A dual-agent driven multi-attribute human-machine collaborative decision-making method and system

Abstract

The present invention relates to a dual-subject driven multi-attribute human-machine collaborative decision-making method and system, the method comprising: decomposing a human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set; obtaining a machine decision preference matrix and a human decision preference matrix; using human trust in the machine, calculating the human-machine hybrid decision preference matrix; obtaining machine decision attribute weights and human decision attribute weights; using the machine decision attribute weights and human decision attribute weights to calculate the distribution attribute weights, and calculating the mixed attribute entropy value according to the human-machine hybrid decision preference matrix; calculating the human-machine hybrid decision attribute weights according to the distribution attribute weights and the mixed attribute entropy value; assembling the human-machine hybrid decision preference matrix, and outputting the decision result. In the decision-making process, the preferences and advantages of the two decision-making subjects, humans and machines, are considered at the same time, the cooperative decision-making of humans and machines is realized, and the intelligent and scientific decision-making is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of human-machine collaboration, and in particular, relates to a dual-agent driven multi-attribute human-machine collaborative decision-making method and system. Background Art

[0002] Traditional decision-making methods rely solely on human evaluation and mainly rely on the achievement of multiple rounds of consensus among experts to ensure the effectiveness of decision results. However, as the correlation factors of decision problems become increasingly complex, the real evaluation space of decision problems in reality has exceeded the cognitive scope of human intelligence, and human decision-making ability is difficult to cover all dimensions. When there is a systematic deviation in expert preferences, it will lead to serious decision-making errors and bring significant losses. Therefore, it is necessary to use machine intelligence to make up for human deficiencies, form a collective wisdom of human-machine integration and symbiosis, and obtain more scientific and highly adaptable decision results.

[0003] Humans and machines have their own areas of expertise in decision-making, that is, machines and humans have different cognitive structures, processing speeds, information capacity, and learning methods, which lead to different contributions and influences on group decision-making. Due to the complexity of real decision-making problems, there are actually many problems that cannot be solved by humans or machines alone. If the interaction between human and machine is not tight, the decision attribute space of human-machine crossover will be ignored, and comprehensive decision information cannot be obtained. In addition, due to the operating principle of machines based on data and logic, they cannot directly give a comprehensive evaluation like humans, and need to be composed of multiple layers of single-dimensional indicators to form a comprehensive indicator. Therefore, the representation of the human-machine cross-decision space is different from the traditional single-layer attribute structure, which will produce a multi-layer, unbalanced evaluation index system, which is reflected as the dimension of decision attributes in multi-attribute decision-making, and will bring a series of problems that need to be solved in the construction of decision models. Summary of the invention

[0004] In order to remedy the deficiencies of the prior art, the present invention provides a dual-agent driven multi-attribute human-machine collaborative decision-making method and system.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] In a first aspect, a dual-agent driven multi-attribute human-machine collaborative decision-making method is provided, comprising:

[0007] Obtaining the decision target and the corresponding human-machine hybrid decision attribute set, and decomposing the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set;

[0008] According to the machine decision attribute set, a machine decision preference matrix is ​​obtained;

[0009] According to the human decision attribute set, a human decision preference matrix is ​​obtained;

[0010] Using the trust of humans in machines, the human-machine cross-decision preference matrix is ​​calculated and integrated with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix.

[0011] Obtaining a machine decision attribute weight according to a machine decision attribute set;

[0012] Obtaining human decision attribute weights according to a human decision attribute set;

[0013] The distribution attribute weights are calculated using the machine decision attribute weights and the human decision attribute weights, and the mixed attribute entropy value is calculated based on the human-machine mixed decision preference matrix;

[0014] The attribute weights of human-machine hybrid decision-making are calculated according to the distribution attribute weights and the entropy values ​​of the hybrid attributes, the human-machine hybrid decision-making preference matrix is ​​assembled, and the decision results are output.

[0015] Furthermore, the decision target and the corresponding human-machine hybrid decision attribute set are obtained, and the human-machine hybrid decision attribute set is decomposed into a machine decision attribute set and a human decision attribute set, including:

[0016] Get the decision target and the human-machine hybrid decision attribute set Q corresponding to the decision target, ,in represents the jth attribute;

[0017] Divide Q into machine decision attribute sets and human decision attribute set ,and , ,in express The jth attribute in ,in express The jth attribute in .

[0018] Furthermore, a machine decision preference matrix is ​​obtained according to the machine decision attribute set, including:

[0019] Set the machine decision attribute Properties in Decompose into multiple single-dimensional indicators , the r represents the number of single-dimensional indicators decomposed from the attributes;

[0020] Data collection, missing value interpolation, indicator extraction and importance measurement are carried out according to the indicator dimension of each single-dimensional indicator;

[0021] According to the importance of each single-dimensional indicator, the evaluation value of the single-dimensional indicator is aggregated as the comprehensive attribute evaluation value to obtain the machine decision preference matrix .

[0022] Furthermore, a human decision preference matrix is ​​obtained according to the human decision attribute set, including:

[0023] According to the human decision attribute set , using cosine similarity to divide M human decision makers into U clusters;

[0024] Calculate the decision preference matrix of each cluster, and use the weighted average operator to aggregate the preference matrices of all clusters in the group to obtain the human decision preference matrix .

[0025] Furthermore, by using the trust of humans in machines, the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set is calculated and integrated with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix, including:

[0026] Gaining human trust in machines ;

[0027] The attribute set of human-machine cross-decision is , the machine independent decision attribute set is , the attribute set of human independent decision making is ;

[0028] According to the machine decision preference matrix Get the machine's attribute set in human-machine cross-decision making Independent decision preference matrix under ;

[0029] According to the human decision preference matrix Get the attribute set of human in human-machine cross-decision Independent decision preference matrix under ;

[0030] use , and , the human-machine cross decision attribute preference matrix is ​​obtained as:

[0031] ;

[0032] The , and Combine to get the human-machine hybrid decision preference matrix for:

[0033] .

[0034] in, Representation attributes Machine preferences, Representation attributes of human preferences.

[0035] Furthermore, the machine decision attribute weights are obtained according to the machine decision attribute set, including:

[0036] From the bottom-level single-dimensional index of the machine decision attribute set, the feature importance is aggregated layer by layer to obtain the machine decision attribute weight .

[0037] Furthermore, the human decision attribute weights are obtained according to the human decision attribute set, including:

[0038] Targeting Human Decision Preference Matrix , the entropy weight method is used to calculate the weight of human decision attributes .

[0039] Furthermore, the distribution attribute weights are calculated using the machine decision attribute weights and the human decision attribute weights, and the mixed attribute entropy value is calculated based on the human-machine hybrid decision preference matrix, including:

[0040] Distribution attribute weights The calculation formula is:

[0041] ;

[0042] in, Representation attributes The machine decision attribute weights, Representation attributes The weight of human decision attributes and the weight of distribution attributes Perform normalization processing to obtain the distribution weight vector S;

[0043] According to the human-machine hybrid decision preference matrix G, the entropy value of the hybrid attribute is calculated using the entropy weight method. .

[0044] Furthermore, the attribute weights of human-machine hybrid decision-making are calculated based on the distribution attribute weights and the hybrid attribute entropy values, and the human-machine hybrid decision-making preference matrix is ​​assembled to output the decision results, including:

[0045] The cosine similarity is introduced to characterize the spatial distance between the attribute weight W of human-machine hybrid decision and the distribution attribute vector S: ;

[0046] The cosine similarity is introduced to characterize the spatial distance between the attribute weight W of human-machine hybrid decision and the hybrid attribute entropy value H: ;

[0047] Build the optimization model:

[0048] ;

[0049] in, is the balance coefficient, The larger the value, the closer the attribute weight of human-machine hybrid decision is to the distribution attribute weight. The smaller it is, the closer the attribute weight of human-machine hybrid decision is to the hybrid attribute entropy value;

[0050] Solve the optimization model to obtain the optimal attribute weights of human-machine hybrid decision-making;

[0051] Based on the optimal attribute weights of human-machine hybrid decision-making, the human-machine hybrid decision-making preference matrix is ​​assembled and the decision results are output.

[0052] In the second aspect, a dual-agent driven multi-attribute human-machine collaborative decision-making system is provided, comprising:

[0053] An attribute set processing module is used to obtain the decision target and the corresponding human-machine hybrid decision attribute set, and decompose the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set;

[0054] A machine preference module, used to obtain a machine decision preference matrix according to a machine decision attribute set;

[0055] A human preference module, used to obtain a human decision preference matrix according to a human decision attribute set;

[0056] The human-machine hybrid preference module is used to calculate the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set by using the human's trust in the machine, and integrate it with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix;

[0057] A machine weight module, used to obtain a machine decision attribute weight according to a machine decision attribute set;

[0058] A human weight module, used to obtain a human decision attribute weight according to a human decision attribute set;

[0059] The weight and entropy value calculation module is used to calculate the distribution attribute weight using the machine decision attribute weight and the human decision attribute weight, and calculate the mixed attribute entropy value according to the human-machine mixed decision preference matrix;

[0060] The decision result calculation module is used to calculate the attribute weights of human-machine hybrid decision-making based on the distribution attribute weights and the mixed attribute entropy values, assemble the human-machine hybrid decision preference matrix, and output the decision results.

[0061] The beneficial effects achieved by the present invention are:

[0062] Obtain the decision target and the corresponding human-machine hybrid decision attribute set, decompose the human-machine hybrid decision attribute set into the machine decision attribute set and the human decision attribute set; obtain the machine decision preference matrix according to the machine decision attribute set; obtain the human decision preference matrix according to the human decision attribute set; use the human trust in the machine to calculate the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set, and integrate it with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix; obtain the machine decision attribute weight according to the machine decision attribute set; obtain the human decision attribute weight according to the human decision attribute set; calculate the distribution attribute weight using the machine decision attribute weight and the human decision attribute weight; calculate the mixed attribute entropy value according to the human-machine hybrid decision preference matrix; calculate the human-machine hybrid decision attribute weight according to the distribution attribute weight and the mixed attribute entropy value, gather all the human-machine hybrid decision preference matrices, and output the decision result. In the decision-making process, the preferences and advantages of both human and machine decision subjects are considered at the same time to achieve cooperative decision-making between humans and machines and promote intelligent and scientific decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flow chart of the dual-agent driven multi-attribute human-machine collaborative decision-making method of the present invention;

[0064] Figure 2 A schematic diagram of a human-machine hybrid decision attribute set;

[0065] Figure 3 This is a structural diagram of the dual-agent driven multi-attribute human-machine collaborative decision-making system of the present invention. DETAILED DESCRIPTION

[0066] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0067] like Figure 1 As shown, the embodiment of the present invention provides a dual-agent driven multi-attribute human-machine collaborative decision-making method, including:

[0068] 101, obtaining a decision target and a corresponding human-machine hybrid decision attribute set, and decomposing the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set;

[0069] Among them, the decision target is obtained and the candidate solution set is , and the human-machine hybrid decision attribute set Q corresponding to the decision target, ,in represents the jth attribute.

[0070] The composition of the human-machine hybrid decision attribute set is as follows: Figure 2 shown.

[0071] Divide Q into machine decision attribute sets and human decision attribute set ,and , ,in express The jth attribute in ,in express The jth attribute in .

[0072] 102, obtaining a machine decision preference matrix according to the machine decision attribute set;

[0073] Among them, the machine decision attribute set Properties in Decompose into multiple single-dimensional indicators , the r represents the number of single-dimensional indicators decomposed from the attributes;

[0074] Data collection, missing value interpolation, indicator extraction and importance measurement are performed according to the indicator dimension of each single-dimensional indicator. The missing value interpolation method adopts the expectation maximization (EM) algorithm. When extracting indicators, the target value is generally defined in advance and the machine decision is used for redundant feature screening and indicator importance calculation. The recursive feature elimination (RFE) algorithm is needed to eliminate redundant features and improve the efficiency of evaluation reasoning; the Random Forest (RF) algorithm is used to mine the importance of indicators;

[0075] According to the importance of each single-dimensional indicator, the evaluation value of the single-dimensional indicator is aggregated as the comprehensive attribute evaluation value to obtain the machine decision preference matrix ,Right now

[0076] .

[0077] 103, obtaining a human decision preference matrix according to the human decision attribute set;

[0078] Among them, according to the human decision attribute set , using cosine similarity to divide M human decision makers into U clusters;

[0079] For each cluster, the preference matrices of all human decision makers contained in it are summed to obtain the decision preference matrix of each cluster. The preference matrices of all clusters in the group are aggregated using the weighted average operator to obtain the human decision preference matrix ,Right now

[0080] .

[0081] 104. Using the trust of humans in machines, the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set is calculated, and integrated with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix;

[0082] The specific process of human-machine decision preference fusion is as follows:

[0083] Gaining human trust in machines ,Human trust in machines can represent humans’ assessment of the reliability and safety of machines, and this trust is crucial for machines to be accepted and used in the decision-making process;

[0084] The attribute set of human-machine cross-decision is , the machine independent decision attribute set is , the attribute set of human independent decision making is ;

[0085] According to the machine decision preference matrix Get the machine's attribute set in human-machine cross-decision making Independent decision preference matrix under ;

[0086] According to the human decision preference matrix Get the attribute set of human in human-machine cross-decision Independent decision preference matrix under ;

[0087] use , and , the human-machine cross decision attribute preference matrix is ​​obtained as:

[0088] ;

[0089] The , and Combine to get the human-machine hybrid decision preference matrix for:

[0090] .

[0091] in, Representation attributes Machine preferences, Representation attributes of human preferences.

[0092] 105, obtaining a machine decision attribute weight according to the machine decision attribute set;

[0093] Among them, according to the multi-level structural characteristics of the machine decision attribute set, from the bottom-level single-dimensional index upward, the machine decision attribute weight is expressed as:

[0094] .

[0095] 106, obtaining a human decision attribute weight according to the human decision attribute set;

[0096] Among them, for the human decision preference matrix , the entropy weight method is used to calculate the weight of human decision attributes .

[0097] The entropy weight method is an objective weighting method based on the principle of information entropy, which is usually used in multi-index comprehensive evaluation and decision analysis. It determines the degree of dispersion of the index by calculating the entropy value of each index. The smaller the information entropy value, the greater the degree of variation of the index, and the greater the impact on the comprehensive evaluation. Therefore, its weight should be higher. Alternatives and The group preference matrix of evaluation attributes , its entropy is defined as:

[0098] ;

[0099] in, ,when hour, .

[0100] property The larger the entropy, the closer the difference between the values ​​of each scheme on this attribute and the average value of this attribute is. The entropy weight can be defined as:

[0101] ;

[0102] Thus, the weight of human decision attributes based on the human preference matrix in the human-machine hybrid decision problem can be obtained. .

[0103] 107, using the machine decision attribute weight and the human decision attribute weight to calculate the distribution attribute weight, and calculating the mixed attribute entropy value according to the human-machine mixed decision preference matrix;

[0104] Among them, the process of fusion of attribute weights for human-machine hybrid decision making is as follows:

[0105] Distribution attribute weights The calculation formula is:

[0106] ;

[0107] in, Representation attributes The machine decision attribute weights, Representation attributes The human decision attribute weights are normalized to obtain the distribution weight vector of the human-machine hybrid attribute set ;

[0108] According to the human-machine hybrid decision preference matrix G, the entropy value of the hybrid attribute is calculated using the entropy weight method . Human-machine hybrid attributes The entropy calculation process is as follows:

[0109] ;

[0110] ;

[0111] in, ,when hour, .

[0112] 108. The attribute weights of the human-machine hybrid decision are calculated according to the distribution attribute weights and the hybrid attribute entropy values, the human-machine hybrid decision preference matrix is ​​assembled, and the decision result is output.

[0113] The cosine similarity is introduced to characterize the spatial distance between the attribute weight W of human-machine hybrid decision and the distribution attribute vector S: ;

[0114] The cosine similarity is introduced to characterize the spatial distance between the attribute weight W of human-machine hybrid decision and the hybrid attribute entropy value H: ;

[0115] Build the optimization model:

[0116] ;

[0117] in, is the balance coefficient, The larger the value, the closer the attribute weight of human-machine hybrid decision is to the distribution attribute weight. The smaller it is, the closer the attribute weight of human-machine hybrid decision is to the hybrid attribute entropy value;

[0118] Solve the optimization model to obtain the optimal attribute weights of human-machine hybrid decision-making;

[0119] Based on the optimal attribute weights of human-machine hybrid decision-making, the human-machine hybrid decision-making preference matrix is ​​assembled and the decision results are output.

[0120] The principles of the embodiments of the present invention are as follows:

[0121] Obtain the decision target and the corresponding human-machine hybrid decision attribute set, decompose the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set; obtain the machine decision preference matrix based on the machine decision attribute set; obtain the human decision preference matrix based on the human decision attribute set; use the human's trust in the machine to calculate the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set, and integrate it with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix; obtain the machine decision attribute weights based on the machine decision attribute set; obtain the human decision attribute weights based on the human decision attribute set; calculate the distribution attribute weights using the machine decision attribute weights and the human decision attribute weights, and calculate the mixed attribute entropy value based on the human-machine hybrid decision preference matrix; calculate the human-machine hybrid decision attribute weights based on the distribution attribute weights and the mixed attribute entropy value, gather all the human-machine hybrid decision preference matrices, and output the decision results.

[0122] The beneficial effect that can be achieved is to establish a set of human-machine hybrid decision-making attributes, use machine learning technology to extract machine preferences from multi-dimensional attributes, adopt traditional group decision-making methods to aggregate human preferences, and realize the convergence of human-machine hybrid decision-making attributes, and finally realize the full play of human-machine hybrid intelligence in multi-attribute group decision-making, and promote intelligent and scientific decision-making.

[0123] In combination with the dual-agent driven multi-attribute human-machine collaborative decision-making method described in the above embodiments, a dual-agent driven multi-attribute human-machine collaborative decision-making system is described below through an embodiment.

[0124] like Figure 3 As shown, the embodiment of the present invention provides a dual-agent driven multi-attribute human-machine collaborative decision-making system, including:

[0125] The attribute set processing module 301 is used to obtain the decision target and the corresponding human-machine hybrid decision attribute set, and decompose the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set;

[0126] A machine preference module 302, used to obtain a machine decision preference matrix according to a machine decision attribute set;

[0127] A human preference module 303, used to obtain a human decision preference matrix according to a human decision attribute set;

[0128] The human-machine hybrid preference module 304 is used to calculate the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set by using the human's trust in the machine, and integrate it with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix;

[0129] A machine weight module 305, used to obtain a machine decision attribute weight according to a machine decision attribute set;

[0130] A human weight module 306, for obtaining a human decision attribute weight according to a human decision attribute set;

[0131] The weight and entropy value calculation module 307 is used to calculate the distribution attribute weight using the machine decision attribute weight and the human decision attribute weight, and calculate the mixed attribute entropy value according to the human-machine mixed decision preference matrix;

[0132] The decision result calculation module 308 is used to calculate the human-machine hybrid decision attribute weight according to the distribution attribute weight and the hybrid attribute entropy value, gather all the human-machine hybrid decision preference matrices, and output the decision result.

[0133] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0137] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

Claims

1. A dual-agent driven multi-attribute human-machine collaborative decision-making method, characterized in that: include: Obtaining a decision target and a corresponding human-machine hybrid decision attribute set, and decomposing the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set; Obtaining a machine decision preference matrix according to the machine decision attribute set; Obtaining a human decision preference matrix according to the human decision attribute set; Using the trust of humans in machines, the human-machine cross-decision attribute preference matrix under the human-machine cross-decision attribute set is calculated, and then integrated with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix; Obtaining a machine decision attribute weight according to the machine decision attribute set; Obtaining a human decision attribute weight according to the human decision attribute set; Calculating the distribution attribute weight using the machine decision attribute weight and the human decision attribute weight, and calculating the hybrid attribute entropy value according to the human-machine hybrid decision preference matrix; Calculate the attribute weight of human-machine hybrid decision-making according to the distribution attribute weight and the hybrid attribute entropy value, gather the human-machine hybrid decision-making preference matrix, and output the decision result; The step of obtaining a decision target and a corresponding human-machine hybrid decision attribute set, and decomposing the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set, includes: Obtain the decision target and the human-machine hybrid decision attribute set corresponding to the decision target Q , It is stated Indicates j Attributes; Divide Q into machine decision attribute sets and human decision attribute set , and the , , where Indicates the Middle j properties, , where Indicates the Middle j Attributes; The obtaining of a machine decision preference matrix according to the machine decision attribute set includes: The machine decision attribute set Properties in Decompose into multiple single-dimensional indicators , r Indicates the number of single-dimensional indicators decomposed from the attributes; Data collection, missing value interpolation, indicator extraction and importance measurement are carried out according to the indicator dimension of each single-dimensional indicator; According to the importance of each single-dimensional indicator, the evaluation value of the single-dimensional indicator is aggregated as the comprehensive attribute evaluation value to obtain the machine decision preference matrix ; The step of obtaining a human decision preference matrix according to the human decision attribute set includes: According to the human decision attribute set, the M human decision makers are divided into U clusters using cosine similarity; Calculate the decision preference matrix of each cluster, and use the weighted average operator to aggregate the preference matrices of all clusters in the group to obtain the human decision preference matrix; The human-machine cross-decision attribute preference matrix under the human-machine cross-decision attribute set is calculated by using the human's trust in the machine, and is integrated with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix, including: According to the human decision attribute set , using cosine similarity to divide M human decision makers into U clusters; Calculate the decision preference matrix of each cluster, and use the weighted average operator to aggregate the preference matrices of all clusters in the group to obtain the human decision preference matrix ; The human-machine cross-decision attribute preference matrix under the human-machine cross-decision attribute set is calculated by using the human's trust in the machine, and is integrated with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix, including: Gaining trust in machines ; The attribute set of human-machine cross-decision is , the machine independent decision attribute set is , the attribute set of human independent decision making is ; According to the machine decision preference matrix Get the machine in the human-machine cross decision attribute set The machine independent decision preference matrix under ; According to the human decision preference matrix Get the human-machine cross-decision attribute set The human independent decision preference matrix under ; Using the , and , the human-machine cross decision attribute preference matrix is ​​obtained as: ； The , and Combine to get the human-machine hybrid decision preference matrix for: ； Among them, the Indicates the attribute The machine preference, Indicates the attribute of human preferences.

2. The dual-agent driven multi-attribute human-machine collaborative decision-making method according to claim 1 is characterized in that: The obtaining the machine decision attribute weight according to the machine decision attribute set includes: From the bottom-level single-dimensional index of the machine decision attribute set, the feature importance is aggregated layer by layer to obtain the machine decision attribute weight .

3. The dual-agent driven multi-attribute human-machine collaborative decision-making method according to claim 2 is characterized in that: The obtaining of the human decision attribute weight according to the human decision attribute set includes: For the human decision preference matrix , the entropy weight method is used to calculate the weight of human decision attributes .

4. The dual-agent driven multi-attribute human-machine collaborative decision-making method according to claim 3 is characterized in that: The use of the machine decision attribute weight and the human decision attribute weights Calculating the distribution attribute weights and calculating the mixed attribute entropy value according to the human-machine hybrid decision preference matrix includes: Distribution attribute weights The calculation formula is: ； Among them, the Indicates the attribute The machine decision attribute weights, Indicates the attribute The human decision attribute weights are normalized to obtain the distribution weight vector of the human-machine hybrid attribute set ; According to the human-machine hybrid decision preference matrix G Calculate the entropy value of mixed attributes using the entropy weight method .

5. The dual-agent driven multi-attribute human-machine collaborative decision-making method according to claim 4 is characterized in that: The method of calculating the attribute weight of the human-machine hybrid decision according to the distribution attribute weight and the hybrid attribute entropy value, aggregating the human-machine hybrid decision preference matrix, and outputting the decision result includes: Introducing cosine similarity to characterize attribute weights of human-machine hybrid decision making W With the distribution attribute vector S The spatial distance between ； Introduce cosine similarity to characterize the attribute weights of the human-machine hybrid decision W With the mixed attribute entropy value H The spatial distance between ; Build the optimization model: ； Among them, the is the balance coefficient, The larger the value, the closer the attribute weight of the human-machine hybrid decision is to the distribution attribute weight. The smaller it is, the closer the attribute weight of the human-machine hybrid decision is to the hybrid attribute entropy value; Solving the optimization model to obtain the optimal human-machine hybrid decision attribute weights; Based on the optimal human-machine hybrid decision attribute weights, all human-machine hybrid decision preference matrices are assembled to output the decision results.

6. A dual-agent driven multi-attribute human-machine collaborative decision-making system, characterized in that: include: An attribute set processing module, used to obtain a decision target and a corresponding human-machine hybrid decision attribute set, and decompose the human-machine hybrid decision attribute set into a machine decision attribute set and a human decision attribute set; A machine preference module, used to obtain a machine decision preference matrix according to the machine decision attribute set; A human preference module, used to obtain a human decision preference matrix according to the human decision attribute set; The human-machine hybrid preference module is used to calculate the human-machine cross-decision preference matrix under the human-machine cross-decision attribute set by using the human's trust in the machine, and integrate it with the machine independent decision preference matrix and the human independent decision preference matrix to obtain the human-machine hybrid decision preference matrix; A machine weight module, used to obtain a machine decision attribute weight according to the machine decision attribute set; A human weight module, used to obtain a human decision attribute weight according to the human decision attribute set; A weight and entropy value calculation module, used to calculate the distribution attribute weight using the machine decision attribute weight and the human decision attribute weight, and calculate the mixed attribute entropy value according to the human-machine mixed decision preference matrix; A decision result calculation module, used to calculate the attribute weight of human-machine hybrid decision according to the distribution attribute weight and the hybrid attribute entropy value, gather the human-machine hybrid decision preference matrix, and output the decision result; The attribute set processing module is specifically used to obtain the decision target and the human-machine hybrid decision attribute set corresponding to the decision target. Q , , where Indicates j attributes; divide the Q into machine decision attribute sets and human decision attribute set , and the , , where Indicates the Middle j properties, , where Indicates the Middle j Attributes; The machine preference module is specifically used to set the machine decision attribute set Properties in Decompose into multiple single-dimensional indicators , r Indicates the number of single-dimensional indicators decomposed from the attributes; performs data collection, missing value interpolation, indicator extraction and importance measurement processing according to the indicator dimension of each single-dimensional indicator; according to the importance of each single-dimensional indicator, aggregates the single-dimensional indicator evaluation value as the comprehensive attribute evaluation value to obtain the machine decision preference matrix ; The human preference module is specifically used to determine the human decision attribute set , use cosine similarity to divide M human decision makers into U clusters; calculate the decision preference matrix of each cluster, and use the weighted average operator to aggregate the preference matrices of all clusters in the group to obtain the human decision preference matrix ; The human-machine hybrid preference module is specifically used to obtain human trust in the machine ; The attribute set of human-machine cross-decision making is , the machine independent decision attribute set is , the attribute set of human independent decision making is ; According to the machine decision preference matrix Get the machine in the human-machine cross decision attribute set The machine independent decision preference matrix under ; According to the human decision preference matrix Get the human-machine cross-decision attribute set The human independent decision preference matrix under ; Using the , and , the human-machine cross decision attribute preference matrix is ​​obtained as: ; , and Combine to get the human-machine hybrid decision preference matrix for: ； Among them, the Indicates the attribute The machine preference, Indicates the attribute of human preferences.

Images