Multi-attribute group decision-making method applied to selection bridge design

By combining the cross-grained semantic vector distance measurement with hesitation and projection value, the problem of large amount of calculation and projection value in the prior art is solved, and the calculation amount and projection value are not considered, achieving more accurate inter-expert distance calculation and Group decision results.

CN118966232APending Publication Date: 2024-11-15DALIAN UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410956612.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing hesitant fuzzy language term set distance measurement has different elements in bridge design and needs to be filled, which leads to large calculations and cannot effectively consider projection values, resulting in large distance differences between experts but the preference sorting is the same. precise.

Method used

A multi-attribute group decision-making method for cross-grained semantic vector distance is proposed. By defining the binary vector concept of hesitant fuzzy language term set, the expert's hesitant degree, distance of hesitant fuzzy language term set and the projection value are combined, and the semantic vector distance is adopted. The distance between different experts is calculated and the final group decision matrix is ​​obtained through clustering and weight calculation.

Benefits of technology

It improves the accuracy and rationality of multi-attribute group decisions, eliminates the situation where there is a large distance difference between experts but the preference for the same sorting, reduces the amount of calculation, ensures data integrity, and obtains more reasonable and accurate decision results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118966232A_ABST
    Figure CN118966232A_ABST
Patent Text Reader

Abstract

The invention provides a multi-attribute group decision-making method applied to cross-granularity semantic vector distance for selecting bridge design, which comprises the following steps: S1, determining evaluation attributes and schemes, and giving an evaluation matrix corresponding to a hesitant fuzzy language term set form by an expert; s2, converting a hesitant fuzzy language term set given by an expert into a binary vector form based on the hesitant fuzzy language term set provided by the invention; s3, according to the semantic vector distance measurement proposed by the invention, solving the distance between different experts, and carrying out clustering; s4, calculating to obtain the weight of each expert in the class; s5, determining a clustered evaluation matrix according to the expert evaluation matrix and the expert weight; s6, calculating to obtain the weight of each cluster; s7, fusing the group decision matrix of each cluster by using the inter-class weight to obtain a final large group decision matrix; and S8, carrying out final scheme sorting by using an approximate positive and negative ideal solution. The method has the advantage of comprehensively considering multiple factors, and can improve the accuracy and reliability of decision making.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of group decision making, and in particular to a multi-attribute group decision making method for selecting cross-granularity semantic vector distances for bridge design. Background Art

[0002] Decision making can be viewed as a problem-solving process to find the best or satisfactory solution among feasible alternatives. Multi-attribute decision making is a branch of operations research. It is widely used in decision making by explicitly evaluating feasible alternatives on multiple conflicting criteria to determine the best solution. In the current bridge design example, we are faced with the challenge of selecting a suitable bridge design scheme. It is essentially a multi-attribute group decision-making problem. The evaluation method is an important exogenous force that affects the decision maker's choice of bridge design scheme. The selection of an effective evaluation method is a key factor for decision makers to select a bridge design scheme that best suits the needs and has good performance.

[0003] In order to describe decision information, in 2012, Rodríguez proposed the concept of hesitant fuzzy language term set based on hesitant fuzzy set and language term set, which can characterize the situation that decision makers are indecisive and difficult to choose between several consecutive language terms. For example, for the taste of food in a restaurant, customers may use words such as "very delicious", "average", "not very delicious" to express their evaluation. Hesitant fuzzy language term set is based on the language expression given by people, which is closer to people's thinking and cognition, can flexibly and comprehensively reflect the real preferences of decision makers, and is more conducive to experts expressing the evaluation information of each attribute in the decision. Therefore, hesitant fuzzy language term set is widely used in multi-attribute group decision-making problems.

[0004] However, there are still some problems with the distance measurement of hesitant fuzzy language term sets: most distance measures are not applicable to hesitant fuzzy language term sets of different granularities; the hesitation is not fully considered; the hesitant fuzzy language term sets with different numbers of elements need to be padded before calculating the distance, which causes a large amount of calculation; when the distance difference between expert evaluation matrices is large but the preference ranking is the same, the hesitant fuzzy language term set does not consider the influence of the projection value. Summary of the invention

[0005] The present application mainly aims at the problem that the existing hesitant fuzzy language term set distance measure has the problem that when selecting a bridge design plan, the distance calculation needs to fill in the elements when the number of elements in the hesitant fuzzy language term set is different, and when making a large group decision, there is a large difference in distance between experts, but the preference ranking is the same. The distance between experts calculated using the existing distance measure will be inaccurate. A multi-attribute group decision-making method for cross-granularity semantic vector distance applied to bridge design selection is provided, a binary vector concept based on the hesitant fuzzy language term set is proposed to consider the factor of projection value, and then a semantic vector distance measure is proposed to combine the expert's hesitation, the distance of the hesitant fuzzy language term set and the projection value, so that the proposed semantic vector distance is applicable to the distance calculation of hesitant fuzzy language term sets of different granularities and different numbers of elements, and it is used to solve the intra-class weight, inter-class weight and comprehensive evaluation value in large group decision-making, so that the final result will be more reasonable and accurate.

[0006] The technical solution adopted by the present invention to solve the technical problem is:

[0007] A multi-attribute group decision-making method for selecting a bridge design comprises the following steps:

[0008] S1: Determine the evaluation attributes and schemes, and the experts give the evaluation matrix in the form of the corresponding hesitant fuzzy language term set;

[0009] S2: converting the hesitant fuzzy language term set given by the expert into a binary vector form based on the hesitant fuzzy language term set proposed in this application;

[0010] S3: Calculate the distance between different experts according to the semantic vector distance measure proposed in this application and perform clustering;

[0011] S4: Calculate the weight of each expert in the class;

[0012] S5: Determine the clustering evaluation matrix according to the expert evaluation matrix and expert weights;

[0013] S6: Calculate the weight of each cluster;

[0014] S7: Use the inter-class weights to fuse the group decision matrix of each cluster to obtain the final large group decision matrix;

[0015] S8: Use the method of approximating positive and negative ideal solutions to sort the final solutions.

[0016] Furthermore, step S1 is specifically as follows:

[0017] The language term set can be expressed as S = {s i |i=1,2,...,g}, where i represents the granularity level; s iA linguistic term for i-level granularity, with granularity g and order, i.e., if θ < τ, then s θ <s τ ;

[0018] Let S = {s1, s2, ..., s g}, then the hesitant fuzzy language term set H s It can be defined as a set of finite ordered continuous terms in S, namely:

[0019] H S ={s i ,s i+1 , ..., s j |s k ∈S,k=i,i+1,...,j};

[0020] Among them, i and j represent the lowest and highest granularity levels of the hesitant fuzzy language term set, respectively.

[0021] Furthermore, the binary vector form based on the hesitant fuzzy language term set in step S2 is:

[0022] In the decision-making process, when experts evaluate the same scheme attributes, if the number of elements in the hesitant fuzzy language term set used is different, the meaning represented will be different, so it is necessary to consider the degree of hesitation.

[0023] First, we define the hesitant fuzzy language term set H S The hesitation μ(H s ):

[0024]

[0025] For a language term set S = {s1, s2, ..., s g}, let the language term in S The semantics of Then the language terms in S evenly cover [0, 1]. S Indicates the number of elements of HFLTS; δ l Indicates H s The subscript for each linguistic term in ; H s Semantics i, j represent the lowest and highest granularity levels of the hesitant fuzzy language term set, respectively;

[0026] A binary vector α based on the hesitant fuzzy language term set is introduced to convert the evaluation matrix of the hesitant fuzzy language term set given by the experts into a vector form:

[0027]

[0028] Furthermore, the semantic vector distance measurement in step S3 calculates the distance between different experts, and the specific description of clustering is:

[0029] Assume e M and e N is the evaluation matrix of two m×n dimensional hesitant fuzzy language term sets, which is converted into a binary vector, and the obtained projection value (directional factor) is:

[0030]

[0031] in, and e M and e N The binary vector of and They are and Model;

[0032]

[0033] Among them, M ab for e M The evaluation value of feasible solution a under attribute b, N ab is the evaluation value of eN for feasible solution a under attribute b; m and n represent the number of solutions and attributes in the decision, respectively; Indicates e M The hesitation of the evaluation value of feasible solution a under attribute b; Indicates e N The hesitation of the evaluation value of feasible solution a under attribute b; Indicates e M The lowest granularity level of the evaluation value of feasible solution a under attribute b; Indicates e N The lowest granularity level of the evaluation value of feasible solution a under attribute b; Indicates e M The granularity used to evaluate the feasible solution a under attribute b; Indicates e N The granularity used to evaluate the feasible solution a under attribute b;

[0034] When the number of elements in the hesitant fuzzy language term sets is different, in order to avoid the problems of large amount of calculation and information loss caused by the completion or transformation process, this application proposes a new distance definition (length factor) between expert evaluation matrices applicable to cross-granularity.

[0035] Definition H s (M ab )={h r(M ab )∈S} is about M ab The hesitant fuzzy language term set, H s (N ab )={h r (N ab )∈S} is about N ab The hesitant fuzzy language term set, where h r is the element form of the hesitant fuzzy language term set; then H s (M ab ) and H s (N ab ), the length factor, is defined as follows:

[0036]

[0037] Among them, #L(M ab ) and #L(N ab ) represent the number of elements in the hesitant fuzzy language term set used by experts M and N to evaluate the feasible solution a under attribute b, L ab is the value when the number of elements is equal;

[0038] Therefore, combining the projection value (direction factor) with the distance (length factor), the final distance formula is:

[0039] d(e M , e N )=η[1-cOs(e M , e N )]+(1-η)D(e M , e N )

[0040] Among them, 0≤η≤1 and η is the distance and direction parameter preference, which represents the decision maker's preference for the projection value or distance, which can be given in advance. In this application, η=0.4, which is more biased towards the result of distance. This formula can be applied to the calculation of hesitant fuzzy language term sets across granularities between different expert evaluation matrices. The distance formulas that appear in the following text are calculated according to this formula.

[0041] According to the newly defined distance formula above, the distance between different experts is calculated, and a 20×20 distance matrix D is obtained and K-medoids clustering is performed (the optimal number of clusters is determined by the sum of squared errors):

[0042] Furthermore, step S4 is specifically as follows:

[0043] The K-medoids clustering algorithm is used to divide p experts into K clusters, and then the similarity of the evaluation matrix of each expert in the cluster is measured to determine the expert weight. The higher the similarity, the more likely the experts in the cluster are to reach a consensus, and the higher the weight should be. p The similarity index can be expressed as:

[0044]

[0045] Among them, f k is cluster C k The number of experts in represents the expert e in cluster k p The total difference with other expert evaluation matrices, i.e.

[0046] Expert p The similarity-based weights are expressed as:

[0047]

[0048] Furthermore, step S5 is specifically as follows:

[0049] According to the expert evaluation matrix V(e p ) and expert weights Determine cluster C k The evaluation matrix G k ;

[0050]

[0051] Furthermore, the weight of each cluster in step S6 is calculated as follows:

[0052] If the expert clustering satisfies "high cohesion and low coupling", the clustering effect will be better. From the perspective of the relationship and behavior of a large group of experts, the more similar the evaluations of the experts in the cluster are, the better the clustering effect will be, and a larger weight can be given, and vice versa. For the experts in the cluster, the preference distance between experts is calculated using the newly defined distance formula to obtain the cluster expert similarity matrix ESM K :

[0053]

[0054] The internal similarity coefficient of clustering ISF (C k ) is expressed as:

[0055]

[0056] in, Represents the number of elements in the expert similarity matrix. The higher the similarity of expert preferences in a cluster, the greater the internal similarity coefficient of the cluster, and the higher the weight should be, and vice versa.

[0057] Calculate the distance between clusters and obtain the similarity matrix SM between clusters:

[0058]

[0059] G X and G Y Represent the evaluation matrices of cluster X and cluster Y respectively;

[0060] The external similarity coefficient ESF(C k ):

[0061]

[0062] The larger the external similarity coefficient of a cluster is, the higher the similarity between the cluster and other clusters is, indicating that the overlap between clusters is relatively high, and the cluster has more similarities with other clusters. The weight should be lower, and vice versa.

[0063] Based on the above, we can know that if the internal similarity coefficient is larger and the external similarity coefficient is lower, the clustering effect is better, and the weight between clusters should be higher, otherwise it is lower. Therefore, the inter-class weight ω(C k ):

[0064]

[0065] Furthermore, the large group decision matrix R in step S7 is:

[0066]

[0067] Furthermore, step S8 is specifically as follows:

[0068] The positive ideal solution A+ and the negative ideal solution A are defined as:

[0069] A + = {R 1+ , R 2+ , ..., R n+} and A - = {R 1- , R 2- , ..., R n-}.

[0070] in,

[0071]

[0072] The distances of each solution of the aggregation matrix to the positive and negative ideal solutions respectively:

[0073]

[0074] ω b Represents the weight vector of the attribute, ω b ≥0 and

[0075] Calculate the comprehensive evaluation value ψ a (a=1,2,...,m), sort them to get the final result:

[0076]

[0077] Compared with the prior art, the beneficial effects of the present invention include:

[0078] (1) A concept of binary vector based on hesitant fuzzy language term set is proposed to introduce the consideration of projection value, which makes the distance between experts more comprehensive. It effectively eliminates the influence of the situation that the expert evaluation matrix has large distance difference but the same preference ranking, improves the rationality and accuracy of multi-attribute group decision-making, and provides new ideas for solving related problems. (2) It is proposed to calculate the distance between different experts based on the semantic vector distance measure. The semantic vector distance measure can calculate the distance between experts of hesitant fuzzy language term sets of different granularity, and there is no need to expand the number of elements of a smaller hesitant fuzzy language term set. It considers the difference in language term values ​​more comprehensively and objectively, making it universally applicable; (3) The influence of the uncertainty difference (hesitation) of individual expert evaluation information on the calculation of the distance between experts is considered, making the distance more objective and reasonable; (4) There is no need to fill in the hesitant fuzzy language term sets with different numbers of elements, which ensures the integrity of the data and reduces the amount of calculation; (5) When there are experts with different strictness, some experts will give lower scores, but their evaluation ranking and hesitation of these attributes are the same as the evaluation ranking and hesitation of experts who give higher scores. In this case, the distance between each expert in the decision-making will change due to the consideration of the projection value, and the final result will be more reasonable and accurate.

[0079] In summary, the present invention utilizes the distance measurement theory of group decision-making and hesitant fuzzy language term sets, aiming to provide scientific decision support for the selection of bridge design schemes, and fully considers the definition and role of hesitant fuzzy language term sets. By proposing a distance measurement of semantic vectors, the actual situation can be better reflected and accurate decision results can be obtained. At the same time, the method takes into account multiple key attributes of alternative solutions, and conducts a comprehensive evaluation in combination with positive and negative ideal solutions, which has certain universality and practicality. The present invention has the advantages of comprehensively considering multiple factors, can improve the accuracy and reliability of decision-making, is more in line with actual decision-making needs, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 The figure is an overall flow chart of the method of the present invention. DETAILED DESCRIPTION

[0081] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0082] The present invention provides a multi-attribute group decision-making method for selecting bridge design based on cross-granularity semantic vector distance, aiming to help decision makers make the most sensible choice of bridge design scheme. First, the distance between the hesitant fuzzy language term sets of different experts is calculated according to the proposed semantic vector distance measure; then, clustering is performed to obtain cluster-related information, and then the intra-class weight and inter-class weight are obtained by calculation; finally, the schemes are sorted using the approximate ideal solution sorting method to obtain the optimal scheme. It provides new ideas for improving the group decision sorting method and provides an innovative method to help decision makers make the most sensible choice of bridge design scheme. The present invention mainly proposes a semantic vector distance formula for experts who may use hesitant fuzzy language term sets of different granularities and different numbers of elements to solve the distance between experts. This formula can calculate the distance between experts of hesitant fuzzy language term sets of different granularities, and there is no need to expand the number of elements of a smaller hesitant fuzzy language term set. It considers the differences in language term values ​​more comprehensively and objectively, and on this basis proposes a concept of binary vector based on hesitant fuzzy language term sets to introduce the consideration of the projection value factor, so that the distance between experts is calculated more comprehensively, and effectively eliminates the influence of the situation where the expert evaluation matrix has a large distance difference but the same preference ranking, making the ranking result more objective and reasonable, and more in line with actual decision-making needs.

[0083] A semantic vector distance formula is proposed to solve the distance between experts who may use hesitant fuzzy language term sets of different granularities and different numbers of elements. The proposed decision-making method of cross-granularity semantic vector distance measurement is used in the context of large group decision-making examples of two different bridge design schemes. Three simulation experiments are conducted to verify the rationality of semantic vector distance measurement in three different situations: no experts with large distance difference but the same preference ranking and using hesitant fuzzy language term sets of the same granularity, experts with large distance difference but the same preference ranking and using hesitant fuzzy language term sets of the same granularity, and experts using hesitant fuzzy language term sets of different granularities.

[0084] In order to prove the rationality and effectiveness of the method in this paper, the semantic vector distance measure proposed in this paper is used for large-scale group decision-making problems in Example 1 and Example 2 for verification.

[0085] Embodiment 1:

[0086] In March 2024, the construction of the Zhangjinggao Yangtze River Bridge, the world's largest span suspension bridge under construction, made important progress. The first section of the upstream and downstream main tower columns of the South Main Tower of the South Channel Bridge was accurately installed, marking the entry of the South Main Tower into a new stage of "growing taller". The Zhangjinggao Yangtze River Bridge connects the three cities of Suzhou, Taizhou and Nantong in Jiangsu Province. It starts from the west side of Shizhuang Town, Rugao, Nantong City, at the intersection of the Shanghai-Shaanxi Expressway, crosses Jiangxinzhou (Jingjiang Minzhusha Island, Taizhou City), and ends at the Chenyang Interchange of the Zhangjiagang Port Expressway in Suzhou City. The total length of the line is 30 kilometers, and the cross-river section is 7,859 meters long. At present, the challenge of choosing a suitable suspension bridge design is faced in bridge construction. In order to make a more informed choice of bridge design, a committee of 20 experts was invited to make decisions on three alternatives: x1 designed as a traditional suspension bridge, x2 designed as a spatial suspension bridge, and x3 designed as a hybrid structure bridge, based on three attributes: C1 construction period, C2 engineering cost, and C3 technical difficulty, where W = (0.13, 0.31, 0.56) T A vector representing the weights of the attributes.

[0087] The following is a detailed introduction to the decision-making process:

[0088] Step 1: Determine the evaluation attributes C1, C2, C3 and the solutions x1, x2, x3. 20 experts give the evaluation matrix in the form of the corresponding hesitant fuzzy language term set:

[0089]

[0090] Step 2: Convert the hesitant fuzzy language term set given by the expert into a binary vector form based on the hesitant fuzzy language term set:

[0091]

[0092] Step 3: Calculate the distance between different experts according to the newly defined distance formula, obtain a 20×20 distance matrix D and perform K-medoids clustering (the optimal number of clusters is determined by the sum of squared errors, K=5):

[0093]

[0094] Step 4: Calculate the weight of each expert in the cluster using the clustering results obtained in step 3;

[0095] Step 5: Get the group evaluation matrix of each cluster based on the expert weights and expert evaluation matrix:

[0096]

[0097]

[0098] Step 6: Get the specific clustering structure and weight information as shown in Table 1:

[0099] Table 1 Expert clustering and weight information of the present invention

[0100]

[0101] Step 7: Get the final group decision matrix:

[0102]

[0103] Step 8: Calculate the distance between each solution and the positive and negative ideal solutions, and then obtain the comprehensive evaluation value:

[0104] ψ1=0.6068, ψ2=0.5266, ψ3=0.4855

[0105] Table 2 Expert distance and weight information using different distance measurement methods

[0106]

[0107] The final decision matrices obtained using the distance measure of this paper, Yang Tingpu, Tan and Wu Peng are denoted as R1, R2, R3 and R4 respectively:

[0108]

[0109] Table 3 Comparison of comprehensive evaluation values ​​and ranking results of different literature methods

[0110]

[0111] In the case of the evaluation matrix given by 20 experts, it can be seen from Table 2 that compared with Yang Tingpu's method and Wu Peng's method that do not consider the hesitation and projection value, and Tan's method that considers the hesitation but does not consider the projection value, the C obtained by these four methods is 1 The maximum deviation of the weight is 0.0091, C 2 The maximum deviation of the weight is 0.0083, C 3 The maximum deviation of the weight is 0.0054, C 4 The maximum deviation of the weight is 0.008, C 5 The maximum deviation of the weight is 0.02, and the difference between the weights of each class is not very large. It can be seen that the clustering results of this paper are exactly the same as Tan's method, and Yang Tingpu's method is exactly the same as Wu Peng's method. The difference between the two clustering results is that e 20 Classified as C1 and C 5 Two different classes will cause C 1 and C 5 The two weight deviations are larger than those of other classes. This is because, unlike Yang Tingpu's method and Wu Peng's method, the distance measurement of this paper and Tan's method both consider the factor of hesitation. As can be seen from Table 3, the final ranking of the schemes is exactly the same, and both choose to set it as a traditional suspension bridge, so the distance measurement of this paper has certain feasibility and rationality.

[0112] Five experts with large distance differences but the same attribute preference rankings are added to the committee of 20 experts in Example 1 above. Their complete initial evaluation matrix is ​​as follows:

[0113]

[0114] The steps are the same as above and will not be described in detail here. The following data can be obtained.

[0115] Table 4 Expert distance and weight information using different distance measurement methods

[0116]

[0117] Table 5 Comparison of comprehensive evaluation values ​​and ranking results of different literature methods

[0118]

[0119] When 25 experts give the evaluation matrix, it can be seen from Table 4 that the C obtained by these four methods 1 The maximum deviation of the weight is 0.0337, C 2 The maximum deviation of the weight is 0.0299, C 3 The maximum deviation of the weight is 0.0063, C 4 The maximum deviation of the weight is 0.0098, C 5 The maximum deviation of the weights is 0.0044. 3 , C 4 and C 5 The clustering results of C are basically the same, but 1 and C 2 The clustering results of cluster C are quite different. 1 The clustering results of the methods used by Yang Tingpu, Tan and this paper are that experts e2, e7, and e 10 and experts 18 Classify into one category, cluster C 3The distance measurement used in this paper is the same as the clustering results of Yang Tingpu and Wu Peng, and both include expert e8. It can be seen that expert e8 should be classified into C 3 Cluster C 2 The distance measure used in this paper and the method of Yang Tingpu and Tan are both to use the expert e 19 With expert e4, expert e9 and expert e 13 Classified into one category, see expert e 1g It should be classified as C 2 According to experts 19 With experts 24 It can be clearly seen from the evaluation matrix given that expert e 19 With experts 24 The evaluation matrices of are only one level different in granularity and have exactly the same hesitation and ranking, so they should be classified into one cluster. However, only the distance measure in this paper classifies them into one category. Therefore, this paper effectively solves the misjudgment problem that occurs when clustering based on the distance between experts without considering the projection value. The clustering result using the distance measure in this paper is more reasonable.

[0120] As shown in Table 5, the optimal solution obtained by the distance measure in this paper is also different from the optimal solution obtained by other distance measures in the literature. In this context, this paper believes that the design of the spatial suspension bridge is the optimal solution. The distance measure of Yang Tingpu and Wu Peng does not consider the hesitation and projection value, and Tan's method considers the hesitation but not the projection value. This paper effectively solves the misjudgment situation that occurs when the projection value is not considered. The factors considered when calculating the distance between experts are more comprehensive, and the results are more reasonable and accurate.

[0121] Embodiment 2:

[0122] In urban traffic planning, a new bridge may need to be designed to improve traffic flow and connectivity. For group decision making, the following different options and their attributes can be considered: There are 3 bridge types to choose from, denoted as x1 arch bridge, x2 cable-stayed bridge, and x3 truss bridge. The following indicators are mainly considered: c1 cost-effectiveness; c2 traffic capacity; c3 disaster resistance; c4 city image. Each attribute has the same weight, and 20 experienced experts with different granularity are invited to participate in the decision-making.

[0123] Table 6 Comparison of comprehensive evaluation values ​​and ranking results of different literature methods

[0124]

[0125] Due to the different habits of experts, the granularity of the hesitant fuzzy language term sets used by different experts may be different. If the distance measure is only applicable to the hesitant fuzzy language term sets of the same granularity, the distance between experts is very limited. Therefore, the distance measure in this paper can be calculated across granularities. The results of the comparison with Xu Xuanhua's method can be seen in Table 6. The distance measure in this paper is consistent with the ranking results of Xu Xuanhua's method. Both of them choose the arch bridge as the optimal solution, which shows that the distance measure in this paper can also be applied to calculate the distance between experts of different granularities.

[0126] Therefore, the method in this paper inherits the excellent characteristics of the previous hesitant fuzzy language term set distance measurement, introduces the consideration of projection value (direction), and proposes a semantic vector distance measurement, which provides a new idea for group decision-making. Experiments and data fully verify the effectiveness of the weight assignment of the method in this paper and the rationality of the decision-making process.

[0127] Obviously, the above examples are only examples for clear explanation, and are not intended to limit the implementation methods. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from them are still within the protection scope of the invention.

Claims

1. A multi-attribute group decision-making method for selecting bridge designs, Its characteristic steps include: S1: Determine the evaluation attributes and schemes, and the experts give the evaluation matrix in the form of the corresponding hesitant fuzzy language term set; S2: Convert the hesitant fuzzy language term set given by the expert into a binary vector form based on the hesitant fuzzy language term set; S3: Calculate the distance between different experts based on the semantic vector distance measurement and perform clustering; S4: Calculate the weight of each expert in the class; S5: Determine the clustering evaluation matrix according to the expert evaluation matrix and expert weights; S6: Calculate the weight of each cluster; S7: Use the inter-class weights to fuse the group decision matrix of each cluster to obtain the final large group decision matrix; S8: Use the method of approximating positive and negative ideal solutions to sort the final solutions.

2. A multi-attribute group decision-making method for selecting bridge design according to claim 1, It is characterized in that Step S1 is specifically as follows: The language term set is represented by S = {s i |i=1,2,...,g}, where i represents the granularity level; s i A linguistic term for level i granularity, with granularity g and order, i.e., if θ < τ, then s θ τ ;​ Let S = {s 1 ,s 2 , ..., s g }, then the hesitant fuzzy language term set H S is defined as the set of finite ordered consecutive terms in S, namely: H S ={s i ,s i+1, ...,s j |s k ∈S,k=i,i+1,...,j}; Among them, i and j represent the lowest and highest granularity levels of the hesitant fuzzy language term set, respectively.

3. A multi-attribute group decision-making method for selecting bridge design according to claim 2, It is characterized in that The binary vector form based on the hesitant fuzzy language term set in step S2 is: First, we define the hesitant fuzzy language term set H S The hesitation is: For a language term set S = {s 1 ,s 2 , ..., s g }, let the language term in S The semantics of Then the language terms in S evenly cover [0,1]; #H s Indicates the number of elements of HFLTS; δ l Indicates H s The subscript for each linguistic term in ; H s Semantics i, j represent the lowest and highest granularity levels of the hesitant fuzzy language term set, respectively; A binary vector α based on the hesitant fuzzy language term set is introduced to convert the evaluation matrix of the hesitant fuzzy language term set given by the experts into a vector form:

4. A multi-attribute group decision-making method for selecting bridge design according to claim 3, It is characterized in that The semantic vector distance measurement in step S3 calculates the distance between different experts and performs clustering, specifically: Assume e M and e N is the evaluation matrix of two m×n dimensional hesitant fuzzy language term sets, which is converted into a binary vector, and the obtained projection value, that is, the direction factor is: in, and e M and e N The binary vector of and They are and Model; Among them, M ab for e M The evaluation value of feasible solution a under attribute b, N ab for e N The evaluation value of feasible solution a under attribute b; m and n represent the number of solutions and attributes in the decision respectively; Indicates e M The hesitation of the evaluation value of feasible solution a under attribute b; Indicates e N The hesitation of the evaluation value of feasible solution a under attribute b; Indicates e M The lowest granularity level of the evaluation value of feasible solution a under attribute b; Indicates e N The lowest granularity level of the evaluation value of feasible solution a under attribute b; Indicates e M The granularity used to evaluate the feasible solution a under attribute b; Indicates e N The granularity used to evaluate the feasible solution a under attribute b; Definition H s (M ab )={h r (M ab )∈S} is about M ab The hesitant fuzzy language term set, H s (N ab )={h r (N ab )∈S} is about N ab The hesitant fuzzy language term set, where h r is the element form of the hesitant fuzzy language term set; then H s (M ab ) and H s (N ab ), the length factor, is defined as follows: Among them, #L(M ab ) and #L(N ab ) represent the number of elements in the hesitant fuzzy language term set used by experts M and N to evaluate the feasible solution a under attribute b, L ab is the value when the number of elements is equal; Combining the projection value, i.e. the direction factor, with the distance, i.e. the length factor, the final distance formula is: d(and M ,And N )=η[1-cos(e M ,And N )]+(1-η)D(e M ,And N ) Where η represents the distance and direction parameter preference, and 0≤η≤1; The distances between different experts are calculated according to the final distance formula, and a 20×20 distance matrix D is obtained and K-medoids clustering is performed, where the optimal number of clusters is determined by the sum of squared errors.

5. A multi-attribute group decision-making method for selecting bridge design according to claim 4, It is characterized in that Step S4 is specifically as follows: The K-medoids clustering algorithm is used to divide p experts into K clusters, and then the similarity of the evaluation matrix of each expert in the cluster is measured to determine the expert weight. The higher the similarity, the more likely the experts in the cluster are to reach a consensus. The higher the weight, the lower the weight. In cluster k, expert e p The similarity index is expressed as: Among them, f k is cluster C k The number of experts in represents the expert e in cluster k p The total difference with other expert evaluation matrices, i.e. Expert p The similarity-based weights are expressed as:

6. A multi-attribute group decision-making method for selecting bridge design according to claim 5, It is characterized in that Step S5 is specifically as follows: According to the expert evaluation matrix V(e p ) and expert weights Determine cluster C k The evaluation matrix G k :

7. A multi-attribute group decision-making method for selecting a bridge design according to claim 6, It is characterized in that The weight of each cluster in step S6 is calculated as follows: The distance formula is used to calculate the preference distance between experts and obtain the clustered expert similarity matrix ESM K : The internal similarity coefficient of clustering ISF (C k ) is expressed as: in, Represents the number of elements in the expert similarity matrix; the higher the similarity of expert preferences in a cluster, the greater the internal similarity coefficient of the cluster and the higher the weight, and vice versa; Calculate the distance between clusters and obtain the similarity matrix SM between clusters: G X and G Y Represent the evaluation matrices of cluster X and cluster Y respectively; The external similarity coefficient ESF(C k ): The larger the external similarity coefficient of a cluster is, the higher the similarity between the cluster and other clusters is, indicating that the overlap between clusters is relatively high, and the cluster has more similarities with other clusters, the lower the weight is, and vice versa; Then, if the internal similarity coefficient is larger and the external similarity coefficient is lower, the clustering effect is better and the weight between clusters should be higher, otherwise it is lower; the inter-class weight ω(C k ):

8. A multi-attribute group decision-making method for selecting bridge design according to claim 7, It is characterized in that The large group decision matrix R in step S7 is expressed as:

9. A multi-attribute group decision-making method for selecting a bridge design according to claim 8, It is characterized in that Step S8 is specifically as follows: Define positive ideal solution A + and negative ideal solution A - They are: A + = {R 1+ , R 2+ ,..., R n+} and A - = {R 1- , R 2- ,..., R n-}. in, The distances of each solution of the aggregation matrix to the positive and negative ideal solutions respectively: w b Represents the weight vector of the attribute, w b ≥0 and Calculate the comprehensive evaluation value ψ a (a=1,2,...,m), sort them to get the final result:

Citation Information

Cited By

  • Medical material selection method and system considering individual semantic continuous learning

    CN120636738A