1. Introduction
The multiple attribute decision making (MADM) problem is a significant area of decision science, whose theories and methods are widely used in engineering, economics, management, the military and many other fields. Generally, decision makers will provide an evaluation of each alternative for every attribute or criterion according to their own cognitive beliefs. The main task of solving MADM problems is sorting a group of choices and finding the best one based on decision information provided by decision makers.
Since being proposed by Zadeh [
1] in 1965, fuzzy theory has been widely used with applications in various areas. However, the fuzzy theory can only express membership—Non-membership cannot be represented. The intuitionistic fuzzy set (IFS) presented by Atanassov [
2], which is an important extension of traditional fuzzy sets, contains membership degree and non-membership degree. Because of the ambiguity of objects and the uncertainty of human thought and cognition, decision makers have difficulty using crisp numbers to evaluate relevant decision making problems, such as student assessment and car performance evaluation. Decision makers are more accustomed to making evaluations directly in linguistic terms, such as good, generally, and not good. Therefore, many methods and models have been developed to solve real problems based on linguistic variables [
3,
4,
5,
6]. For example, since the type 2 fuzzy sets (T2FSs) could better represent the indeterminacy and simplify the calculation process, the interval type 2 hesitant fuzzy sets (IT2HFSs) can reflect the uncertainty of inaccurate information more effectively. Deveci et al. [
7] proposed a method including T2FSs and the IT2HFSs to access airlines’ service quality. Finally, accurate data results and practical implications were obtained.
Herrera and Martinez [
8] proposed the concept of 2-tuples made up of a linguistic variable and a numerical value to prevent loss of information when addressing MADM problems. Subsequently, many operators and methods based on 2-tuples have been proposed. On the basis of the power average (PA) operator, Xu and Wang [
9] studied several 2-tuple linguistic power average (2TLPA) operators which could alleviate the impact of partial arguments on the aggregated consequences. Furthermore, the method proposed in the paper considered all the decision parameters and the interrelationships of each other. However, there is a slight disadvantage in that it ignores the relationship between the two parameters. In addition, considering the significance of different parameters, a 2-tuple linguistic weighted PA (2TLWPA) operator was proposed. Wei and Zhao [
10] came up with series aggregation operators according to 2-tuple linguistic information and a dependent operator that eliminates the influence of unjust 2-tuple linguistic parameters on the aggregation results. Jiang and Wei [
11] developed a 2-tuple linguistic Bonferroni mean (2TLBM) operator based on the Bonferroni mean (BM)operator and a 2-tuple linguistic weighted BM (2TLWBM) operator to account for the different importance of the input parameters. Merigó et al. [
12] introduced some aggregation operators based on 2-tuple linguistic information that provide a more complete understanding of the situation being considered. Moreover, the authors also studied the applicability of the novel method in different fields. A modified composite scale that can enhance the precision of decision making was developed by Wang et al. [
13] to overcome the limitation of the 2-tuple linguistic representation model. Qin and Liu [
14] proposed several operators based on 2-tuple linguistic information and the Muirhead mean (MM) operator. It is known as a mean type aggregation operator that can utilize the intact relation between the multi-input parameters. Meanwhile, they applied the method proposed in the paper for supplier selection.
As the decision environment and content become increasingly complex, the use of 2-tuple linguistic variables alone fails to accurately describe ambiguous and fragmentary cognitive information. Cuong [
15] developed the picture fizzy set (PFS) to express uncertain cognitive information characterized by three degrees: A positive membership degree
, a neutral membership degree
and a negative membership degree
. Therefore, PFS allows several types of answers when solving decision making problems, such as yes, abstain, no, and refusal. Many research achievements have been made in the field of PFS theory. Singh [
16] applied the correlation coefficient to clustering analysis where the attribute values are in the form of PFS. Because the PFS contains more information about people’s evaluation than IFS, the proposed correlation coefficients are a further generalization of IFSs. Yang et al. [
17] proposed picture fuzzy soft sets and studied their relevant properties. In particular, there is a method based on adjustable soft discernibility matrix which could obtain a sequential relationship between all objects. Son [
18] proposed a generalized distance measure for pictures and the method of hierarchical picture clustering (HPC). Wei [
19] proposed picture fuzzy cross entropy to address the MADM problem which can reflect the fuzziness of subjective judgment easily. Thong and Son [
20] developed a novel hybrid model including picture fuzzy clustering and intuitionistic fuzzy recommender systems that are applicable to health care support systems. These models not only improve the accuracy of medical diagnosis but also guarantee the development of a medical security system. But the limitations of these models are the time complexity and the capability of the model when new patients are added to the system.
Archimedean
t-norms and
t-conorms (ATT) are types of
t-norms and
t-conorms that have become important tools for explaining the conjunction, and the operational rules have been defined. Beliakov et al. [
21] used ATT to calculate the IFS, thus simplifying and extending the existing constructions. Liu [
22] developed single-valued neutrosophic number operators based on ATT which are able to extend to most of the existing
t-norms and
t-conorms and single-valued neutrosophic numbers (SVNNs). Liu [
23] developed some operators based on ATT and PFS and studied several properties and particular cases of the operators.
The aggregation operator is a crucial tool for addressing MADM problems. Many effective aggregation operators have been developed for situations where the input arguments have some relations. Yager [
24] introduced the PA operator. In the process of aggregation, parameter values support each other. Tan and Chen [
25] investigated the induced Choquet ordered averaging (I-COA) operator and demonstrated its relationship to the induced ordered weighted averaging operator. Bonferroni [
26] developed the Bonferroni mean (BM) operator, which can effectively address the relationships among input parameters. Liu et al. [
27] presented several intuitionistic uncertain linguistic Bonferroni OWA (IULBOWA) operators that can aggregate max and min operators and introduced relevant score functions, accuracy functions, and comparative methods. Li and Liu [
28] proposed novel aggregation operators according to the Heronian mean (HM) operator that considered the interrelationships of attribute values. BM and HM operators can only account for relationships between input arguments and not the correlation between multiple arguments. To overcome this limitation, Maclaurin [
29] proposed a Maclaurin symmetric mean (MSM) operator to capture the relationships among multiple input arguments. Qin and Liu [
30] solved MADM problems based on MSM operators under a hesitant fuzzy environment. Wang et al. [
31] extended MSM aggregation operators with single-valued neutrosophic linguistic variables and developed methods for multiple-criteria decision making (MCDM). Wei and Lu [
32] proposed the Pythagorean fuzzy MSM (PFMSM) and Pythagorean fuzzy weighted MSM (PFWMSM) operators and discussed their desirable properties. Liu and Zhang [
33] extended MSM operators with the single-valued trapezoidal neutrosophic number (SVTNNs) to not only account for the correlation between multi-input arguments but also conveniently depict uncertain information in the decision making process.
Inspired by the above analysis, although the PFS can address complex and uncertain problem flexibly, PFS has difficulty expressing cognitive information. Therefore, we use the picture 2-tuple linguistic set based on PFS and 2-tuple linguistic information to address MADM problems, thereby overcoming the above limitation and preventing loss of information in the calculation and aggregation processes. In addition, we apply the ATT to address MADM problems described by picture 2-tuple linguistic numbers (P2TLNs). Then, we extend MSM operators under a P2TLN environment, such as ATT-P2TLMSM and ATT-P2TLGMSM, to capture the interrelationships among multiple input parameters. In cases where the input parameters have different significances, ATT-P2TLWMSM and ATT-P2TLGWMSM are proposed. Based on the above operators, we propose a method to handle MADM problems.
The framework of this paper is as follows. In
Section 2, we explain several basic concepts and theories. In
Section 3, a novel operation for picture 2-tuple linguistic sets based on ATT is proposed. In
Section 4, we develop novel P2TLMSM operators. In
Section 5, we present models based on the ATT-P2TLWMSM operator and the ATT-P2TLGWMSM operator to solve the MADM problems. Finally, an expositive instance is provided in
Section 6.
4. Picture 2-Tuple Linguistic MSM Operators Based on ATT
Based on the new operations for picture 2-tuple linguistic sets and MSM operators, we developed a few novel P2TLMSM operators.
4.1. The ATT-P2TLMSM and ATT-P2TLGMSM Operators
Definition 17. Let be a set of P2TLNs. Then, the ATT-P2TLMSM operator is as follows.whereis a collection of P2TLNs and.
According to the operation rules of P2TLNs in Definition 16, the ATT-P2TLMSM operators are shown below.
Theorem 1. Let be a collection of P2TLNs with . The result of aggregating by Definition 17 is still a P2TLN.wheretraverses all m-tuple combinations ofandis the binomial coefficient. Proof. The flowchart of proof is shown in
Figure 1.
The specific proof process is as follows.
Because
Therefore, Theorem 1 proved to be correct. □
Property 1. Let and be sets of P2TLNs. then has several properties.
- (1)
Idempotency: If the P2TLNsfor all, then.
- (2)
Commutativity: Assumeis a permutation offor all; then,.
- (3)
Monotonicity: If,,andfor each, thenand.
- (4)
Boundedness: Ifand, then.
Proof. Since each
, that is,
This property is obvious, and we do not prove it here.
If , , , and for each and ,, according to idempotency, and .
According to idempotency, let
and
. Based on the monotonicity, if
and
for each
, then we have
and
. Therefore, the following conclusion can be obtained.
□
In the following, we present a detailed formula as an example to introduce the
P2TLMSM operator in the context MADM. When
, based on Formula (10), we can obtain:
Next, we study some special cases of the P2TLMSM operator with respect to the parameter m.
- (1)
When
m = 1, Equation (10) degrades to the following formula.
- (2)
When
m = 2, Equation (10) degrades to the following formula.
- (3)
When
m =
n, Equation (10) degrades to the following formula.
Definition 18. Let be a set of P2TLNs. Then, the ATT-P2TLGMSM operator is as follows.whereis a collection of P2TLNs and.
On the basis of the operation rules of P2TLNs in Definition 16, the ATT-P2TLMSM operators are presented below.
Theorem 2. Let be a set of P2TLNs and . Then, the aggregation result from Definition 17 is also a P2TLN.wheretraverses all m-tuple combinations ofandis the binomial coefficient. Because the proof is analogous to Theorem 1, we do not repeat it here.
Property 2. Let and be collections of P2TLNs. then has a number of properties.
- (1)
Idempotency: If the P2TLNsfor all, then.
- (2)
Commutativity: Assumeis a permutation offor all; then,.
- (3)
Monotonicity: If,,andfor each, thenand.
- (4)
Boundedness: Ifand, then.
Because the proof is analogous to Property 1, we do not repeat it here.
In the following, we present a detailed formula as an example to introduce the
P2TLMSM operator in the context of MADM. When
, based on Formula (16), we obtain:
Next, we study several specific cases of the P2TLGMSM operator with respect to the argument m.
- (1)
When
m = 1, Equation (17) degrades to the following formula.
- (2)
When
m = 2, Equation (17) degrades to the following formula.
- (3)
When
m =
n, Equation (17) degrades to the following formula.
4.2. The ATT-P2TLWMSM and ATT-P2TLWGMSM Operators
We now introduce the weighted ATT-P2TLMSM and ATT-P2TLGMSM operators to improve the decision making accuracy.
Definition 19. Let be a set of P2TLNs and be the weight vector, where represents the importance degree of , satisfying with . Then, the ATT-P2TLWMSM operator is as follows.whereis a set of P2TLNs and.
On the basis of the operation rules of P2TLNs in Definition 16, the ATT-P2TLWMSM operators are shown below.
Theorem 3. Let be a set of P2TLNs and . Then, the result of aggregating via Definition 19 is also a P2TLN.wheretraverses all m-tuple combinations ofandis the binomial coefficient. The proof of Theorem 3 is analogous to that of Theorem 1, and we do not repeat the proof here. Property 3. Let and be sets of P2TLNs. then has a number of properties.
- (1)
Monotonicity: If,,andfor each, thenand.
- (2)
Boundedness: Ifand, then.
The proof is analogous to that of Property 1 and is therefore omitted.
In the following, we present a detailed formula as an example to introduce the
P2TLMSM operator in the context of MADM. When
, based on Formula (22), we obtain:
If we consider some specific values of m, the following formulas can be obtained.
- (1)
When
m = 1, Equation (23) degrades to the following formula.
- (2)
When
m = 2, Equation (23) degrades to the following formula.
- (3)
When
m =
n, Equation (23) degrades to the following formula.
Definition 20. Let be a set of P2TLNs and be the weight vector, where represents the importance degree of , satisfying with . Then, the ATT-P2TLGWMSM operator is as follows.whereis a collection of P2TLNs and.
On the basis of the operation rules of P2TLNs in Definition 16, the ATT-P2TLGWMSM operator is presented below.
Theorem 4. Let be a collection of P2TLNs and . Then, the result of aggregating with Definition 20 is also a P2TLN.wheretraverses all m-tuple combinations ofandis the binomial coefficient. The proof of Theorem 4 is similar to that of Theorem 1 and is therefore omitted. Property 4. Let and be sets of P2TLNs. has the following important properties.
- (1)
Monotonicity: If,,andfor each, thenand.
- (2)
Boundedness: Ifand, then.
The proofs are analogous to those of Property 1 and are therefore omitted.
In the following, we present a detailed formula as an example to introduce the
P2TLGWMSM operator in the context of MADM. When
, based on Formula (28), we obtain:
If we consider some special values of m, the following formulas are obtained.
- (1)
When
m = 1, Equation (29) degrades to the following formula.
- (2)
When
m = 2, Equation (23) degrades to the following formula.
- (3)
When
m =
n, Equation (23) degrades to the following formula.
5. MADM Based on the ATT-P2TLMSM Operator
Based on the ATT-P2TLWMSM and ATT-P2TLGWMSM operators, in this section, we address the MADM problems in which the attribute preference values take the form of picture 2-tuple linguistic variables.
Let be a discrete set of alternatives, be the set of attributes and be the weighting vector of the attributes , where , . For the alternative of the attribute , the decision maker provides an attribute value , which is a picture 2-tuple linguistic variable. Each attribute value constitutes the decision matrix .
Next, we apply the
ATT-P2TLWMSM and
ATT-P2TLGWMSM operators to solve MADM problems in which the attribute values take the form of P2TLNs. The flowchart of the method is shown in
Figure 2.
Step 1. Aggregate all P2TLNs via the ATT-P2TLWMSM or ATT-P2TLGWMSM operator to derive the aggregation results of the alternatives .
Step 2. Calculate the scores of the P2TLNs and rank the alternatives . If is equal to , the accuracy degrees and must be calculated. Then, rank the alternatives according to and .
Step 3. Sort the alternatives and select the best choice with .
Step 4. End.