Mathematical Problems in Engineering

Mathematical Problems in Engineering / 2021 / Article
Special Issue

Advanced Intelligent Fuzzy Systems Modeling Technologies for Smart Cities

View this Special Issue

Research Article | Open Access

Volume 2021 |Article ID 6618333 | https://doi.org/10.1155/2021/6618333

Zhili Huang, Qinglan Chen, Liu Chen, Qinyuan Liu, "Relative Similarity Programming Model for Uncertain Multiple Attribute Decision-Making Objects and Its Application", Mathematical Problems in Engineering, vol. 2021, Article ID 6618333, 16 pages, 2021. https://doi.org/10.1155/2021/6618333

Relative Similarity Programming Model for Uncertain Multiple Attribute Decision-Making Objects and Its Application

Academic Editor: Yi-Zhang Jiang
Received06 Nov 2020
Revised04 Dec 2020
Accepted13 Feb 2021
Published09 Mar 2021

Abstract

This paper is concerned with the uncertain multiattribute decision-making (UMADM) of which the attribute value is triangular fuzzy number. Firstly, the max-relative similarity degree and min-relative similarity degree of alternative decision-making objects are given based on the relative similarity degree of triangular fuzzy number, the advantage relation theories to comparative relative similarity degree of triangular fuzzy number are proposed, and some good properties, relations, and conclusions are derived. Secondly, in order to determine the attribute weight vector, a triangular fuzzy number-based decision-making object relative similarity programming model is established with the help of maximizing possibility degree algorithm rules in the cooperative game theory. Subsequently, by aggregating the comparison overall relative similarity degree values of all decision-making objects, we could pick over and sort the set of alternative objects and gather a new model algorithm for the relative similarity programming of triangular fuzzy number-based multiple attribute decision-making alternatives. Finally, an example is given to illustrate the feasibility and practicability of the model algorithm presented in this paper.

1. Introduction

Uncertain multiattribute decision-making (UMADM) is one of the main contents of decision analysis, management, control engineering, and system engineering [1], and the acquisition of attribute weight information is a hot research field in UMADM. The research on uncertain multiattribute decision-making mainly focuses on supplier selection, benefit evaluation of logistics network, transportation optimization, enterprise production benefit and risk evaluation, multiagent system, waste management site selection, logistics support evaluation, and other disciplines [27]. Due to objective factors, complexity, randomness, uncertainty and disturbance of surrounding environment, subjectivity, fuzziness and preference of thinking, and empirical, stereotyped, and limited of cognition, it is usually impossible or difficult for people to express and describe the specific attribute utility measure information with accurate numbers when they are dealing with solving UMADM problems or objects, especially those with outstanding development and innovation. Therefore, their quantification usually takes the form of interval numerical value, multivariate interval value, triangular fuzzy number value, or uncertain language value. The research on this kind of UMADM problem has attracted many experts and scholars at home and abroad. Van Laarhoven, a Dutch expert who is one of the famous scholars, put forward the comparative judgment method of describing fuzzy uncertain information with triangular fuzzy number [8]. Compared with the comparative judgment [9, 10], the comparative judgment of triangular fuzzy number can more consistently meet the objective uncertainty of the environment and the subjective fuzziness of logical thinking and more scientifically and accurately describe the quality judgment of uncertain multiple attribute decision-making objects. The acquisition of attribute optimization weight information is an important research content in UMADM problem. It is easy to find that most scholars focus on the following weight measurement methods: VIKOR method [11], AHP-VIKOR method [12], relative dominance relation method [13], relative similarity degree relation method [14, 15], quadratic programming and relative superiority degree method [16], grey target decision-making method [17], grey relational analysis method [18], Grey relation-projection pursuit dynamic cluster method [19], similarity measurement of cloud model [20], rough hesitant fuzzy method based on rough sets [21], generalized complex intuitionistic fuzzy aggregation operators [22], robust averaging-geometric aggregation operators for complex intuitionistic fuzzy sets [23], Pythagorean fuzzy aggregation operators [24, 25], Pythagorean hybrid weighted aggregation operators with Pythagorean fuzzy numbers [26], distance measures of complex Pythagorean fuzzy sets [27], q-rung orthopair fuzzy point weighted aggregation operators [28], power aggregation operators for complex q-rung orthopair fuzzy sets [29], aggregation operators based on complex q-rung orthopair uncertain linguistic information [30], aggregation operators based on complex interval-valued q-rung orthopair fuzzy sets [31], and so on. Then, the relevant decision information is aggregated by combining the respective proposed models and algorithms to pick over and sort the set of alternative objects. Although literature [15] proposed a method to determine the attribute weight based on relative similarity degree relation with the help of maximizing possibility degree algorithm in cooperative game, it failed to give a specific solution process to determine the attribute weight. In the literature [32], maximizing possibility degree algorithm rules in the cooperative game theory is adopted to obtain the attribute optimization weight vector: from the perspective of picking over and sorting decision-making objects alternatives, the greater the total attribute comparative possibility degree value of decision-making objects is, the greater the assignment of attribute weight is, and the smaller the total attribute comparative possibility degree value of decision-making objects is, the smaller the assignment of attribute weight is; therefore, the comprehensive attribute value difference of all alternative decision objects assembled and merged under the optimal weighting is further expanded. On the other hand, [33] gives a too simple and loose deviation formula to express the definition of similarity based on the similarity principle, which aggregates and merges different attribute data types. For example, in the special case that the length of triangular fuzzy numbers degenerated into real numbers or interval numbers is 0, it is easy to cause the accumulation of errors in the decision-making process, which eventually leads to the failure of information fusion and the error of picking over and sorting decision-making objects alternatives.

For UMADM problem whose attribute value is triangular fuzzy number, the paper gives two new definitions: max- and min-relative similarity degree of triangular fuzzy number and decision-making object of max-relative similarity degree and min-relative similarity degree from the basic idea of measuring best-worst scaling between alternative decision objects and calculating the similarity degree of triangular fuzzy number, which better integrates the geometric position relationship of decision object attribute utility measure data sequence, the close level of curve shape and the similarity degree, and is convenient to characterize the connection and difference between decision objects, thus overcoming the problems in literature [33]. On the basis of the above definition, we introduce a new attribute weighting rule based on the relative similarity programming model of decision-making objects: after unifying the incommensurability of attribute data, the comprehensive relative similarity degree of evaluation values of decision objects under the same attribute utility measure is higher (the difference of attribute measure values is smaller), and this attribute has less influence on picking over and sorting alternatives, so the degree of emphasis should be reduced in the decision-making process, and the measure value of attribute weight should be reduced. Especially, under the same attribute utility measure, the comprehensive relative similarity of evaluation values reaches the highest; namely, it is equal to 1 (there is no difference in attribute measure values); then this attribute will have no effect on picking over and sorting decision-making alternatives, and the measure value of attribute weight should be zero. On the contrary, the comprehensive relative similarity degree of evaluation values of decision-making alternatives under the same attribute utility measure is relatively low (the difference of attribute measurement values is large), so the attribute has a greater impact on picking over and sorting the decision-making alternatives, the degree of emphasis should be increased, and the measurement value of attribute weight should be increased in the decision-making process, because the changes of attribute utility measure data is the fundamental reason that causes the change of the order and selection of decision objects. In this paper, using the abovementioned weighting thought and the advantage relation theories to comparative relative similarity degree of triangular fuzzy number, under the attribute weight assignment based on decision making objects relative similarity programming model, the comprehensive relative similarity degree value between decision-making objects is reduced, and the sort difference between decision-making objects is increased, by reconciling the information of all decision-making objects. According to the overall relative similarity degree value between each decision object and the ideal optimal object in the set of decision objects to pick over and sort, then decision-making object relative similarity programming model algorithm of UMADM is obtained.

In a word, the main work of this paper is to fully consider the influence of the relative similarity measurement information between attribute values on the attribute itself in the evaluation process such as UMADM and redesign and construct a new decision-making object relative similarity programming model (DMORSPM) based on the triangular fuzzy number attribute value comparative similarity relationship to obtain more optimized attribute weighting formula. In order to solve the problem of low overall discrimination after the feature, information of decision-making object is gathered and fused. The main advantage of TFN-DMORSPM algorithm proposed in this paper is that it can expand the discrimination between decision-making objects under the condition of processing data information and very high similarity degree of attribute measurement values, facilitate the accurate calculation and aggregation of data, and promote the optimal decision-making and sorting of decision-making objects.

2. The Advantage Relation Theories to Comparative Relative Similarity Degree of Triangular Fuzzy Number

2.1. Comparative Relative Similarity Degree of Triangular Fuzzy Number

Definition 1. If , is a triangular fuzzy number (TFN) [7, 8, 15], where and are the lower and upper bounds of triangular fuzzy number, which are generally called small elements and large elements. is the median value supported by triangular fuzzy number (i.e., information tendency preference value, which usually refers to the number with the highest probability of being selected in this interval), generally called special element. If also satisfies , then is called a normalized triangular fuzzy number. If or , degenerates into an interval number or , in particular, if , then is reduced to an ordinary positive real number; that is, .
For the convenience of follow-up analysis, the operation rules of triangular fuzzy numbers are introduced: let , , and then we have(i)Rule 1: (ii)Rule 2: (iii)Rule 3: (iv)Rule 4: (v)Rule 5: if and only if , , , then

Definition 2. Let two arbitrary triangular fuzzy numbers and . Then,is called Max-relative similarity degree between triangular fuzzy numbers and , oris called Min-relativesimilarity degree between triangular fuzzy numbers and . Apparently, the greater the or , the greater the similarity degree between and . In special cases, when or , one has; that is, the triangular fuzzy number is completely similar to .
According to the max-relative and min-relative similarity degree of triangular fuzzy numbers given by Definition 2, the following properties can be revealed.

Theorem 1. Let any given triangular fuzzy number be set as , , ; then we have the following:(1)Boundedness, , (2)Reflexivity, , (3)Symmetry property, , , (4)Transitivity, , , , , , , that is, if and are completely similar, and are completely similar, then and are completely similar(5)If , , , then is closer to compared with ; if , , , , then is closer to compared with According to the definition of max-relative and min-relative similarity degree of triangular fuzzy numbers, it is easy to prove that the conclusion of Theorem 1 is valid. The process of proof is omitted.

Definition 3. Let alternative decision-making objects be composed of triangular fuzzy number sequence set as , . Then, one haswhere is called max-relative similarity degree based decision-making objects and ,andwhere is called min-relative similarity degree based decision-making objects and .
By letting normalized triangular fuzzy number-based decision-making objects matrix , where , , , we can give the following definition.

Definition 4. is the triangular fuzzy number-based positive ideal decision-making objects composed of positive ideal point sequence, whereand is the triangular fuzzy number-based negative ideal decision-making objects composed of negative ideal point sequence

2.2. The Advantage Relation to Comparative Relative Similarity Degree of Triangular Fuzzy Number

According to the concept of max-relative similarity degree and min-relative similarity degree of triangular fuzzy number, the definition and main results of the advantage relation of comparative relative similarity of triangular fuzzy number sequence and triangular fuzzy number are as follows.

Definition 5. Let any two triangular fuzzy numbers and and two triangular fuzzy number-based positive and negative ideal points and ; ifthen the triangular fuzzy number is called superior to , which is denoted as . Obviously, the greater the max-relative similarity degree with the positive ideal point triangular fuzzy number or the smaller the min-relative similarity degree with the negative ideal point triangular fuzzy number is, the greater the dominance degree of the corresponding triangular fuzzy number is.

Theorem 2. If and only if the positive and negative ideal points of triangular fuzzy number are optimal decision points, then

Proof. Obviously, according to Definition 5,When positive and negative ideal points are optimal decision points, and are the ideal point triangular fuzzy number; and hence we haveand then we can getAccording to formulas (1) and (2), it can be derived thatTherefore,Thus, formulas (8) hold. Proof is completed.
We can use results in Theorem 2 to compare the dominance relation between different triangular fuzzy numbers, 2 that is, by calculating the max-relative similarity degree or min-relative similarity degree between triangular fuzzy numbers and ideal points or by comparing the sum of attribute value relations of small, special, and large elementsof triangular fuzzy numbers.

Definition 6. Let alternative decision-making objects composed of triangular fuzzy number sequence are and . The triangular fuzzy number-based positive and negative ideal decision-making object composed of positive and negative ideal point sequence is defined as and , where, , , , , ; ifthen the alternative decision-making object is called superior to , defined as . Obviously, the greater the max-relative similarity degree with the positive ideal decision-making object of triangular fuzzy number is or the smaller the max-relative similarity degree with the negative ideal decision-making object of triangular fuzzy number is, the greater the superiority of the alternative decision-making objects is.

Theorem 3. If and only if positive and negative ideal decision-making objects of triangular fuzzy number are the optimal decision-making objects,

Proof. Obviously, according to Definition 6,As and are, respectively, triangular fuzzy number based positive and negative ideal decision-making objects composed of positive and negative ideal points sequence, and are triangular fuzzy number-based positive and negative ideal points; it can be seen thatand, therefore,According to formulas (3) and (4), we can getAs a result,Now, we can see that formula (15) holds, which completes the proof.
The above theoremcan be exploited to compare the advantages between alternative decision objects, by comparing the Max-relative or Min-relative similarity degree between alternative decision objects and ideal decision objects, the sum of attribute value sequence relations of triangular fuzzy number small, special, and large elements of alternative decision objects.
In the process of analyzing the advantages and disadvantages of the UMADM problem with unknown attribute weight and without any decision object preference we could assume that the set of alternative decision objects in the decision space is . It is generally considered that the greater the max-relative similarity degree between the alternative decision-making objects and the positive ideal optimal decision-making objects is, the smaller the min-relative similarity degree between the alternative decision-making objects and the negative ideal optimal decision-making objects is, and the better the corresponding alternative decision-making object is. However, it is pointed out in [14] that the alternative decision-making objects are not necessarily close to the positive ideal optimal object and far away from the negative ideal optimal object at the same time. In order to obtain the optimal approach point to the positive and negative ideal decision-making objects, a concept of overall relative similarity degree [7, 15] (Definition 7) is introduced to represent the degree of relative difference between an alternative decision-making object approaching the optimal ideal point, namely, the point close to the positive ideal object, and that far away from the negative ideal optimal object point.

Definition 7. Let the decision-making object composed of triangular fuzzy number sequence be , and the positive and negative ideal decision-making objects composed of positive and negative ideal point sequence of triangular fuzzy number be, respectively, and ; thenwhere is the overall relative similarity degree of the comparison of Max-relative and Min- relative similarity degrees in the set of the alternative decision-making objects between the selected decision-making objects and the positive and negative ideal decision-making objects.

Theorem 4. , .

Proof. According to formula (21), it is easy to knowWe can getwhich completes the proof.
If the alternative decision object also satisfiesthen ; whichreaches the maximum. At this time, the max-relative similarity between the alternative decision-making object and the positive ideal optimal decision-making object is the largest and the most approximate, while t, the max-relative similarity between the alternative decision-making object and the positive ideal optimal decision-making object, is the smallest and farthest away. If gradually decreases,the alternative decision object deviates from the positive ideal optimal object point and gets close to the negative ideal optimal object point, making it difficult to meet the requirements of decision-makers. Therefore, the overall relative similarity degree given in this paper excels in picking and sorting the alternative decision objects . Alternative decision objects are better than , marked as ().

3. A Triangular Fuzzy Number-Based Decision-Making Object Relative Similarity Programming Model

For the UMADM problem with unknown index weight and no preference for decision-making objects, we combine the relevant theory of the weighting algorithm for maximizing the comparative possibility in cooperative game theory [23] with the advantage relation theories to comparative relative similarity degree of triangular fuzzy number presented in this paper. By considering the risk attitude of decision makers, a new attribute weighting rule based on relative similar programming model of triangular fuzzy number decision-making objects is proposed from the perspective of determining the advantages and disadvantages of the alternatives [7, 15]: after eliminating the influence of different physical dimensions, namely, unifying the incommensurability between different attribute utility measures, if the comprehensive relative similarity between all alternative decision-making objects under the same attribute utility is smaller (i.e., the difference between attribute measure values is greater), the attribute is weighted. In the contrary, if the comprehensive relative similarity of all candidate decision-making objects is greater (i.e., the difference of attribute measure values is smaller) under the same attribute utility, the weighting value of attributes should be reduced accordingly. In particular, if the comprehensive relative similarity of all the candidate decision-making objects synthesized among the measure values under the same attribute utility reaches the maximum—that is, it is equal to 1 (i.e., there is no difference in attribute measure values)—then the attribute weighting value can be assigned to zero accordingly. From the perspective of similarity measure principle, the comparative relative similarity value assembled between the attribute measure values of the alternative decision objects, that is, the difference degree of the attribute measure value data, is the main basis and key factor for judging the quality of the decision results. Next, the paper will establish a comparative relative similarity planning model by using the attribute utility measure value information of the decision objects and determine the optimal attribute weight vector by solving it. Then, the decision information gathered and fused under the weight of the vector can not only reduce the similarity among the alternative decision objects but also expand the difference and discrimination among the alternative decision objects, which is more conducive to the comparative judgment, and picking and sorting of the optimal decision objects.

In the process of judging the merits and demerits of decision objects for a UMADM problem, it is assumed that all alternative decision-making objects in the utility measurement of each attribute will obtain a matrix composed of the measured values of the attributes of all the initial decision objects about , which is called the initial triangular fuzzy number decision matrix. Here, where , . Let denote the subscript set of the most common benefit attributes and cost benefit. It is easy to see. In order to fuse the incommensurability and contradiction between the utility measure data of different attributes and eliminate the influence of different physical dimensions on the judgment and ranking results of decision objects, we adopt the following formulas to transform the initial triangular fuzzy number decision matrix into the standard triangular fuzzy number decision matrix [7, 15]:where is a normalized triangular fuzzy number and is norm of a vector, , and. According to the operation relation rule of triangular fuzzy numbers, formulas (25) and (26) can be changed into the following formulas:

Considering the risk attitude of decision-makers, we apply the theory of the advantage relation theories to comparative relative similarity degree of triangular fuzzy number and investigate the jth attribute, , and the triangular fuzzy number decision matrix . The comprehensive relative similarity degree between the alternative decision objects and other alternative decision objects is

Therefore, for the th attribute, the total comprehensive relative similarity degree between the alternative decision objects and other alternative decision objects is

For a UMADM problem whose attribute weight information is completely unknown, we can assume its attribute weight vector is expressed by , , , and satisfyingthe constraints .

According to the attribute weighting rules based on decision-making object relative similarity programming model proposed in this paper, under the consideration of the risk awareness of decision-makers and unknown attribute weights and no preference for decision objects, the determination of the optimal solution of weighting vectors should make the weighted sum of the total comprehensive relative similarity of all attributes to all alternative decision objects minimum under the action of weighting vectors . In other words, the optimal solution of the weighting vector should be determined so that the weighted sum of the reciprocal of the total comprehensive relative similarity of all attributes to all alternative decision objects must be the largest under the action of the weighting vector . In order to obtain the optimal weight vector , we can construct the triangular fuzzy number-based decision-making object relative similarity programming model of UMADM (TFN-DMORSPM):

By solving the optimization model, the optimal solution is

In order to keep consistent with the traditional normalization usage, we will normalize the unit weight vector lettingand then we can get

According to formula (33), it is easy to know that under the same attribute measure; the sum of the comprehensive relative similarity values compared among all the candidate decision objects is inversely proportional to the size of the attribute weight value.

Remark 1. The values of comprehensive relative similarity degree in formulas (29) ∼ (33) are as follows:

The choice of value depends on the risk attitude of decision-makers: when , decision-makers tend to avoid risks; when , decision-makers tend to be risk-neutral; when , decision-makers tend to risk preference. In particular, when , , is conservative max-relative similarity degree; when , , , is adventurous min-relative similarity degree.

4. Implementation Steps and Examples of the Model

The implementation steps of UMADM’s triangular fuzzy number-based decision-making objects relative similarity programming model (TFN-DMORSPM) given in this paper are as follows.Step 1. To fuse the incommensurability and contradiction between the utility measure data of different attributes which eliminates the influence of different physical dimensions on the decision-making object’s judgment and ranking results, the initial triangular fuzzy number type decision matrix is converted into a normalized triangular fuzzy number type decision matrix according to formulas (27) and (28), where is a normalized triangular fuzzy number.Step 2. Analyze and measure the normalized triangular fuzzy number based decision-making matrix which can reflect the attribute eigenvalue information of all alternative decision-making objects and calculate the max-relative similarity degree and the min-relative similarity degree between the attribute values of each decision-making object by using formulas (1) and (2). According to the constructed TFN-DMORSPM, the attribute weight measurement value is obtained by calculating formulas (29)~(34) and by applyingthe normalized triangular fuzzy number-based decision-making matrix inStep 1, we can get the constructed matrixwhich is the weighted normalized triangular fuzzy number decision-making matrix.Step 3. According to the weighted normal triangular fuzzy number type decision-making matrix obtained in Step 2, the triangular fuzzy number-based positive and negative ideal decision-making objects and composed of positive and negative ideal point sequences can be obtained by using formulas (5) and (6).Step 4.Calculate all the max-relative similarity degree and min-relative similarity degree of alternative decision-making objects and triangular fuzzy number-based positive and negative ideal decision-making objects by using formulas (3) and (4).Step 5. Using formula (21) to calculate the overall relative similarity degree by all the max-relative similarity and min-relative similarity based positive and negative ideal decision-making objects , and alternative decision making objects in the alternative set.Step 6.According to the overall relative similarity degree value , pick over and sort the set of the alternative decision-making objects in descending order.

Example 1. In order to illustrate the practicability and effectiveness of TFN-DMORSPM, in this paper we uses the example of scheduling problem about carrier-based aircraft models in literature [34] to analyze. The performance parameters affecting carrier-based aircraft models, i.e. maximum speed (), free voyage across the sea (), maximum net load (), purchase cost (), reliability (), and maneuverability (), We consider that there are four types of aircraft to choose. It is assumed that after standardized statistical processing, the utility measure values of performance parameters of various models under various indexes (attributes) are expressed in the form of triangular fuzzy numbers (all the given data is benefit data), as shown in Table 1.Step 1. In order to unify the incommensurability and contradiction of different physical dimensions among the utility measures of initial attributes, we assume that the normalized triangular fuzzy number decision matrix obtained by standardized statistical processing of (27) and (28), is shown in Table 1.Step 2. According to the standard triangle fuzzy number decision matrix composed of attribute utility measure value data in Table 1, the max-relative similarity and min-relative similarity is obtained by (1) and (2), and then the weight measurement value of attributes is calculated by formulas (29)–(34) according to TFN-DMORSPM for simplicity, we let; the decision-makers tend to be risk neutral) as follows:</