ 𝛥−1 𝛼2 ×
𝐻𝐸𝑛1
+
𝐻𝐸𝑛2
⎝ (𝑥,𝜉)∈𝑎𝑝𝑟(𝑥𝑗 ,𝜉𝑗 ) ⎠ ( )
𝛥−1 𝛼
1 2 𝛥−1 𝛼1 𝛥−1 𝛼2
̃ 𝑗 )− 𝐻𝐴(𝐶1 , 𝐶2 , … , 𝐶𝑟 ) = 𝛾𝑗 𝐶𝑗 (16)
resenting a scale belong to the interval 𝛥−1 (𝑥𝐸𝑥𝑗 , 𝐸𝑥
] 𝑗=1
−1 ̃ −1
3𝐸𝑛𝑗 , 𝛥 (𝑥𝐸𝑥𝑗 , 𝐸𝑥𝑗 ) + 3𝐸𝑛𝑗 . Then 3𝐸𝑛𝑗 = (𝛥 (𝑥𝑗 , 𝜉𝑗 ) − Here [𝛾1 , 𝛾2 , … , 𝛾𝑟 ] is the weight vector satisfying 𝛾𝑗 ∈ [0, 1] and
∑𝑟
𝛥−1 (𝑥𝑗 , 𝜉𝑗 ))∕2 can be computed, where (𝑥𝑗 , 𝜉𝑗 ) and (𝑥𝑗 , 𝜉𝑗 ) 𝑗=1 𝛾𝑗 = 1. The weight vector can be computed using Eqs. (17) and
(18).
are the upper and lower boundaries of the determined ( ) ( )
2tLRN. Thus, the entropy value of 2tL cloud can be com- 𝑗 𝑗−1
𝛾𝑗 = ℎ −ℎ (17)
puted using Eq. (12). 𝑟 𝑟
⎧0, 𝑥<𝛼
𝛥−1 (𝑥𝑗 , 𝜉𝑗 ) − 𝛥−1 (𝑥𝑗 , 𝜉𝑗 ) ⎪ 𝑥−𝛼
𝐸𝑛𝑗 = (12) ℎ(𝑥) = ⎨ 𝛽−𝛼 , 𝛼≤𝑥≤𝛽 (18)
6 ⎪0, 𝑥>𝛽
⎩
• Compute hyper entropy 𝐻𝐸𝑛𝑗
The parameters 𝛼 and 𝛽 are evaluated by DMrs and can be varied with
Hyper entropy 𝐻𝐸𝑛 represents the uncertainty of entropy the situation. In this research study, we will consider 𝛼 = 0.15 and
𝐸𝑛, computed from 𝐸𝑛. According to the theory presented 𝛽 = 0.9.
by Zhang et al. (2020)( about cloud droplets,) we can say
̃
that a 2tL cloud 𝐶 = (𝑥𝐸𝑥 , 𝐸𝑥)𝐸𝑥, 𝐸𝑛, 𝐻𝐸𝑛 also follows 4. The proposed D2tLRN cloud HCWM methodology
a non-strict normal distribution if 3𝐻𝐸𝑛 < 𝐸𝑛, otherwise it
is fogged if 3𝐻𝐸𝑛 > 𝐸𝑛. Hence, for any 2tL cloud 𝐶𝑗 , 𝐻𝐸𝑛𝑗 In this section, an improved HCWM methodology for waste man-
can be computed using the relation 𝐻𝐸𝑛𝑗 = 𝛼 × 𝐸𝑛𝑗 (0 < agement based on D2tLRN cloud model and extended TOPSIS method
𝛼 < 1∕3). is illustrated in detail.
5
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
1. Firstly, the possible waste factors and treatment technologies are the HCW treatment alternatives. The classical HCWM schemes utilizes
determined by the DMrs. An improved twofold waste factor system is four main waste factors: Eco (economic), Envi (environmental), Tech
designed at this stage. (technical) and Soc (social). This approach is not rational as it cannot
2. Secondly, the 2tL evaluations from DMrs are collected and then comprehensively and systematically examine the waste treatments and
transformed into D2tLRN clouds. Cloud model has the ability to de- ignores various sub-levels that effect the waste factors. This study uses
scribe the relation between qualitative concept and qualitative values a twofold waste factor hierarchical system based on four waste factors
considering both randomness and fuzziness in the data. Meanwhile, the Eco, Envi, Tech and Soc as shown in Fig. 2.
2tL variables are utilized to express uncertain knowledge DMrs have In this hierarchical system, Eco, Envi, Tech and Soc are level-
about the problem. Considering the advantages of both theories along 1 waste factors and level-2 contains eight waste factors. The waste
with rough set theory, this research study proposes rough integrated factor Eco considers net cost per heap (𝑊 𝑓1 ) as its level-2 factor,
clouds under 2tL information. while factor Envi is about the effects of the target HCW treatment
3. Thirdly, an improved dynamic weighting scheme is adopted technology on the atmosphere and contains three level-2 factors: waste
considering the confidence level of DMrs to compute the evaluation residuals (𝑊 𝑓2 ), noise (𝑊 𝑓3 ) and health effects (𝑊 𝑓4 ). The factor Tech
weights using HA operator. The importance of waste factors is calcu- is sub-divided into three level-2 factors: reliability (𝑊 𝑓5 ), treatment
lated by integrated approach considering both objective and subjective effectiveness (𝑊 𝑓6 ) and occupational hazard (𝑊 𝑓7 ). The treatment
aspects of uncertainty. The 2tL values converted from DMrs’ evalu- level of the waste is reduced if the disposal technologies and waste
ations corresponding to each waste factor from multiple assessment factors are easily identified. The level-1 factor Soc relates and refers
matrix is exactly where the TOPSIS method can utilize its superiority to the hardship of maintenance or repair of public property after the
into full play. Finally, the proposed D2tLRN cloud TOPSIS approach is management and treatment of waste. This is the reason Soc takes
implemented to compute the benefit level of treatment technology. The one sub-factor: public acceptance (𝑊 𝑓8 ). This twofold hierarchy can
procedure of the proposed approach is illustrated in this section. be successfully applied to various HCWM methods. It also provides a
The proposed D2tLRN cloud HCWM approach is explained in five criteria for the decision analysis of machines for the disposal of medical
different stages as shown in Fig. 1. wastes which is the main focus of this research study.
Defining waste factors and treatment technologies Collection and transformation of 2tL information into D2tLRN clouds
The first step of HCWM method is to determine form the team of In this step, the given 2tL evaluation data is converted into 2tL
experts (DMrs) whose practical experience, knowledge is relevant to rough numbers (2tLRNs) and then further transformed into 2tL clouds.
the target problem. Then the development of D2tLRN cloud TOPSIS Then the 𝑠 matrices containing 2tL cloud values, corresponding to 𝑠
method begins with the determination of important waste factor and DMrs, are aggregated into a single matrices using HA operator.
treatment or disposal technologies.
Collection of 2tL information: The DMrs express their judgements of
Make the team of DMrs: Due to the uncertainty of the problem and evaluation objects using 2tL terms to provide effective and reasonable
limited knowledge and experience of single DM, a team of experienced assessment data. During this process, the DMrs evaluate the disposal
experts is needed to fully capture the problem and improve the accu- technologies for each waste factor based on a questionnaire survey.
racy and objectivity of the results. The team of DMrs can be established For a certain assessment problem, assume that there are 𝑝 treatment
based on three main principles. technologies 𝑇 𝑇1 , 𝑇 𝑇2 , … , 𝑇 𝑇𝑝 , 𝑟 twofold waste factors 𝑊 𝑓1 , 𝑊 𝑓2 , … ,
1. Experience and capability of DMrs: The members of the team 𝑊 𝑓𝑟 and 𝑠 DMrs 𝑑1 , 𝑑2 , … , 𝑑𝑠 . Convert the 2tL terms into 2tL variables
should have at least five years experience of working in scientific re- using given criteria like Table 2. For 𝑗th treatment technology and 𝑘th
search projects, testing, design, maintenance, installation and construc- waste factor, the 2tL assessment value given by 𝑖th DMr can be denoted
tion in the relevant field. They must be capable of current operations 𝑖 = (𝑥𝑖 , 𝜉 𝑖 ), where 𝑥𝑖 refers to central linguistic term of the
as 𝛿𝑗𝑘 𝑗𝑘 𝑗𝑘 𝑗𝑘
and design of target system to make reasonable judgements (Cai et al., assessment value, 𝜉𝑗𝑘 𝑖 ∈ [−0.5, 0.5] represents the deviation from the
2013). central value 𝑥𝑖𝑗𝑘 . The DMrs can use more than one linguistic sets to
2. Number of DMrs: The numbers of DMrs in the team should not be express their evaluations, it will effect in the aggregation process to
very small to ensure the rationality and reliability of the results. Usu- reflect the efficacy of assessment by the DMrs. The 2tL initial relation
ally, three to ten members are selected to provide credible judgements matrix by 𝑖th DM, for 𝑖 = 1, 2, … , 𝑠, is given in Eq. (19).
in the waste management problem (Yazdi et al., 2020; Zarei et al.,
𝑖 𝑖 (𝑥𝑖12 , 𝜉12
𝑖 ) (𝑥𝑖1𝑟 , 𝜉1𝑟
𝑖 )
2021). ⎡(𝑥11 , 𝜉11 ) … ⎤
⎢ 𝑖 𝑖 𝑖 )⎥
3. Diversity of composition structure: To avoid extreme convergence (𝑥 , 𝜉 ) (𝑥𝑖22 , 𝜉22
𝑖 ) … (𝑥𝑖2𝑟 , 𝜉2𝑟
2𝑡𝐿𝑖 = ⎢ 21 21 ⎥ (19)
of opinions, a diverse team must be organized considering the het- ⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢ 𝑖 𝑖 𝑖 )⎥
erogeneous expert composition. The DMrs with different knowledge, ⎣(𝑥𝑝1 , 𝜉𝑝1 ) (𝑥𝑖𝑝2 , 𝜉𝑝2
𝑖 ) … (𝑥𝑖𝑝𝑟 , 𝜉𝑝𝑟 ⎦
experience, position and medical waste disposal analysis hold different
point of views on the same issue which is beneficial to improve the Obtain 2tLRN relation matrix by converting 2tL variables into 2tL
efficacy of the evaluations. rough numbers: The data collected by DMrs judgements contain un-
certain and ambiguous information due to diversity of knowledge and
Identify the possible treatment technologies: The purpose HCWM
experience. Rough approximations can extract{and cope with uncer-
team is to identify the currently used HCWM treatment technologies 𝑖 = (𝑥𝑖 , 𝜉 𝑖 ) ∣ 𝑖 = 1,
based on target area, practical experiences, data from environmental tainty without additional parameters. Let 𝛺𝑗𝑘 = 𝛿𝑗𝑘
} 𝑗𝑘 𝑗𝑘
protection bureau and the company responsible for collecting and 2, … , 𝑠 be the universal set consisting of all 2tL assessment values
cleaning the waste. In this step, the potential treatment technologies for 𝑗th treatment technology under 𝑘th level-2 waste factor, where 𝑥𝑖𝑗𝑘
in the waste management system are determined using HCWM team and 𝜉𝑗𝑘 𝑖 are the central linguistic variable and deviation number of
analysis results. 2tL assessment value. The upper and lower approximations of 𝛿𝑗𝑘 𝑖 =
𝑖 𝑖
(𝑥𝑗𝑘 , 𝜉𝑗𝑘 ) can be computed using Eqs. (20) and (21).
Analysis and identification of waste factors: Usually, various quali-
tative and quantitative criteria are considered in analysing the suitable { }
HCW disposal technology. Based on HCWM team reviews, system- 𝑎𝑝𝑟(𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖
) = ∪ (𝑥, 𝜉) ∣ (𝑥, 𝜉) ≥ (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖
), (𝑥, 𝜉) ∈ 𝛺𝑗𝑘 (20)
atic literature review, social, technical, environmental and economic { }
𝑎𝑝𝑟(𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖
) = ∪ (𝑥, 𝜉) ∣ (𝑥, 𝜉) ≤ (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖
), (𝑥, 𝜉) ∈ 𝛺𝑗𝑘 (21)
dimensions, the relevant waste criteria are determined to evaluate
6
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
𝑖
[ ]
𝑖 ) and 𝛿 𝑖
The upper and lower limits 𝛿 𝑗𝑘 = (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘 = (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖 ) of
𝑗𝑘
𝑖 ), (𝑥𝑖 , 𝜉 𝑖 ) . The 2tLR (2tL rough) relation matrix for 𝑖th DMr
(𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘 𝑗𝑘 𝑗𝑘
𝑖 = (𝑥𝑖 , 𝜉 𝑖 ) are obtained in Eqs. (22) and (23).
𝛿𝑗𝑘 𝑗𝑘 𝑗𝑘 is written in Eq. (24).
⎛ ⎞
𝑖 ⎜1 ∑ ⎟ ⎡[(𝑥𝑖 , 𝜉 𝑖 ), (𝑥𝑖 , 𝜉 𝑖 )] [(𝑥𝑖12 , 𝜉12
𝑖
), (𝑥𝑖12 , 𝜉12
𝑖
)] … [(𝑥𝑖1𝑟 , 𝜉1𝑟
𝑖
), (𝑥𝑖1𝑟 , 𝜉1𝑟
𝑖
)]⎤
−1
𝛿 𝑗𝑘 = 𝛥⎜ 𝑢 𝛥 (𝑥, 𝜉)⎟ (22) ⎢ 11 11 11 11 ⎥
⎜𝑛 (𝑥,𝜉)∈𝑎𝑝𝑟(𝑥𝑖𝑗𝑘 ,𝜉𝑗𝑘
𝑖 ) ⎟ ⎢ [(𝑥𝑖 , 𝜉 𝑖 ), (𝑥𝑖 , 𝜉 𝑖 )] [(𝑥𝑖22 , 𝜉22
𝑖
), (𝑥𝑖1𝑟 , 𝜉1𝑟
𝑖
)] … [(𝑥2𝑟 , 𝜉2𝑟 ), (𝑥1𝑟 , 𝜉1𝑟 )]⎥
𝑖 𝑖 𝑖 𝑖
⎝ ⎠ 2𝑡𝐿𝑅𝑖 = ⎢ 21 21 1𝑟 1𝑟 ⎥
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎛ ⎞ ⎢ ⎥
⎜1 ∑ ⎟ ⎢ [(𝑥𝑖𝑝1 , 𝜉𝑝1
𝑖
), (𝑥𝑖1𝑟 , 𝜉1𝑟
𝑖
)] [(𝑥𝑖𝑝2 , 𝜉𝑝2
𝑖
), (𝑥𝑖1𝑟 , 𝜉1𝑟
𝑖
)] … 𝑖 ), (𝑥𝑖 , 𝜉 𝑖 )]⎥
[(𝑥𝑖𝑝𝑟 , 𝜉𝑝𝑟
𝛿 𝑖𝑗𝑘 = 𝛥⎜ 𝑙 −1
𝛥 (𝑥, 𝜉)⎟ (23) ⎣ 𝑝𝑟 𝑝𝑟 ⎦
⎜𝑛 (𝑥,𝜉)∈𝑎𝑝𝑟(𝑥𝑖𝑗𝑘 ,𝜉𝑗𝑘
𝑖 ) ⎟ (24)
⎝ ⎠
where 𝑛𝑢 denotes the number of elements in the upper approximation
𝑎𝑝𝑟(𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖 ) and 𝑛𝑙 denotes the number of elements in the lower ap- Compute 2tL cloud evaluations of 2tLR relation matrix: For detailed
analysis, the obtained 2tLR evaluations are further converted into 2tL
proximation 𝑎𝑝𝑟(𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖 ). Note that the upper and lower limits (𝑥𝑖 , 𝜉 𝑖 )
𝑗𝑘 𝑗𝑘 clouds to discuss randomness and fuzziness effectively in the assess-
and (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖 ) are also 2tL terms. The 2tL rough number (2tLRN)
ment values. The obtained 2tL clouds are known as D2tLRN( clouds.
𝑖 ̃ 𝑖 ),
𝑖 ) is written as 2𝑡𝐿𝑅𝑁(𝑥𝑖 , 𝜉 𝑖 ) = [𝛿 𝑖 , 𝛿 ] =
of 2tL value (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘 A 2tL cloud of 2𝑡𝐿𝑅𝑁(𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖 ) is represented as 𝑖 = (ℎ
𝑗𝑘 𝐸𝑥 𝑖 , 𝐸𝑥 𝑗𝑘
𝑗𝑘 𝑗𝑘 𝑗𝑘 𝑗𝑘 𝑗𝑘
7
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
( ) ( ) ( )
⎡ (𝑥 𝑖 , 𝜉 𝑖 ), 𝐸𝑛𝑖 , 𝐻𝐸𝑛𝑖 (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖12 , 𝐻𝐸𝑛𝑖12 … (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖1𝑟 , 𝐻𝐸𝑛𝑖1𝑟 ⎤
⎢( 𝐸𝑥11 𝐸𝑥11 11 11
) ( 12 12
) ( 1𝑟 1𝑟
)⎥⎥
⎢
(𝑥 ,𝜉 ), 𝐸𝑛𝑖21 , 𝐻𝐸𝑛𝑖21 (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖22 , 𝐻𝐸𝑛𝑖22 … (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖2𝑛 , 𝐻𝐸𝑛𝑖2𝑛 ⎥
𝐷2𝑡𝐿𝑅𝑁𝐶 = ⎢ 𝐸𝑥𝑖21 𝐸𝑥𝑖21
𝑖
22 22 2𝑟 2𝑟 (27)
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢( ) ( ) ( )⎥
⎢ (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖 , 𝐻𝐸𝑛𝑖 (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖𝑝2 , 𝐻𝐸𝑛𝑖𝑝2 … (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑝𝑟 , 𝐻𝐸𝑛𝑝𝑟 ⎥
𝑖 𝑖
⎣ 𝑝1 𝑝1 𝑝1 𝑝1 𝑝2 𝑝2 𝑝𝑟 𝑝𝑟 ⎦
Box I.
) ( )
̃ 𝑖 ), 𝐸𝑛𝑖 and 𝐻𝐸𝑛𝑖 are the 2tL expec-
𝐸𝑛𝑖𝑗𝑘 , 𝐻𝐸𝑛𝑖𝑗𝑘 , where (ℎ𝐸𝑥𝑖 , 𝐸𝑥 (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖𝑗𝑘 , 𝐻𝐸𝑛𝑖𝑗𝑘 , 1 ≤ 𝑖 ≤ 𝑠 are aggregated using For-
𝑗𝑘
𝑗𝑘 𝑗𝑘 𝑗𝑘 𝑗𝑘 𝑗𝑘
tation, entropy and hyper entropy of 2tL cloud 𝐶𝑗𝑘 𝑖 . Based on Section 3,
[ ] mula (28).
𝑖 ), (𝑥𝑖 , 𝜉 𝑖 ) is the translation
the 2tL expectation of 2tLRN (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘 𝑗𝑘 𝑗𝑘 ∑
𝑠
1 2 𝑠 𝑠
𝑗𝑘 = 𝐻𝐴(𝑗𝑘 , 𝑗𝑘 , … , 𝑗𝑘 )= 𝛾𝑖 𝑗𝑘 (28)
function value of the average of reverse function values of upper and 𝑖=1
lower limits as illustrated in Eq. (25). Here [𝛾1 , 𝛾2 , … , 𝛾𝑠 ] is the weight vector satisfying 𝛾𝑖 ∈ [0, 1] and
∑𝑠
𝑖=1 𝛾𝑖 = 1. The weight vector can be computed using Eqs. (17)
⎛ 𝛥−1 (𝑥𝑖 , 𝜉 𝑖 ) + 𝛥−1 (𝑥𝑖 , 𝜉 𝑖 ) ⎞ and (18). These weight vectors are unique and can be determined
̃ 𝑖 ⎜ 𝑗𝑘 𝑗𝑘 𝑗𝑘 𝑗𝑘 ⎟
(𝑥𝐸𝑥𝑖 , 𝐸𝑥𝑗𝑘 ) = 𝛥 ⎜ ⎟ (25) by different methods. The aggregated(D2tLRN cloud relation matrix )
𝑗𝑘
⎜ 2 ⎟ [ ]
⎝ ⎠ 𝐷2𝑡𝐿𝑅𝑁𝐶 = 𝑗𝑘 𝑝×𝑟 , where 𝑗𝑘 = (𝑥𝐸𝑥𝑗𝑘 , 𝜉𝐸𝑥𝑗𝑘 ), 𝐸𝑛𝑗𝑘 , 𝐻𝐸𝑛𝑗𝑘 , is
written as Eq. (29) given in Box II.
The entropy value of 2tL cloud can be computed using Eq. (26).
𝑖 ) − 𝛥−1 (𝑥𝑖 , 𝜉 𝑖 )
𝛥−1 (𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘 𝑗𝑘 𝑗𝑘
Compute hybrid weights of waste factors
𝐸𝑛𝑖𝑗𝑘 = (26)
6
To compute the hybrid importance coefficients of waste factors,
Hyper entropy 𝐻𝐸𝑛 reflects the vagueness of entropy value 𝐸𝑛, so the objective, subjective and hybrid importance coefficients of weight
it can be computed from 𝐸𝑛. As described in Section 3, for any 2tL factors are computed based on the importance of factors evaluated by
𝑖 , hyper entropy 𝐻𝐸𝑛𝑖 can be obtained using the relation
cloud 𝑗𝑘 𝑗𝑘 DMrs. The team of DMrs give importance to waste factors in the form of
𝐻𝐸𝑛𝑖𝑗𝑘 = 𝛼 × 𝐸𝑛𝑖𝑗𝑘 (0 < 𝛼 < 1∕3). The D2tLRN cloud relation matrix 2tL values described in Table 2. The 2tL relative importance of 𝑟 waste
[ ] ( )
𝐷2𝑡𝐿𝑅𝑁𝐶 𝑖 = 𝑗𝑘 𝑖 𝑖 = (𝑥
, where 𝑗𝑘 𝑖 𝑖 factors provided by the DMrs is given in Eq. (30).
𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑗𝑘 , 𝐻𝐸𝑛𝑗𝑘 , for
𝑝×𝑟 𝑗𝑘 𝑗𝑘
( )
𝑖th DMr is written as Eq. (27) given in Box I. (𝑥𝑘 , 𝜉𝑘 ) = (𝑥1𝑘 , 𝜉𝑘1 ), (𝑥2𝑘 , 𝜉𝑘2 ), … , (𝑥𝑖𝑘 , 𝜉𝑘𝑖 ), … , (𝑥𝑠𝑘 , 𝜉𝑘𝑠 ) (30)
8
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
( ) ( ) ( )
⎡ (𝑥 ,𝜉 ), 𝐸𝑛11 , 𝐻𝐸𝑛11 (𝑥𝐸𝑥12 , 𝜉𝐸𝑥12 ), 𝐸𝑛12 , 𝐻𝐸𝑛12 … (𝑥𝐸𝑥1𝑟 , 𝜉𝐸𝑥1𝑟 ), 𝐸𝑛1𝑟 , 𝐻𝐸𝑛1𝑟 ⎤
⎢( 𝐸𝑥11 𝐸𝑥11 ) ( ) ( )⎥⎥
⎢
𝐷2𝑡𝐿𝑅𝑁𝐶 = ⎢ (𝑥𝐸𝑥21
, 𝜉𝐸𝑥21
), 𝐸𝑛21 , 𝐻𝐸𝑛21 (𝑥𝐸𝑥22 , 𝜉𝐸𝑥22 ), 𝐸𝑛22 , 𝐻𝐸𝑛22 … (𝑥𝐸𝑥2𝑟 , 𝜉𝐸𝑥2𝑟 ), 𝐸𝑛2𝑛 , 𝐻𝐸𝑛2𝑛 ⎥ (29)
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢( ) ( ) ( )⎥
⎢ (𝑥𝐸𝑥 , 𝜉𝐸𝑥 ), 𝐸𝑛𝑝1 , 𝐻𝐸𝑛𝑝1 (𝑥𝐸𝑥𝑝2 , 𝜉𝐸𝑥𝑝2 ), 𝐸𝑛𝑝2 , 𝐻𝐸𝑛𝑝2 … (𝑥𝐸𝑥𝑝𝑟 , 𝜉𝐸𝑥𝑝𝑟 ), 𝐸𝑛𝑝𝑟 , 𝐻𝐸𝑛𝑝𝑟 ⎥⎦
⎣ 𝑝1 𝑝1
Box II.
[ ]
where the 2tLRN 2𝑡𝐿𝑅𝑁(𝑥𝑖𝑘 , 𝜉𝑘𝑖 ) = (𝑥𝑖𝑘 , 𝜉𝑘𝑖 ), (𝑥𝑖𝑘 , 𝜉𝑘𝑖 ) is a 2tLRN of The objective importance coefficient 𝛿𝑘𝑜 of 𝑘th weight factor can be
determined using Eq. (39).
𝑘th waste factor given by 𝑖th DM. The 2tLRNs in Eq. (31) are then
transformed into 2tL cloud evaluations as given in Eq. (32). 𝜎2
𝛿𝑘𝑜 = ∑𝑟 𝑘 . (39)
( ) 𝜎𝑘2
2𝑡𝐿𝐶(2𝑡𝐿𝑅𝑁(𝑥𝑘 , 𝜉𝑘 )) = 𝑘 = 𝑘1 , 𝑘2 , … , 𝑘𝑠 𝑘=1
(( )
= (𝑥𝐸𝑥1 , 𝜉𝐸𝑥1 ), 𝐸𝑛1𝑘 , 𝐻𝐸𝑛1𝑘 , Here 𝜎𝑘2 is the variance of aggregated 2tL cloud evaluation 𝑘 for
( 𝑘 𝑘
) each 1 ≤ 𝑘 ≤ 𝑟. The hybrid importance coefficient of 𝑘th weight
(𝑥𝐸𝑥2 , 𝜉𝐸𝑥2 ), 𝐸𝑛2𝑘 , 𝐻𝐸𝑛2𝑘 , … , factor computed using subjective and objective importance coefficients
( 𝑘 𝑘
)) is given in Formula (40).
(𝑥𝐸𝑥𝑠 , 𝜉𝐸𝑥𝑠 ), 𝐸𝑛𝑠𝑘 , 𝐻𝐸𝑛𝑠𝑘 (32) ( )
𝑘 𝑘
𝛿𝑘𝑜 × 𝛥−1 𝛿𝑘𝑠
( ) 𝛿𝑘 = 𝛥 ∑𝑟 𝑜 −1 𝑠
. (40)
where 𝑘𝑖 = (𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖𝑘 , 𝐻𝐸𝑛𝑖𝑘 is the 2tL cloud importance 𝑘=1 𝛿𝑘 × 𝛥 𝛿𝑘
𝑘 𝑘
9
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
( ) ( ) ( )
⎡ (𝑥 ,𝜉 )′ , 𝐸𝑛′11 , 𝐻𝐸𝑛′11 (𝑥𝐸𝑥12 , 𝜉𝐸𝑥12 )′ , 𝐸𝑛′12 , 𝐻𝐸𝑛′12 … (𝑥𝐸𝑥1𝑟 , 𝜉𝐸𝑥1𝑟 )′ , 𝐸𝑛′1𝑟 , 𝐻𝐸𝑛′1𝑟 ⎤
⎢( 𝐸𝑥11 𝐸𝑥11 ) ( ) ( )⎥⎥
⎢ ′ , 𝐸𝑛′ , 𝐻𝐸𝑛′
′
[𝑗𝑘 ]𝑝×𝑟 = ⎢ (𝑥𝐸𝑥21
, 𝜉𝐸𝑥21
) 21 21
(𝑥𝐸𝑥22 , 𝜉𝐸𝑥22 )′ , 𝐸𝑛′22 , 𝐻𝐸𝑛′22 … ′ ′ ′
(𝑥𝐸𝑥2𝑟 , 𝜉𝐸𝑥2𝑟 ) , 𝐸𝑛2𝑛 , 𝐻𝐸𝑛2𝑛 ⎥ (42)
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢( ) ( ) ( )⎥
⎢ (𝑥𝐸𝑥 , 𝜉𝐸𝑥 )′ , 𝐸𝑛′ , 𝐻𝐸𝑛′ (𝑥𝐸𝑥𝑝2 , 𝜉𝐸𝑥𝑝2 )′ , 𝐸𝑛′𝑝2 , 𝐻𝐸𝑛′𝑝2 … (𝑥𝐸𝑥𝑝𝑟 , 𝜉𝐸𝑥𝑝𝑟 ) , 𝐸𝑛𝑝𝑟 , 𝐻𝐸𝑛𝑝𝑟 ⎥⎦
′ ′ ′
⎣ 𝑝1 𝑝1 𝑝1 𝑝1
Box III.
10
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Table 5
2tL evaluations of treatment alternatives by five DMrs.
DMrs Alternatives Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4 𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 (𝑥1 , 0) (𝑥2.5 , 0) (𝑥1.5 , 0) (𝑥𝑜 , 0) (𝑥3 , 0) (𝑥4 , 0) (𝑥2.5 , 0) (𝑥1 , 0)
𝑇 𝑇2 (𝑥2.5 , 0) (𝑥4 , 0) – (𝑥2.5 , 0) (𝑥3 , 0) (𝑥2.5 , 0) (𝑥4 , 0) (𝑥2.5 , 0)
𝑑1
𝑇 𝑇3 (𝑥1.5 , 0) (𝑥3 , 0) (𝑥2.5 , 0) (𝑥3 , 0) (𝑥2 , 0) (𝑥3 , 0) (𝑥3 , 0) (𝑥2 , 0)
𝑇 𝑇4 (𝑥2 , 0) (𝑥1 , 0) (𝑥𝑜 , 0) (𝑥1.5 , 0) (𝑥1 , 0) (𝑥𝑜 , 0) (𝑥1.5 , 0) (𝑥𝑜 , 0)
𝑇 𝑇1 (𝑦1 , 0) (𝑦4 , 0) (𝑦1 , 0) (𝑦1 , 0) (𝑦6 , 0) (𝑦5 , 0) (𝑦5 , 0) (𝑦2.5 , 0)
𝑇 𝑇2 (𝑦4.5 , 0) (𝑦5 , 0) (𝑦4.5 , 0) (𝑦5.5 , 0) (𝑦6 , 0) (𝑦4 , 0) (𝑦5 , 0) (𝑦5 , 0)
𝑑2
𝑇 𝑇3 (𝑦3 , 0) (𝑦5 , 0) (𝑦4 , 0) (𝑦5 , 0) (𝑦3 , 0) – (𝑦4 , 0) (𝑦4 , 0)
𝑇 𝑇4 (𝑦5 , 0) (𝑦1 , 0) (𝑦6 , 0) (𝑦1 , 0) (𝑦2 , 0) (𝑦1 , 0) (𝑦1 , 0) (𝑦1 , 0)
𝑇 𝑇1 (𝑧2.5 , 0) (𝑧6 , 0) (𝑧3 , 0) (𝑧5.5 , 0) (𝑧7 , 0) (𝑧6.5 , 0) (𝑧5 , 0) (𝑧3 , 0)
𝑇 𝑇2 (𝑧5 , 0) (𝑧5 , 0) (𝑧6 , 0) (𝑧5.5 , 0) (𝑧7 , 0) (𝑧6.5 , 0) (𝑧5 , 0) (𝑧3 , 0)
𝑑3
𝑇 𝑇3 (𝑧5 , 0) (𝑧7 , 0) (𝑧7 , 0) (𝑧6 , 0) (𝑧5 , 0) (𝑧5 , 0) (𝑧5.5 , 0) (𝑧6 , 0)
𝑇 𝑇4 (𝑧6 , 0) (𝑧1 , 0) (𝑧5 , 0) (𝑧3 , 0) (𝑧3 , 0) (𝑧7 , 0) (𝑧6 , 0) (𝑧2 , 0)
𝑇 𝑇1 (𝑦1 , 0) (𝑦4 , 0) (𝑦1 , 0) (𝑦1 , 0) (𝑦5 , 0) (𝑦4 , 0) (𝑦3 , 0) (𝑦1 , 0)
𝑇 𝑇2 (𝑦4.5 , 0) (𝑦5 , 0) (𝑦5.5 , 0) (𝑦4 , 0) (𝑦4 , 0) (𝑦3 , 0) (𝑦2.5 , 0) (𝑦3.5 , 0)
𝑑2
𝑇 𝑇3 – (𝑦5 , 0) (𝑦3 , 0) (𝑦5 , 0) (𝑦4 , 0) (𝑦4.5 , 0) (𝑦2 , 0) (𝑦1 , 0)
𝑇 𝑇4 (𝑦5 , 0) (𝑦1 , 0) (𝑦3 , −0.5) (𝑦2 , 0) (𝑦5 , 0) (𝑦1 , 0) (𝑦3 , 0) (𝑦1 , 0)
𝑇 𝑇1 (𝑥𝑜 , 0) (𝑥2 , 0) (𝑥2 , 0) (𝑥1.5 , 0) (𝑥4 , 0) (𝑥3 , 0) (𝑥1.5 , 0) (𝑥𝑜 , 0)
𝑇 𝑇2 (𝑥2.5 , 0) (𝑥3 , 0) (𝑥3 , 0) (𝑥2.5 , 0) (𝑥2.5 , 0) (𝑥2.5 , 0) (𝑥4 , 0) (𝑥3 , 0)
𝑑5
𝑇 𝑇3 (𝑥1 , 0) (𝑥3 , 0) (𝑥2 , 0) (𝑥3 , 0) (𝑥2 , 0) (𝑥2 , 0) – (𝑥2.5 , 0)
𝑇 𝑇4 (𝑥2.5 , 0) (𝑥1.5 , 0) (𝑥2.5 , 0) (𝑥1 , 0) (𝑥2 , 0) (𝑥1 , 0) (𝑥1.5 , 0) (𝑥1 , 0)
Table 6
2tLR numbers of treatment technologies for DMr 𝑑1 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑇 𝑇1 [𝛥(0.1458), 𝛥(0.2813)] [𝛥(0.5625), 𝛥(0.6771)] [𝛥(0.2708), 𝛥(0.4167)] [𝛥(0), 𝛥(0.2792)]
𝑇 𝑇2 [𝛥(0.6250), 𝛥(0.6750)] [𝛥(0.8083), 𝛥(1)] [𝛥(0), 𝛥(0.6333)] [𝛥(0.5833), 𝛥(0.6510)]
𝑇 𝑇3 [𝛥(0.2083), 𝛥(0.5)] [𝛥(0.75), 𝛥(0.8083)] [𝛥(0.5417), 𝛥(0.7222)] [𝛥(0.75), 𝛥(0.7833)]
𝑇 𝑇4 [𝛥(0.5), 𝛥(0.7083)] [𝛥(0.1771), 𝛥(0.3125)] [𝛥(0), 𝛥(0.5333)] [𝛥(0.3), 𝛥(0.3750)]
Table 7
2tLR numbers of treatment technologies for DMr 𝑑1 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 [𝛥(0.75), 𝛥(0.8917)] [𝛥(0.8125), 𝛥(1)] [𝛥(0.5313), 𝛥(0.6944)] [𝛥(0.1389), 𝛥(0.3472)]
𝑇 𝑇2 [𝛥(0.6806), 𝛥(0.8750)] [𝛥(0.5833), 𝛥(0.6823)] [𝛥(0.6917), 𝛥(1)] [𝛥(0.5278), 𝛥(0.7361)]
𝑇 𝑇3 [𝛥(0.5), 𝛥(0.5583)] [𝛥(0.5250), 𝛥(0.75)] [𝛥(0.4875), 𝛥(0.75)] [𝛥(0.3333), 𝛥(0.6354)]
𝑇 𝑇4 [𝛥(0.25), 𝛥(0.4583)] [𝛥(0), 𝛥(0.2917)] [𝛥(0.3056), 𝛥(0.5)] [𝛥(0), 𝛥(0.1667)]
Table 8
2tLR numbers of treatment technologies for DMr 𝑑2 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑇 𝑇1 [𝛥(0), 𝛥(0.1792)] [𝛥(0.5), 𝛥(0.6417)] [𝛥(0.3167), 𝛥(0.5)] [𝛥(0.1771), 𝛥(0.5313)]
𝑇 𝑇2 [𝛥(0.6250), 𝛥(0.6750)] [𝛥(0.6875), 𝛥(0.8542)] [𝛥(0.5625), 𝛥(0.7917)] [𝛥(0.5833), 𝛥(0.6510)]
𝑇 𝑇3 [𝛥(0.1250), 𝛥(0.4375)] [𝛥(0.75), 𝛥(0.8083)] [𝛥(0.5), 𝛥(0.6333)] [𝛥(0.75), 𝛥(0.7833)]
𝑇 𝑇4 [𝛥(0.5625), 𝛥(0.7604)] [𝛥(0.2167), 𝛥(0.3750)] [𝛥(0.4167), 𝛥(0.75)] [𝛥(0.2083), 𝛥(0.3333)]
Table 9
2tLR numbers of treatment technologies for DMr 𝑑2 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 [𝛥(0.8917), 𝛥(1)] [𝛥(0.7083), 𝛥(0.8490)] [𝛥(0.3750), 𝛥(0.5917)] [𝛥(0), 𝛥(0.2417)]
𝑇 𝑇2 [𝛥(0.6250), 𝛥(0.7833)] [𝛥(0.5833), 𝛥(0.6823)] [𝛥(0.6917), 𝛥(1)] [𝛥(0.5833), 𝛥(0.7917)]
𝑇 𝑇3 [𝛥(0.5), 𝛥(0.5583)] [𝛥(0.25), 𝛥(0.6563)] [𝛥(0), 𝛥(0.4875)] [𝛥(0.4306), 𝛥(0.6806)]
𝑇 𝑇4 [𝛥(0.3646), 𝛥(0.6667)] [𝛥(0.1458), 𝛥(0.5625)] [𝛥(0.3056), 𝛥(0.5000)] [𝛥(0.1667), 𝛥(0.2500)]
Using Eqs. (25), (26), the 2tL cloud evaluations of 2tLR num- Based on the calculations of Liu et al. (2014), the weights of five
bers, known as D2tLRN clouds, for DMrs 𝑑1 and 𝑑5 are computed in DMrs are given based on their different expertise and domain as 𝛾1 =
Tables 10–13. 0.15, 𝛾2 = 0.2, 𝛾3 = 0.3, 𝛾4 = 0.2 and 𝛾5 = 0.15. Using Eq. (28), the
11
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Table 10
D2tLRN clouds of treatment technologies for DMr 𝑑1 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑇 𝑇1 (𝛥(0.2135), 0.0226, 0.0068) (𝛥(0.6198), 0.0191, 0.0057) (𝛥(0.3438), 0.0243, 0.0073) (𝛥(0.1396), 0.0465, 0.0140)
𝑇 𝑇2 (𝛥(0.65), 0.0083, 0.0025) (𝛥(0.9042), 0.0319, 0.0096) (𝛥(0.3167), 0.1056, 0.0317) (𝛥(0.6172), 0.0113, 0.0034)
𝑇 𝑇3 (𝛥(0.3542), 0.0486, 0.0146) (𝛥(0.7792), 0.0097, 0.0029) (𝛥(0.6319), 0.0301, 0.0090) (0.7667, 0.0056, 0.0017)
𝑇 𝑇4 (𝛥(0.6042), 0.0347, 0.0104) (𝛥(0.2448), 0.0226, 0.0068) (𝛥(0.2667), 0.0889, 0.0267) (𝛥(0.3375), 0.0125, 0.0038)
Table 11
D2tLRN clouds of treatment technologies for DMr 𝑑1 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 (𝛥(0.8208), 0.0236, 0.0071) (𝛥(0.9063), 0.0313, 0.0094) (𝛥(0.6128), 0.0272, 0.0082) (𝛥(0.2431), 0.0347, 0.010)
𝑇 𝑇2 (𝛥(0.7778), 0.0324, 0.0097) (𝛥(0.6328), 0.0165, 0.0049) (𝛥(0.8458), 0.0514, 0.0154) (𝛥(0.6319), 0.0347, 0.0104)
𝑇 𝑇3 (𝛥(0.5292), 0.0097, 0.0029) (𝛥(0.6375), 0.0375, 0.0113) (𝛥(0.6188), 0.0438, 0.0131) (𝛥(0.4844), 0.0503, 0.0151)
𝑇 𝑇4 (𝛥(0.3542), 0.0347, 0.0104) (𝛥(0.1458), 0.0486, 0.0146) (𝛥(0.4028), 0.0324, 0.0097) (𝛥(0.0833), 0.0278, 0.0083)
Table 12
2tLR numbers of treatment technologies for DMr 𝑑5 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑇 𝑇1 (𝛥(0.0896), 0.0299, 0.0090) (𝛥(0.5708), 0.0236, 0.0071) (𝛥(0.4083), 0.0306, 0.0092) (𝛥(0.3542), 0.0590, 0.0177)
𝑇 𝑇2 (𝛥(0.65), 0.0083, 0.0025) (𝛥(0.7708), 0.0278, 0.0083) (𝛥(0.6771), 0.0382, 0.0115) (𝛥(0.6172), 0.0113, 0.0034)
𝑇 𝑇3 (𝛥(0.2813), 0.0521, 0.0156) (𝛥(0.7792), 0.0097, 0.0029) (𝛥(0.5667), 0.0222, 0.0067) (𝛥(0.7667), 0.0056, 0.0017)
𝑇 𝑇4 (𝛥(0.6615), 0.0330, 0.0099) (𝛥(0.2958), 0.0264, 0.0079) (𝛥(0.5833), 0.0556, 0.0167) (𝛥(0.2708), 0.0208, 0.0063)
Table 13
D2tLRN clouds of treatment technologies for DMr 𝑑5 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 (𝛥(0.9458), 0.0181, 0.0054) (𝛥(0.7786), 0.0234, 0.0070) (𝛥(0.4833), 0.0361, 0.0108) (𝛥(0.1208), 0.0403, 0.0121)
𝑇 𝑇2 (𝛥(0.7042), 0.0264, 0.0079) (𝛥(0.6328), 0.0165, 0.0049) (𝛥(0.8458), 0.0514, 0.0154) (𝛥(0.6875), 0.0347, 0.0104)
𝑇 𝑇3 (𝛥(0.5292), 0.0097, 0.0029) (𝛥(0.4531), 0.0677, 0.0203) (𝛥(0.2438), 0.0813, 0.0244) (𝛥(0.5556), 0.0417, 0.0125)
𝑇 𝑇4 (𝛥(0.5156), 0.0503, 0.0151) (𝛥(0.3542), 0.0694, 0.0208) (𝛥(0.4028), 0.0324, 0.0097) (𝛥(0.2083), 0.0139, 0.0042)
Table 14
Aggregated D2tLR cloud relation matrix of treatment technologies for 𝑊 𝑓1 to 𝑊 𝑓4 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑇 𝑇1 (𝛥(0.1862),0.0493,0.0148) (𝛥(0.6491),0.0383,0.0115) (𝛥(0.3126),0.0569,0.0171) (𝛥(0.3111),0.1136,0.0341)
𝑇 𝑇2 (𝛥(0.6750),0.0222,0.0067) (𝛥(0.7961),0.0590,0.0177) (𝛥(0.6426),0.1148,0.0344) (𝛥(0.6216),0.0293,0.0088)
𝑇 𝑇3 (𝛥(0.3609),0.1112,0.0333) (𝛥(0.8140),0.0219,0.0066) (𝛥(0.6538),0.0670,0.0201) (𝛥(0.7833),0.0148,0.0044)
𝑇 𝑇4 (𝛥(0.7128),0.0613,0.0184) (𝛥(0.2108),0.0403,0.0121) (𝛥(0.5433),0.1553,0.0466) (𝛥(0.3003),0.0380,0.0114)
Table 15
Aggregated D2tLR cloud relation matrix of treatment technologies for 𝑊 𝑓5 to 𝑊 𝑓8 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 (𝛥(0.8927), 0.0470, 0.0141) (𝛥(0.8126), 0.0560, 0.0168) (𝛥(0.5991), 0.0730, 0.0219) (𝛥(0.2480), 0.0765, 0.0229)
𝑇 𝑇2 (𝛥(0.7975), 0.0711, 0.0213) (𝛥(0.6576), 0.0490, 0.0147) (𝛥(0.6425), 0.1759, 0.0528) (𝛥(0.6135), 0.0817, 0.0245)
𝑇 𝑇3 (𝛥(0.5636), 0.0304, 0.0091) (𝛥(0.5061), 0.1271, 0.0381) (𝛥(0.4815), 0.1359, 0.0408) (𝛥(0.5404), 0.0989, 0.0297)
𝑇 𝑇4 (𝛥(0.4732), 0.1, 0.03) (𝛥(0.3451), 0.1368, 0.0410) (𝛥(0.4563), 0.0941, 0.0282) (𝛥(0.1701), 0.0382, 0.0115)
12
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Algorithm The proposed D2tLRN cloud HCWM method. methodology is considered ‘‘very good’’ in terms of cost, it shows that
the cost of this HCWM methodology is very low. Based on Eqs. (41)
1. Identify 𝑝 treatment technologies 𝑇 𝑇1 , 𝑇 𝑇2 , … 𝑇 𝑇𝑝 against and (43), the normalized cloud decision matrix of Tables 14 and 15
𝑟 waste factors 𝑊 𝑓1 , 𝑊 𝑓2 , … , 𝑊 𝑓𝑟 provided by 𝑠 experts are computed in Tables 19 and 20. Table 19 contains the weighted
𝑑1 , 𝑑2 , … , 𝑑𝑠 . normalized 2tL cloud values by multiplying weight values (Table 17)
[ 𝑖 ( )] with normalized D2tLRN clouds of treatment technologies for waste
2. Construct the initial decision matrix 2𝑡𝐿𝑖 = 𝛿𝑗𝑘 = 𝑥𝑖𝑗𝑘 , 𝜉𝑗𝑘
𝑖
𝑝×𝑟 factors 𝑊 𝑓1 , 𝑊 𝑓2 , 𝑊 𝑓3 , 𝑊 𝑓4 using operations of 2tL clouds given in
of treatment technologies given in Eq. (19) based on the scales
Table 3. Table 20 contains the weighted normalized 2tL cloud values
described in Table 2, evaluated by ith expert, 1 ≤ 𝑖 ≤ 𝑠.
for waste factors 𝑊 𝑓5 , 𝑊 𝑓6 , 𝑊 𝑓7 , 𝑊 𝑓8 using the same multiplication
3. For each 1 ≤ 𝑖 ≤ 𝑠, convert each [initial decision matrix] 2𝑡𝐿𝑖 into operations of 2tL clouds. The PI and NI solutions of waste factors also
[( 𝑖 𝑖 ) ( 𝑖 𝑖 )]
2tLR relation matrix 2𝑡𝐿𝑅𝑖 = 𝑥𝑗𝑘 , 𝜉𝑗𝑘 , 𝑥𝑗𝑘 , 𝜉𝑗𝑘 using computed in Tables 19 and 20 from weighted normalized 2tL cloud
𝑟×𝑛
Eqs. (20), (21), (22), (23). values.
4. For each 1 ≤ 𝑖 ≤ 𝑖, transform the 2tLR relation ma- Using Eqs. (45), (46), the distances of weighted normalized D2tLRN
[
trix into D2tlRN relation matrix 𝐷2𝑡𝐿𝑅𝑁𝐶 𝑖 = 𝑖
𝑗𝑘 = cloud evaluations of treatment technologies for each waste factor from
( 𝑖 )] PI and NI solutions are computed in Tables 21 and 22. Table 21 con-
̃ 𝑖 𝑖
(ℎ𝐸𝑥𝑖 , 𝐸𝑥𝑗𝑘 ), 𝐸𝑛𝑗𝑘 , 𝐻𝐸𝑛𝑗𝑘 using Eqs. (25), (26) as ex- tains the distances of weighted normalized 2tL cloud values (Table 19)
𝑗𝑘 𝑟×𝑛
plained in Section 3. from PI and NI solutions for waste factors 𝑊 𝑓1 , 𝑊 𝑓2 , 𝑊 𝑓3 and 𝑊 𝑓4 .
5. Aggregate the D2tLRN cloud evaluations of different experts to The resulting distance values are 2tL values which are given in the
compute aggregated D2tLRN cloud]relation matrix 𝐷2𝑡𝐿𝑅𝑁𝐶 = form of 𝛥(𝛼), 𝛼 ∈ [0, 1]. Table 22 contains the distances of weighted
[ ( )
𝑗𝑘 = (ℎ𝐸𝑥 , 𝐸𝑥̃ 𝑗𝑘 ), 𝐸𝑛𝑗𝑘 , 𝐻𝐸𝑛𝑗𝑘 using averaging operator normalized 2tL cloud values (Table 20) from PI and NI solutions for
𝑗𝑘 𝑟×𝑛 waste factors 𝑊 𝑓5 , 𝑊 𝑓6 , 𝑊 𝑓7 and 𝑊 𝑓8 .
given in Eq. (28) and Definition 7.
Using Eqs. (47)–(49), the distances of treatment technologies from
6. Convert[ the relative
] importance of waste factors given
[ by
PI and NI solutions and closeness degrees are computed in Table 23.
𝑖
DMrs (𝑥𝑘 , 𝜉𝑘 ) 𝑖 into D2tLRN cloud evaluations 𝑘 = 𝑖 In Table 23, the distance of each treatment technology from PI and
𝑟×𝑠 ] NI solutions. These positive and negative distances are combined to
( )
(𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖𝑘 , 𝐻𝐸𝑛𝑖𝑘 using Eqs. (31), (32), (33), (34), compute closeness degree in order to rank the treatment technologies.
𝑘 𝑘
𝑟×𝑠 The treatment technologies are ranked as 𝑇 𝑇3 ≻ 𝑇 𝑇2 ≻ 𝑇 𝑇1 ≻ 𝑇 𝑇4
(35), (36).
[ 𝑖] which shows that 𝑇 𝑇3 is the best and 𝑇 𝑇4 is the lowest waste disposal
7. Aggregate the D2tLRN cloud evaluations 𝑘 𝑟×𝑠 =
[( )] technology among all the alternatives.
(𝑥𝐸𝑥𝑖 , 𝜉𝐸𝑥𝑖 ), 𝐸𝑛𝑖𝑘 , 𝐻𝐸𝑛𝑖𝑘 of relative importance of waste
𝑘 𝑘 𝑟×𝑠
factors of different experts into single aggregated 2tL assessment 6. Comparison analysis
[ ] [( )]
values 𝑘 = (𝑥𝐸𝑥𝑘 , 𝜉𝐸𝑥𝑘 ), 𝐸𝑛𝑘 , 𝐻𝐸𝑛𝑘 using averaging
𝑟×1 𝑟×1
operator given in Eq. (36) and Definition 7. This section is devoted to the comparison analysis of treatment al-
8. Compute subjective weights 𝛿𝑘𝑠 , objective weights 𝛿𝑘𝑜 and hybrid ternatives for HCWM using different multi-criteria decision approaches
weights 𝛿𝑘 , for each 1 ≤ 𝑘 ≤ 𝑟, using Eqs. (37), (38), (39), with the proposed D2tLRN cloud TOPSIS method to verify its ratio-
[ (40).
] nality and reliability. It integrates the benefits of 2tL information with
9. Normalize the aggregated D2tLRN cloud relation matrix 𝑗𝑘
[ ] [( )] 𝑟×𝑛 rough approximations and cloud model theory to cope with objective
into 𝑗𝑘′ = (𝑥𝐸𝑥𝑗𝑘 , 𝜉𝐸𝑥𝑗𝑘 )′ , 𝐸𝑛′𝑗𝑘 , 𝐻𝐸𝑛′𝑗𝑘 using Eq. (41). uncertainty and randomness in DMrs evaluations and the merits of
[ ] TOPSIS method to compute the rankings of HCWM technologies. The
10. Convert the normalized matrix 𝑗𝑘 ′ into weighted normalized
[ ] [ ] rankings of HCWM technologies by four different approaches are given
matrix 𝑗𝑘 ∗ = 𝛿𝑘 𝑗𝑘′ = in Table 24.
[( 𝑟×𝑛 𝑟×𝑛
√ √ )] In Table 24, the rankings of treatment alternatives by five different
𝛥(𝛿𝑘 𝛥 (𝑥𝐸𝑥𝑗𝑘 , 𝜉𝐸𝑥𝑗𝑘 ) ), 𝛿𝑘 𝐸𝑛′𝑗𝑘 , 𝛿𝑘 𝐻𝐸𝑛′𝑗𝑘
−1 ′ using Eq.
𝑟×𝑛 methods are given. It can be seen that there are similarities and differ-
(43). ences between the results of all the methods. 𝑇 𝑇2 and 𝑇 𝑇3 are ranked as
11. Compute PI (positive ideal) and NI (negative ideal) solutions of first and second in all the methods, however, in the proposed methods,
the PI and NI solutions of waste factors 𝑊 𝑓𝑘 , 1 ≤ 𝑘 ≤ 𝑟, are 𝑇 𝑇3 is ranked first and 𝑇 𝑇2 is ranked second. 𝑇 𝑇1 is ranked third and
computed as given in Eqs. (45) and (46). 𝑇 𝑇4 is ranked fourth in fuzzy VIKOR (Liu et al., 2013), interval 2tL
12. Computing the Hamming distance of treatment technologies MULTIMOORA (Liu et al., 2014) and the proposed HCWM method.
from PI and NI solutions using Eqs. (13), (14), (15), (47), (48). However, these are ranked fourth and third, respectively, in fuzzy
13. Determine the closeness degree 𝑐𝑑𝑗 , 1 ≤ 𝑗 ≤ 𝑟, of each treatment OWA (Dursun et al., 2011b) and modified fuzzy TOPSIS (Dursun et al.,
technology using Eq. (49) which represents the strength of waste 2011a). In fuzzy OWA (Dursun et al., 2011b), fuzzy VIKOR (Liu et al.,
management priority. The greater the value of 𝑐𝑑𝑗 means the jth 2013) and modified fuzzy TOPSIS (Dursun et al., 2011a) methods, the
treatment technology is more beneficial. uncertainty is discussed using fuzzy numbers. The comparison of results
by different approaches is also illustrated in Fig. 3.
To discuss and illustrate the significance and out-performance of
the proposed D2tLRN cloud TOPSIS method, a detailed comparison
factor is computed from these 2tL cloud evaluations and then further analysis, statistical analysis and sensitivity analysis is performed in
used for determining objective weights. The hybrid weights using the this section. The comparison is done with four different approaches
product of subjective and objective weights. Table 18 contains the based on two main theories including fuzzy numbers based methods
relative importance of DMrs for waste factors 𝑊 𝑓5 , 𝑊 𝑓6 , 𝑊 𝑓7 , 𝑊 𝑓8 in and interval 2tL information.
the form of D2tLRN clouds. These 2tL clouds are utilized similarly to The first comparison is made with fuzzy number based approaches
compute hybrid weights of DMrs for waste factors 𝑊 𝑓5 , 𝑊 𝑓6 , 𝑊 𝑓7 and including fuzzy OWA (Dursun et al., 2011b), fuzzy VIKOR (Liu et al.,
𝑊 𝑓8 . 2013) and modified fuzzy TOPSIS (Dursun et al., 2011a). In these
HCWM methods, the initial decision matrix containing linguistic or
Based on the evaluations by DMrs in Liu et al. (2014), it is as- numeric data is converted into fuzzy numbers using pre-defined mea-
sumed that all criteria are benefit criteria. For instance, if an HCWM surement scales. These measurement scales are usually the suppositions
13
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Table 17
Objective and subjective weights of waste factors 𝑊 𝑓1 to 𝑊 𝑓4 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑑1 (𝛥(0.8), 0.0111, 0.0033) (𝛥(0.9833), 0.05, 0.0150) (𝛥(0.6333), 0.0111, 0.0033) (𝛥(0.8167), 0.0056, 0.0017)
𝑑2 (𝛥(0.7167), 0.0167, 0.0050) (𝛥(1.1875), 0.0486, 0.0146) (𝛥(0.55), 0.0167, 0.0050) (𝛥(0.8167), 0.0056, 0.0017)
𝑑3 (𝛥(0.8), 0.0111, 0.0033) (𝛥(1.1875), 0.0486, 0.0146) (𝛥(0.6333), 0.0111, 0.0033) (𝛥(0.8167), 0.0056, 0.0017)
𝑑4 (𝛥(0.8), 0.0111, 0.0033) (𝛥(1.3167), 0.0611, 0.0183) (𝛥(0.55), 0.0167, 0.0050) (𝛥(0.7333), 0.0222, 0.0067)
𝑑5 (𝛥(0.7167), 0.0167, 0.0050) (𝛥(0.9833), 0.0500, 0.0150) (𝛥(0.6333), 0.0111, 0.0033) (𝛥(0.8167), 0.0056, 0.0017)
Aggregated (𝛥(0.7708), 0.0293, 0.0088) (𝛥(1.1521), 0.1144, 0.0343) (𝛥(0.6), 0.0296, 0.0089) (𝛥(0.8), 0.0198, 0.0059)
𝛿𝑘𝑠 (𝛥(0.1125), 0.0048, 0.0014) (𝛥(0.1681), 0.0170, 0.0051) (𝛥(0.0876), 0.0047, 0.0014) (𝛥(0.1168), 0.0037, 0.0011)
𝜎𝑘2 0.0009 0.0143 0.0010 0.0004
𝛿𝑘𝑜 0.0465 0.7102 0.0475 0.0212
𝛿𝑘𝑠 × 𝛿𝑘𝑜 𝛥(0.0052) 𝛥(0.1194) 𝛥(0.0042) 𝛥(0.0025)
𝛿𝑘 𝛥(0.0342) 𝛥(0.7817) 𝛥(0.0272) 𝛥(0.0162)
Table 18
Objective and subjective weights of waste factors 𝑊 𝑓5 to 𝑊 𝑓8 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑑1 (𝛥(0.8167),0.0056,0.0017) (𝛥(1),0,0) (𝛥(0.8333),0.0139,0.0042) (𝛥(0.8667),0.0111,0.0033)
𝑑2 (𝛥(0.8167),0.0056,0.0017) (𝛥(1),0,0) (𝛥(0.8333),0.0139,0.0042) (𝛥(0.95),0.0167,0.0050)
𝑑3 (𝛥(0.8167),0.0056,0.0017) (𝛥(1),0,0) (𝛥(0.75),0.0278,0.0083) (𝛥(0.8667),0.0111,0.0033)
𝑑4 (𝛥(0.8167),0.0056,0.0017) (𝛥(1),0,0) (𝛥(0.9167),0.0278,0.0083) (𝛥(0.95),0.0167,0.0050)
𝑑5 (𝛥(0.7333),0.0222,0.0067) (𝛥(1),0,0) (𝛥(0.8333),0.0139,0.0042) (𝛥(0.8667),0.0111,0.0033)
Aggregated (𝛥(0.8042),0.0188,0.0056) (𝛥(1),0,0) (𝛥(0.8250),0.0446,0.0134) (𝛥(0.9),0.0296,0.0089)
𝛿𝑘𝑠 (𝛥(0.1174),0.0036,0.0011) (𝛥(0.1459),0.0029,0.0009) (𝛥(0.1204),0.0069,0.0021) (𝛥(0.1313),0.0050,0.0015)
𝜎𝑘2 0.0004 0 0.0022 0.0010
𝛿𝑘𝑜 0.0191 0 0.1079 0.0475
𝛿𝑘𝑠 × 𝛿𝑘𝑜 𝛥(0.0022) 𝛥(0) 𝛥(0.0130) 𝛥(0.0062)
𝛿𝑘 𝛥(0.0147) 𝛥(0) 𝛥(0.0851) 𝛥(0.0409)
Table 19
Weighted normalized D2tLRN cloud decision matrix for 𝑊 𝑓1 to 𝑊 𝑓4 .
TTs Waste factors
𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
𝑇 𝑇1 (𝛥(0.0089),0.0129,0.0003) (𝛥(0.6234),0.0458,0.0002) (𝛥(0.0130),0.0159,0.0003) (𝛥(0.0064),0.0185,0.0006)
𝑇 𝑇2 (𝛥(0.0324),0.0079,0) (𝛥(0.7645),0.0682,0.0005) (𝛥(0.0268),0.0322,0.0005) (𝛥(0.0129),0.0051,0)
𝑇 𝑇3 (𝛥(0.0173),0.0290,0.0008) (𝛥(0.7817),0.0336,0.0001) (𝛥(0.0272),0.0222,0.0002) (𝛥(0.0162),0.0034,0)
𝑇 𝑇4 (𝛥(0.0342),0.0169,0.0001) (𝛥(0.2024),0.0442,0.0008) (𝛥(0.0226),0.0410,0.0011) (𝛥(0.0062),0.0062,0.0001)
𝑆+ (𝛥(0.0342),0.0079,0) (𝛥(0.7817),0.0336,0.0001) (𝛥(0.0272),0.0159,0.0002) (𝛥(0.0162),0.0034,0)
𝑆− (𝛥(0.0089),0.0290,0.0008) (𝛥(0.2024),0.0682,0.0008) (𝛥(0.0130),0.0410,0.0011) (𝛥(0.0062),0.0185,0.0006)
Table 20
Weighted normalized D2tLRN cloud decision matrix for 𝑊 𝑓5 to 𝑊 𝑓8 .
TTs Waste factors
𝑊 𝑓5 𝑊 𝑓6 𝑊 𝑓7 𝑊 𝑓8
𝑇 𝑇1 (𝛥(0.0147),0.0076,0) (𝛥(0),0,0) (𝛥(0.0793),0.0453,0.0005) (𝛥(0.0165),0.0257,0.0007)
𝑇 𝑇2 (𝛥(0.0131),0.0103,0.0001) (𝛥(0),0,0) (𝛥(0.0851),0.0865,0.0021) (𝛥(0.0409),0.0297,0.0004)
𝑇 𝑇3 (𝛥(0.0093),0.0049,0) (𝛥(0),0,0) (𝛥(0.0638),0.0665,0.0017) (𝛥(0.0360),0.0344,0.0006)
𝑇 𝑇4 (𝛥(0.0078),0.0138,0.0003) (𝛥(0),0,0) (𝛥(0.0604),0.0488,0.0009) (𝛥(0.0113),0.0131,0.0003)
𝑆+ (𝛥(0.0147),0.0049,0) (𝛥(0),0,0) (𝛥(0.0851),0.0453,0.0005) (𝛥(0.0409),0.0131,0.0003)
𝑆− (𝛥(0.0078),0.0138,0.0003) (𝛥(0),0,0) (𝛥(0.0604),0.0865,0.0021) (𝛥(0.0113),0.0344,0.0007)
Table 21
Distances of weighted normalized 2tL cloud evaluations from PI and NI solutions for 𝑊 𝑓1 to 𝑊 𝑓4 .
Treatment 𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
technologies Positive Negative Positive Negative Positive Negative Positive Negative
𝑇 𝑇1 𝛥(0.0258) 𝛥(0.0161) 𝛥(0.1588) 𝛥(0.4215) 𝛥(0.0142) 𝛥(0.0251) 𝛥(0.0180) 𝛥(0.0002)
𝑇 𝑇2 𝛥(0.0018) 𝛥(0.0315) 𝛥(0.0386) 𝛥(0.5621) 𝛥(0.0163) 𝛥(0.0163) 𝛥(0.0038) 𝛥(0.0149)
𝑇 𝑇3 𝛥(0.0270) 𝛥(0.0084) 𝛥(0) 𝛥(0.5803) 𝛥(0.0063) 𝛥(0.0236) 𝛥(0) 𝛥(0.0181)
𝑇 𝑇4 𝛥(0.0090) 𝛥(0.0280) 𝛥(0.5794) 𝛥(0.0240) 𝛥(0.0255) 𝛥(0.0096) 𝛥(0.0104) 𝛥(0.0123)
and has no link with the given data. The fuzzy number scales are discuss uncertainty from initial data without pre-defined assumptions.
usually the same against each type of data. This way of inserting Rough approximations play an important part to overcome the limita-
assumed measurement scales in linguistic data ignores the original tions of fuzzy-based approaches and cope with uncertainty using initial
nature of the data and can increase uncertainty. The best way is to data. However, rough numbers cannot deal with randomness in DMrs
14
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Table 22
Distances of weighted normalized 2tL cloud evaluations from PI and NI solutions for 𝑊 𝑓5 to 𝑊 𝑓8 .
Treatment 𝑊 𝑓1 𝑊 𝑓2 𝑊 𝑓3 𝑊 𝑓4
technologies Positive Negative Positive Negative Positive Negative Positive Negative
𝑇 𝑇1 𝛥(0.0027) 𝛥(0.0092) 𝛥(0) 𝛥(0) 𝛥(0.0057) 𝛥(0.0453) 𝛥(0.0274) 𝛥(0.0101)
𝑇 𝑇2 𝛥(0.0057) 𝛥(0.0063) 𝛥(0) 𝛥(0) 𝛥(0.0412) 𝛥(0.0247) 𝛥(0.0166) 𝛥(0.0299)
𝑇 𝑇3 𝛥(0.0054) 𝛥(0.0090) 𝛥(0) 𝛥(0) 𝛥(0.0301) 𝛥(0.0202) 𝛥(0.0219) 𝛥(0.0247)
𝑇 𝑇4 𝛥(0.0112) 𝛥(0) 𝛥(0) 𝛥(0) 𝛥(0.0249) 𝛥(0.0377) 𝛥(0.0295) 𝛥(0.0214)
15
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Table 24
Comparison of proposed HCWM mode with existing approaches.
Treatment Fuzzy OWA Fuzzy VIKOR Modified fuzzy TOPSIS Interval 2tL MULTIMOORA Proposed D2tLRN
alternatives (Dursun et al., 2011b) (Liu et al., 2013) (Dursun et al., 2011a) (Liu et al., 2014) cloud TOPSIS
FI values Ranking 𝑄 values Ranking Closeness degree Ranking 𝑑𝑖 values Ranking Closeness degree Ranking
𝑇 𝑇1 0.694 4 0.806 3 0.433 4 𝛥(0.0776) 3 𝛥(0.2795) 3
𝑇 𝑇2 0.717 1 0 1 0.706 1 𝛥(0.0141) 1 𝛥(0.0982) 2
𝑇 𝑇3 0.711 2 0.411 2 0.681 2 𝛥(0.0381) 2 𝛥(0.0743) 1
𝑇 𝑇4 0.698 3 0.841 4 0.485 3 𝛥(0.0827) 4 𝛥(0.9079) 4
Fig. 4. Sensitivity analysis of treatment alternatives for different variations of DMrs weights.
Based on the ranking obtained by D2tLRN cloud TOPSIS method, 8. Conclusion and future directions
microwave (𝑇 𝑇3 ) is ranked as the optimal HCWM technology, which
is then followed by steam sterilization (𝑇 𝑇2 ), incineration (𝑇 𝑇1 ) and The selection of suitable disposal technology for HCWM is a re-
landfill (𝑇 𝑇4 ). Microwave is computed as a best treatment technology strained and difficult task for the civil authorities of developed coun-
for HCWM as it maintains public health and minimizes environmental tries. In this research article, a novel mathematical model is proposed
effect. The non-incineration alternatives, that is, microwave and steam by integrating 2-tuple linguistic setting into rough set theory and cloud
sterilization are ranked as first and second HCWM technologies due theory. Two novel linguistic manipulation models, namely, 2-tuple
to their fewer pollutants and non-hazardous leftovers (Dursun et al., linguistic rough numbers and dual 2-tuple linguistic rough number
2011b). Moreover, non-incineration technologies are ranked before (D2tLRN) clouds are constructed to handle uncertainty with random-
incineration alternatives because the leftover matters and chemical ness and multi-granularity simultaneously. The idea of 2tL cloud model
compounds of incinerators have negative impacts on the environment with certain binary operations is also presented in this paper. A hybrid
and public health (Diaz et al., 2005). However, land has low cost and is weighting scheme is utilized to evaluate the relative importance of
most economical HCWM technology as compared to the others, but its waste factors using both subjective and objective aspects of uncertainty.
use is very limited due to its environmental and public effects are more The novel D2tLRN cloud TOPSIS method is presented to design a
adverse (Liu et al., 2013). Conclusively, the construction of microwave novel evaluation approach for the selection of HCWM technologies. The
can be proved as the most cost effective as well as most suitable in initial 2tL data is converted into 2tLR numbers and then D2tLRN cloud
terms of environmental, technical and social aspects. values to cope with multi-granularity using rough approximations and
The existing approaches handle different types of uncertainties with normal 2tL clouds. Using statistical variance, distance formulae and
hybrid mathematical structures as the uncertainty existing in real-world arithmetic operations of 2tL clouds, 2tL cloud positive and 2tL cloud
data cannot be dealt only with a single mathematical approach. The negative ideal solutions are computed to rank the treatment technolo-
concept of multi-granulation can be dealt with 2tL variables; rough gies. The importance and significance of the proposed method is studied
numbers can be utilized to extract uncertainty without pre-defined with an empirical case study of HCWM in a most crowded municipality.
parameters; fuzzy numbers, rough fuzzy numbers, fuzzy rough num- The results of the proposed technique show an outstanding correlation
bers are implemented to cope with intrapersonal and interpersonal with the results obtained by previous approaches.
uncertainties; CM theory handles randomness in DMrs judgements; Although the proposed waste management model provides contri-
interval rough numbers deal with interval data instead of single fixed butions in theory an practice. It still has some shortcomings enhancing
16
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Table 26
Ranking results of treatment alternatives under different variations of DMrs weights.
Treatment 𝑆1 = {0.5, 0.1, 0.1, 0.1, 0.1} 𝑆2 = {0.1, 0.1, 0.1, 0.1, 0.5} 𝑆3 = {0.3, 0.1, 0.2, 0.1, 0.3} 𝑆4 = {0.3, 0.1, 0.2, 0.1, 0.3}
technology 𝑐𝑑 Ranking 𝑐𝑑 Ranking 𝑐𝑑 Ranking 𝑐𝑑 Ranking
𝑇 𝑇1 𝛥(0.3595) 3 𝛥(0.3523) 3 𝛥(0.3254) 3 𝛥(0.2795) 3
𝑇 𝑇2 𝛥(0.0883) 1 𝛥(0.0911) 1 𝛥(0.0895) 2 𝛥(0.0982) 2
𝑇 𝑇3 𝛥(0.1173) 2 𝛥(0.1021) 2 𝛥(0.0877) 1 𝛥(0.0743) 1
𝑇 𝑇4 𝛥(0.9170) 4 𝛥(0.9053) 4 𝛥(0.9122) 4 𝛥(0.9079) 4
the need of future studies. The evaluations of only five DMrs are Zhai, L.Y., Khoo, L.P., Zhong, Z.W., 2008. A rough set enhanced fuzzy approach to
considered in this research study. The increase in the team of DMrs can quality function deployment. Int. J. Adv. Manuf. Technol. 37 (5), 613–624.
Li, D., Haijun, M., Xuemei, S., 1995b. Membership clouds and membership cloud
enhance the accuracy and reliability of the results. Only four decision
generators. J. Comput. Res. Dev. 32 (6), 15–20.
factors are considered as performance indicator for treatment tech- Li, J., Fang, H., Song, W., 2019b. Sustainable supplier selection based on SSCM
nologies. These decision factors can be increased including engineering practices: A rough cloud TOPSIS approach. J. Clean. Prod. 222, 606–621.
factors, legal factors or process factors etc, in the future studies for Li, J., Fang, H., Song, W., 2019a. Modified failure mode and effects analysis un-
der uncertainty: A rough cloud theory-based approach. Appl. Soft Comput. 78,
a more comprehensive assessment criteria system. The future study
195–208.
will be to analysing and discussing group decision making problems Jianxing, Y., Shibo, W., Haicheng, C., Yang, Y., Haizhao, F., Jiahao, L., 2021. Risk
for various real-world applications including risk assessment, factory assessment of submarine pipelines using modified FMEA approach based on cloud
location problem, robot evaluation and wind farm selection in different model and extended VIKOR method. Process Saf. Environ. Prot. 155, 555–574.
areas based on multi-granulation of Pythagorean fuzzy information, Gong, X., Yu, C., Min, L., 2021. A cloud theory-based multi-objective portfolio selection
model with variable risk appetite. Expert Syst. Appl. 176, 114911.
roughness of intuitionistic fuzzy numbers and q-rung orthopair fuzzy
Zhu, G.N., Hu, J., Ren, H., 2020. A fuzzy rough number-based AHP-TOPSIS for design
environment. concept evaluation under uncertain environments. Appl. Soft Comput. 91, 106228.
Chen, Z., Ming, X., 2020. A rough-fuzzy approach integrating best-worst method and
Ethical approval data envelopment analysis to multi-criteria selection of smart product service
module. Appl. Soft Comput. 94, 106479. http://dx.doi.org/10.1016/j.asoc.2020.
106479.
This article does not contain any studies with human participants Huang, G., Xiao, L., Zhang, G., 2021a. Assessment and prioritization method of key
or animals performed by the author. engineering characteristics for complex products based on cloud rough numbers.
Adv. Eng. Inform. 49, 101309.
Huang, G., Xiao, L., Zhang, G., 2021b. Decision-making model of machine tool
CRediT authorship contribution statement remanufacturing alternatives based on dual interval rough number clouds. Eng.
Appl. Artif. Intell. 104, 104392.
Musavarah Sarwar: Modeling, Analysis, Application of proposed Xiao, L., Huang, G., Zhang, G., 2021. Improved assessment model for candidate design
schemes with an interval rough integrated cloud model under uncertain group
improved assessment model for health-care waste management based
environment. Eng. Appl. Artif. Intell. 104, 104352.
on dual 2-tuple linguistic rough number clouds, Applied it on a real- Rao, C., Gao, M., Wen, J., Goh, M., 2022. Multi-attribute group decision making method
world problem and proved its effectiveness with various tests, Proof with dual comprehensive clouds under information environment of dual uncertain
read the article and checked the validity of the method. Z-numbers. Inform. Sci. 602, 106–127.
Huang, G., Xiao, L., Pedrycz, W., Pamucar, D., Zhang, G., Martínez, L., 2022.
Design alternative assessment and selection: A novel Z-cloud rough number-based
Declaration of competing interest BWM-MABAC model. Inform. Sci. 603, 149–189.
Wu, Y., Chen, K., Zeng, B., Yang, M., Li, L., Zhang, H., 2017. A cloud decision
The author declares no known competing financial interests or framework in pure 2-tuple linguistic setting and its application for low-speed wind
farm site selection. J. Clean. Prod. 142, 2154–2165.
personal relationships that could have appeared toinfluence the work
Wan, S.-P., Xu, G.-L., Dong, J.-Y., 2017. Supplier selection using ANP and ELECTRE II
reported in this paper. in interval 2-tuple linguistic environment. Inform. Sci. 385, 19–38.
Akram, M., Noreen, U., Deveci, M., 2023b. Enhanced ELECTRE II method with 2-tuple
Funding linguistic m-polar fuzzy sets for multi-criteria group decision making. Expert Syst.
Appl. 213, 119237. http://dx.doi.org/10.1016/j.eswa.2022.119237.
Akram, M., Zahid, K., Kahraman, C., 2022b. A PROMETHEE based outranking approach
There is no specific funding for this project. for the construction of fangcang shelter hospital using spherical fuzzy sets. Artif.
Intell. Med. 135, 102456. http://dx.doi.org/10.1016/j.artmed.2022.102456.
Data availability Irfan, M., Elavarasan, R.M., Ahmad, M., Mohsin, M., Dagar, V., Hao, Y., 2022.
Prioritizing and overcoming biomass energy barriers: Application of AHP and
G-TOPSIS approaches. Technol. Forecast. Soc. Change 177, 121524.
No data was used for the research described in the article. Yang, S., Pan, Y., Zeng, S., 2022. Decision making framework based fermatean fuzzy
integrated weighted distance and TOPSIS for green low-carbon port evaluation.
Eng. Appl. Artif. Intell. 114, 105048.
References
Akram, M., Naz, S., Feng, F., Shafiq, A., 2022a. Assessment of hydropower plants in
Pakistan: Muirhead mean-based 2-tuple linguistic t-spherical fuzzy model combin-
Liu, H.C., Wu, J., Li, P., 2013. Assessment of health-care waste disposal methods using ing SWARA with COPRAS. Arab. J. Sci. Eng. http://dx.doi.org/10.1007/s13369-
a VIKOR-based fuzzy multi-criteria decision making method. Waste Manage. 33 022-07081-0.
(12), 2744–2751. Rong, Y., Yu, L., Niu, W., Liu, Y., Senapati, T., Mishra, A.R., 2022. MARCOS approach
Dursun, M., Karsak, E.E., Karadayi, M.A., 2011a. Assessment of health-care waste treat- based upon cubic Fermatean fuzzy set and its application in evaluation and
ment alternatives using fuzzy multi-criteria decision making approaches. Resour. selecting cold chain logistics distribution center. Eng. Appl. Artif. Intell. 116,
Conserv. Recy. 57, 98–107. 105401.
Hsu, P.F., Wu, C.R., Li, Y.T., 2008. Selection of infectious medical waste disposal firms Mishra, A.R., Chen, S.M., Rani, P., 2022. Multiattribute decision making based on
by using the analytic hierarchy process and sensitivity analysis. Waste Manage. 28 Fermatean hesitant fuzzy sets and modified VIKOR method. Inform. Sci. 607,
(8), 1386–1394. 1532–1549.
Karagiannidis, A., Papageorgiou, A., Perkoulidis, G., Sanida, G., Samaras, P., 2010. A Sarwar, M., Akram, M., Liu, P., 2021. An integrated rough ELECTRE II approach for
multi-criteria assessment of scenarios on thermal processing of infectious hospital risk evaluation and effects analysis in automatic manufacturing process. Artif. Intell.
wastes: A case study for central Macedonia. Waste Manage. 30 (2), 251–262. Rev. 54 (6), 4449–4481.
Dursun, M., Karsak, E.E., Karadayi, M.A., 2011b. A fuzzy multi-criteria group decision Khan, A., Akram, M., Ahmad, U., Al-Shamiri, M.M.A., 2023. A new multi-objective
making framework for evaluating health-care waste disposal alternatives. Expert optimization ratio analysis plus full multiplication form method for the selection
Syst. Appl. 38 (9), 11453–11462. of an appropriate mining method based on 2-tuple spherical fuzzy linguistic sets.
Pawlak, Z., 1982. Rough sets. Int. J. Comput. Inf. Sci. 11 (5), 341–356. Math. Biosci. Eng. 20 (1), 456–488.
17
M. Sarwar Engineering Applications of Artificial Intelligence 123 (2023) 106255
Akram, M., Niaz, Z., Feng, F., 2022c. Extended CODAS method for multi-attribute group Sarwar, M., Akram, M., Shahzadi, S., 2023a. Distance measures and 𝛿-approximations
decision-making based on 2-tuple linguistic fermatean fuzzy hamacher aggregation with rough complex fuzzy models. Granul. Comput. http://dx.doi.org/10.1007/
operators. Granul. Comput. http://dx.doi.org/10.1007/s41066-022-00332-3. s41066-023-00371-4.
Ghoushchi, S.J., Bonab, S.R., Ghiaci, A.M., Haseli, G., Tomaskova, H., Hajiaghaei- Sarwar, M., Ali, G., Chaudhry, R.N., 2023b. Decision-making model for failure modes
Keshteli, M., 2021. Landfill site selection for medical waste using an integrated and effect analysis based on rough fuzzy integrated clouds. Appl. Soft Comput.
SWARA-WASPAS framework based on spherical fuzzy set. Sustainability 13 (24), 136, 110148.
13950. http://dx.doi.org/10.3390/su132413950. Sarwar, M., 2020. Decision-making approaches based on color spectrum and D-TOPSIS
Haseli, G., Ghoushchi, S.J., 2022. Extended base-criterion method based on the
method under rough environment. Comput. Appl. Math. 39 (4), 1–32.
spherical fuzzy sets to evaluate waste management. Soft Comput. 26 (19),
Sarwar, M., 2022. Decision making model for design concept evaluation based on
9979–9992.
interval rough integrated cloud VIKOR. J. Ambient Intell. Humaniz. Comput.
Gokasar, I., Pamucar, D., Deveci, M., Ding, W., 2023. A novel rough numbers based
http://dx.doi.org/10.1007/s12652-022-04459-8.
extended MACBETH method for the prioritization of the connected autonomous
vehicles in real-time traffic management. Expert Syst. Appl. 211, 118445. Zhang, H., 2012. The multiattribute group decision making method based on aggrega-
Wang, J.Q., Wang, D.D., Zhang, H.Y., Chen, X.H., 2015. Multi-criteria group decision tion operators with interval-valued 2-tuple linguistic information. Math. Comput.
making method based on interval 2-tuple linguistic information and choquet Modelling 56 (1–2), 27–35.
integral aggregation operators. Soft Comput. 19 (2), 389–405. Chen, C.T., Tai, W.S, 2005. Measuring the intellectual capital performance based on
Deveci, M., Pamucar, D., Oguz, E., 2022b. Floating photovoltaic site selection using 2-tuple fuzzy linguistic information. In: Proceedings of the 10th Annual Meeting of
fuzzy rough numbers based LAAW and RAFSI model. Appl. Energy 324, 119597. Asia, Vol. 20. Pacific Region of Deicision Sciences Institute, Taiwan.
Akyurt, İ.Z., Pamucar, D., Deveci, M., Kalan, O., Kuvvetli, Y., 2021. A flight base Tai, W.S., Chen, C.T., 2009. A new evaluation model for intellectual capital based on
selection for flight academy using a rough macbeth and rafsi based decision-making computing with linguistic variable. Expert Syst. Appl. 36 (2), 3483–3488.
analysis. IEEE Trans. Eng. Manage. 1–16. http://dx.doi.org/10.1109/TEM.2021. Herrera, F., Martínez, L., 2000. A 2-tuple fuzzy linguistic representation model for
3119659. computing with words. IEEE Trans. Fuzzy Syst. 8 (6), 746–752.
Deveci, M., Brito-Parada, P.R., Pamucar, D., Varouchakis, E.A., 2022a. Rough sets based Li, D., Haijun, M., Xuemei, S., 1995a. Membership clouds and membership cloud
ordinal priority approach to evaluate sustainable development goals (SDGs) for generators. J. Comput. Res. Dev. 32 (6), 15–20.
sustainable mining. Resour. Policy 79, 103049. Li, D., Liu, C., Gan, W., 2009. A new cognitive model: Cloud model. Int. J. Intell. Syst.
Akram, M., Ali, U., Santos-Garcıa, G., Niaz, Z., 2023a. 2-tuple linguistic fermatean
24 (3), 357–375.
fuzzy MAGDM based on the WASPAS method for selection of solid waste disposal
Liu, H.C., Qin, J.T., Mao, L.X., Zhang, Z.Y., 2015. Personnel selection using interval
location. Math. Biosci. Eng. 20 (2), 3811–3837.
2-tuple linguistic VIKOR method. Hum. Factors Ergon. Manuf. Serv. Ind. 25 (3),
Akram, M., Noreen, U., Pamucar, D., 2022d. Extended PROMETHEE approach with 2–
370–384.
tuple linguistic m–polar fuzzy sets for selection of elliptical cardio machine. Expert
Zhang, H., 2013. Some interval-valued 2-tuple linguistic aggregation operators and
Syst. e13178. http://dx.doi.org/10.1111/exsy.13178.
Xiao, F., Wen, J., Pedrycz, W., 2022b. Generalized divergence-based decision making application in multiattribute group decision making. Appl. Math. Model. 37 (6),
method with an application to pattern classification. IEEE Trans. Knowl. Data Eng. 4269–4282.
http://dx.doi.org/10.1109/TKDE.2022.3177896. Cui, C.Q., Wang, B., Zhao, Y.X., Wang, Q., Sun, Z.M., 2019. China’s regional sustainabil-
Sarwar, M., Akram, M., 2023. Certain hybrid rough models with type-2 soft information. ity assessment on mineral resources: Results from an improved analytic hierarchy
J. Multiple-Valued Logic Soft Comput. 1–35, in press. process-based normal cloud model. J. Clean. Prod. 210, 105–120.
Xiao, F., 2022a. GEJS: A generalized evidential divergence measure for multisource Zhang, S., Xiang, M., Xu, Z., Wang, L., Zhang, C., 2020. Evaluation of water cycle
information fusion. IEEE Trans. Syst. Man Cybern. Syst. 1–13. http://dx.doi.org/ health status based on a cloud model. J. Clean. Prod. 245, 118850.
10.1109/TSMC.2022.3211498. Cai, B., Liu, Y., Zhang, Y., Fan, Q., Liu, Z., Tian, X., 2013. A dynamic Bayesian networks
Xiao, F., 2022c. Generalized quantum evidence theory. Appl. Intell. 1–16. http://dx. modeling of human factors on offshore blowouts. J. Loss Prev. Process Ind. 26 (4),
doi.org/10.1007/s10489-022-04181-0. 639–649.
Liu, H.C., Shi, H., Li, Z., Duan, C.Y., 2022. An integrated behavior decision-making Yazdi, M., Khan, F., Abbassi, R., Rusli, R., 2020. Improved DEMATEL methodology for
approach for large group quality function deployment. Inform. Sci. 582, 334–348. effective safety management decision-making. Saf. Sci. 127, 104705.
Akram, M., Khan, A., Luqman, A., Senapati, T., Pamucar, D., 2023c. An extended MAR-
Zarei, E., Khan, F., Yazdi, M., 2021. A dynamic risk model to analyze hydrogen
COS method for MCGDM under 2-tuple linguistic q-rung picture fuzzy environment.
infrastructure. Int. J. Hydrogen Energy 46 (5), 4626–4643.
Eng. Appl. Artif. Intell. 120, 105892.
Li, D., Du, Y., 2017. Artificial Intelligence with Uncertainty. CRC Press.
Akram, M., Ramzan, N., Deveci, M., 2023d. Linguistic pythagorean fuzzy CRITIC-EDAS
Liu, H.C., You, J.X., Lu, C., Shan, M.M., 2014. Application of interval 2-tuple linguistic
method for multiple-attribute group decision analysis. Eng. Appl. Artif. Intell. 119,
multimoora method for health-care waste treatment technology evaluation and
105777.
Akram, M., Bibi, R., Deveci, M., 2023e. An outranking approach with 2-tuple linguistic selection. Waste Manage. 34 (11), 2355–2364.
fermatean fuzzy sets for multi-attribute group decision-making. Eng. Appl. Artif. Diaz, L.F., Savage, G.M., Eggerth, L.L., 2005. Alternatives for the treatment and disposal
Intell. 121, 105992. of healthcare wastes in developing countries. Waste Manage. 25 (6), 626–637.
18