Cloud model-based index maximum entropy weight rapid configuration system, method, device and medium

Abstract

The application discloses a kind of based on cloud model's index maximum entropy weight fast configuration system, method, device and medium, belong to the cross field of uncertain artificial intelligence, intelligent optimization and efficiency evaluation, including: maximum entropy weight configuration module, relative comprehensive capacity quantification module, evaluation index weight optimization module;Maximum entropy weight configuration module is used to write first program using computer, and first program is run on processor, according to maximum entropy principle to build objective function, the prior weight information of evaluation index is converted into quantifiable digital constraint condition with the relative capacity ordering between multiple evaluation objects, form the weight configuration optimization model considering objectivity and subjective experience etc.The present application is realized based on computer device and program code, and each evaluation index weight can be calculated scientifically and quickly using subjective and objective factors.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of uncertainty artificial intelligence, intelligent optimization and performance evaluation, and more specifically, to a system, method, apparatus and medium for rapid configuration of index maximum entropy weights based on cloud models. Background Technology

[0002] Performance evaluation typically assesses the overall performance / capabilities of an evaluation object based on quantified indicator values ​​and corresponding weights using weighted sums. The accuracy and objectivity of the weights directly impact the evaluation results. Current weighting methods are mainly divided into subjective and objective methods. Subjective weighting methods, such as the Analytic Hierarchy Process (AHP), Attribute Hierarchy Model (AHM), Delphi method, and expert scoring, incorporate expert experience and judgment, but the weighting results are prone to individual variations and significant subjective bias. Objective weighting methods, such as weighting based on inter-standard conflict and correlation, grey relational analysis, similarity coefficient method, and entropy weight method, are susceptible to data quality issues, leading to the weakening of some important indicators or the exaggeration of secondary indicators. These problems can distort the evaluation results, necessitating a weighting method that integrates subjective and objective approaches, leveraging the advantages of each to achieve scientific and objective weighting.

[0003] The maximum entropy-based weight allocation method fully combines subjective expert knowledge and objective quantitative data to achieve maximum differentiation of each indicator under multiple constraints, effectively solving the long-standing problem of insufficiently scientific and objective weight settings for evaluation indicators. However, when the quantitative results of evaluation indicators are random and fuzzy, the optimization problem under the constraint of relative ability ranking of multiple evaluation objects becomes a non-convex optimization problem, and convex optimization methods cannot be directly used to find the optimal weight allocation scheme. If traditional optimization algorithms such as simulated annealing are used, a large number of iterations are required to find the optimal weight. Therefore, it is urgent to study a subjective and objective weighting scheme with a maximum entropy weight allocation architecture, multi-dimensional non-convex optimization constraints, and fast global optimization capability to configure objective and accurate weight values ​​for each evaluation indicator. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a system, method, device and medium for rapid configuration of index maximum entropy weights based on cloud models. It is implemented based on computer devices and program code, and ultimately achieves the ability to scientifically and quickly calculate the weights of each evaluation index using subjective and objective factors.

[0005] The objective of this invention is achieved through the following solution: A cloud-based system for rapid configuration of maximum entropy weights for indicators includes: a maximum entropy weight configuration module, a relative comprehensive capability quantification module, and an evaluation indicator weight optimization module. The maximum entropy weight configuration module is used to write a first program using a computer and run the first program on the processor. Based on the principle of maximum entropy, it constructs an objective function and transforms the prior weight information of the evaluation index and the relative ability ranking among multiple evaluation objects into quantifiable numerical constraints, forming a weight configuration optimization model that combines objectivity and subjective experience. The relative comprehensive capability quantification module is used to write a second program on a computer and run the second program on the processor. Based on the TOPSIS approximation ideal solution ranking method and the cloud model, it integrates the cosine distance metric to convert the mean, entropy and hyperentropy of each evaluation index of each evaluation object into the distance between each evaluation object and the positive and negative ideal solutions, so as to achieve accurate quantification and ranking of comprehensive capabilities. The results are fed back to the maximum entropy weight configuration module to dynamically correct the constraint expression of relative capability ranking and enhance the adaptability and consistency of the model. The evaluation index weight optimization module is used to write a third program on a computer and run the third program on the processor. It adopts the particle swarm optimization algorithm to simulate the behavior of swarm intelligence, introduces a linear decreasing inertial weight strategy, and efficiently iterates in the search space to gradually approach the global optimal weight combination, and finally outputs a scientific and reasonable weight configuration scheme.

[0006] A fast configuration method for maximum entropy weights of indicators based on a cloud model, which is based on the fast configuration system for maximum entropy weights of indicators based on a cloud model as described above, includes the following steps: Step 1: Based on the principle of maximum entropy, a priori cognition is fused and constraint modeling is performed. A weight modeling framework based on the principle of maximum entropy is constructed, which transforms the subjective preferences of experts for each evaluation index and the relative ranking relationship between evaluation objects into quantifiable mathematical constraints, thereby establishing a highly adaptive weight optimization model. Step 2: The relative comprehensive capability quantification module uses the cloud model to transform quantitative data with randomness and fuzziness into qualitative concepts. Under the TOPSIS method framework, it constructs a relative comprehensive capability quantification model for multiple assessment objects through positive and negative ideal solutions and a sub-functional module based on positive and negative distances. This model is then used as a constraint to update the quantification formula for the relative capability ranking constraint of multiple assessment objects in the maximum entropy weight configuration construction module. Step 3: Evaluation index weight global optimization and convergence module. Based on the adaptive particle swarm optimization mechanism, it performs global random optimization and finally obtains the optimal weight solution.

[0007] Furthermore, in step 1, the prior cognition fusion and constraint modeling based on the maximum entropy principle constructs a weight modeling framework based on the maximum entropy principle. This transforms the experts' subjective preferences for each evaluation indicator and the relative ranking relationship between evaluation objects into quantifiable mathematical constraints, thereby establishing a highly adaptive weight optimization model. Specifically, this includes the following sub-steps: (1) Construct the objective function: For a hierarchical set of capability assessment indicators, it is necessary to aggregate capabilities using the quantified values ​​and absolute weights of the underlying capability assessment indicators, with the absolute weights of the capability assessment indicators as input parameters. Assume the number of underlying capability indicators in the constructed assessment indicator set is... Then, the objective function for constructing the weight allocation model under the maximum entropy criterion is: ; in, Let be the objective function. For the first The absolute weight of each evaluation indicator for The absolute weight vector of the evaluation index of dimension, i.e. .

[0008] (2) Constructing constraints based on prior knowledge: ① When addressing evaluation issues, the indicator weights should be the normalized absolute weights of each indicator; therefore, their sum should be 1. ; ② The absolute weights of each indicator after normalization are not allowed to be less than zero and must all be less than 1, that is... ; ③ Based on industry experts' understanding of the evaluation objects, and acquiring accumulated generally accepted knowledge—that is, if the overall ability of one evaluation object is better than another—then the knowledge, after integration, can be converted into a mathematical expression as follows: ; in, For the first The relative ability value of each assessment subject. The number of objects in this recognized knowledge; ④ Based on the subjective preferences and prior knowledge of industry experts regarding each evaluation indicator, some indicators are defined as highly important indicators, meaning that certain indicators are far more important than others in the evaluation process. This prior knowledge is comprehensively quantified and expressed as a mathematical formula: ; in, This refers to the set of indicators that experts have identified as either highly important or low in importance. This is the indicator importance coefficient, with a value greater than 1. The final optimization problem, transformed from the weight allocation problem, is expressed mathematically as follows: .

[0009] Further, in step 2, the relative comprehensive capability quantification module uses a cloud model to transform quantitative data with randomness and fuzziness into qualitative concepts. Under the TOPSIS method framework, it constructs a relative comprehensive capability quantification model for multiple assessment objects through a sub-functional module that builds positive and negative ideal solutions and assesses relative comprehensive capability based on positive and negative distances. This model is then used as a constraint to update the quantification formula for the relative capability ranking constraint of multiple assessment objects in the maximum entropy weight configuration construction module. Specifically, this includes the following sub-steps: (1) Construction of positive and negative ideal solutions: Assume the first The first assessment subject The number of quantitative sample data for each underlying capability indicator is: Based on cloud model theory, a reverse cloud generator is used to transform multiple sets of sample data into qualitative concepts with digital characteristics. ,in , and The first The first of the assessment subjects The performance decision cloud matrix of the evaluated object is determined by the mean, entropy, and hyperentropy of each evaluation metric. , ; The mean describes the basic certainty of a qualitative concept; a higher value indicates a higher basic certainty of the quantified data. Entropy and hyperentropy describe the degree of uncertainty of a qualitative concept; lower values ​​indicate lower fuzziness and randomness in the quantified data. Based on the definitions of these three numerical characteristics and following the construction principles of the positive ideal solution in the TOPSIS algorithm, the constructed positive ideal solution is as follows: ; Among them, the positive ideal mean solution positive ideal entropy solution and positive ideal superentropy solution The representation is as follows: ; Following the principle that the construction method for negative ideal solutions is the opposite of that for positive ideal solutions, the expression for negative ideal solutions is constructed as follows: ; In the above formula, the negative ideal mean solution Negative ideal entropy solution and negative ideal hyperentropy solution The representation is as follows: ; (2) Relative comprehensive ability assessment based on positive and negative cosine distance: ① Positive distance

[0010] positive distance Used to characterize the Weighted evaluation object decision cloud matrix of each evaluation object Weighted positive ideal cloud The absolute difference, the weighted evaluation object decision cloud matrix Decision cloud matrix for evaluating object performance Based on this, absolute weights are introduced for each underlying capability indicator, specifically expressed as follows: ; Weighted Positive Ideal Cloud In the positive ideal solution Based on this, the absolute weights of each underlying capability indicator are introduced, expressed as follows: ; positive distance Based on this, we introduce cosine distance, which is expressed as: ; in, For vectors, 1-norm For the first Weighted positive difference matrix of each assessment subject The Middle Column vector, and the first Positive distance of each performance metric Correspondingly, the weighted positive difference matrix of the evaluation object The specific form is as follows: ; The similarity and distance difference between the numerical characteristics of each indicator of the object and the positive ideal solution are evaluated using cosine distance characteristics, as detailed below: ; ② Negative distance

[0011] Negative distance Decision cloud matrix representing weighted evaluation objects Weighted Negative Ideal Cloud The absolute difference, negative ideal cloud In negative ideal solution Based on this, absolute weights are introduced for each underlying capability indicator, specifically in the following form: ; Negative distance Based on this, we introduce cosine distance, which takes the following form: ; in, For the first Weighted negative difference matrix of each assessment subject The Middle Column vector, and the first Negative distance of each performance metric Correspondingly, the weighted negative difference matrix of the evaluation object The specific form is as follows: ; Among them, the cosine distance difference between the numerical characteristics of each indicator of the evaluation object and the negative ideal solution. Specifically, it takes the following form: ; ③ The relative comprehensive ability of the assessment object

[0012] According to the TOPSIS method for calculating relative ability, then the first The relative comprehensive ability of each assessment subject is: ; The above formula is used to calculate the relative comprehensive capabilities of multiple assessment objects, and the results are used to rank the capabilities, achieving a quantitative transformation of expert perception. This is then used as a constraint to update the quantitative formula for the relative capability ranking constraint of multiple assessment objects in the maximum entropy weight configuration construction module. The final weight configuration model under uncertain information is as follows: .

[0013] Furthermore, in step 3, the global optimization and convergence module for the evaluation index weights performs global random optimization based on an adaptive particle swarm optimization mechanism to finally obtain the optimal weight solution, specifically including the following sub-steps: (1) Initialize the particle swarm, including the swarm size. Number of iterations T The position of each particle and speed and individual extreme values and global extrema Among them, the group size Based on the number of indicators M Determined, for each particle Generate random vectors ,in The initial position of the particle is determined by performing normalization on the random vector. The specific formula is as follows: ; initial velocity ,in , Take a smaller random initial value and determine whether the initial particle swarm satisfies constraints ①, ②, ③, and ④. If it does, proceed with the subsequent process operations. If it does not, re-initialize the particle swarm until a particle swarm that meets the constraints is generated. (2) Calculate the fitness value of each particle. The specific formula is as follows: ; (3) For each particle, use its fitness value and individual extreme values In comparison, In the case of, use Replace ; (4) For each particle, use its fitness and global extrema In comparison, In the case of, use Replace ; (5) Perform T In the next iteration, the particle velocity is updated. and location The speed update formula is expressed as follows: ; in, Indicates the first i The particle in the first t The speed of each iteration Indicates the first i The particle in the first t The position of the next iteration. Indicates the first i The historical best position of each particle This represents the current position of the globally optimal particle. Two independent random numbers are used to control the randomness. , These represent two learning factors. This represents dynamically adjusted inertia weights, using a linearly decreasing inertia weight strategy, which decreases with the number of iterations. Down to The formula is expressed as follows: ; The position update formula is expressed as follows: Represents the 1-norm ; (6) After updating, check whether the position of the particle satisfies the constraint boundary conditions ①, ②, ③, and ④. If it does, continue to execute the subsequent steps (7); otherwise, correct the update process. (7) Determine whether the algorithm termination condition is met. If so, end the algorithm and output the optimal weight solution. Otherwise, return to step (2).

[0014] Furthermore, group size .

[0015] A cloud-based metric maximum entropy weight fast configuration device includes a processor and a memory, wherein the memory stores a computer program that, when loaded by the processor, executes the method described in any of the preceding methods.

[0016] A computer-readable storage medium storing a computer program that, when loaded by a processor, executes the method described in any of the preceding claims.

[0017] The beneficial effects of this invention include: This invention, based on cloud model theory and the TOPSIS framework, writes code and runs on a computer device to construct a quantitative model of the relative comprehensive capabilities of multiple evaluation objects by integrating cosine distance. It eliminates the need to normalize the mean, entropy, and hyperentropy of each evaluation index, and analyzes the positive and negative distances of each evaluation index from the perspective of directional trends. This avoids the problem of quantitative distortion of comprehensive capabilities caused by the inaccuracy of "normalization + Euclidean distance", and improves the scientific nature of transforming qualitative cognition and experience knowledge of experts in the field into calculable quantitative mathematical expressions.

[0018] The weight configuration model proposed in this invention, based on an improved particle swarm optimization algorithm, is coded and run on a computer device. It breaks through the application limitations of traditional methods in non-convex optimization problems, and compared with traditional typical optimization methods, it can search for the optimal weight configuration scheme faster, significantly improving the accuracy and practicality of the evaluation system.

[0019] The proposed maximum entropy weight fast configuration method, which is coded and run on a computer device, realizes a closed-loop integration from prior knowledge fusion and dynamic constraint update to intelligent optimization solution, significantly improving the reliability and practicality of multi-attribute decision-making process. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the steps of the method in an embodiment of the present invention. Detailed Implementation

[0022] All features disclosed in all embodiments of this specification, or steps in all methods or processes implied in the disclosure, may be combined and / or extended or replaced in any way, except for mutually exclusive features and / or steps.

[0023] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0024] As a first aspect of the present invention, a cloud model-based system for rapid configuration of index maximum entropy weights is provided, comprising: a maximum entropy weight configuration module, a relative comprehensive capability quantification module, and an evaluation index weight optimization module. Further, according to an embodiment of the present invention, the maximum entropy weight configuration module constructs an objective function based on the maximum entropy principle, transforming the prior weight information of the evaluation indexes and the relative capability ranking among multiple evaluation objects into quantifiable numerical constraints, forming a weight configuration optimization model that combines objectivity and subjective experience. The relative comprehensive capability quantification module, based on the Approximation to Ideal Solution Ranking (TOPSIS) method and the cloud model, integrates a cosine distance metric to convert the mean, entropy, and hyperentropy of each evaluation index of each evaluation object into the distance between each evaluation object and the positive and negative ideal solutions, achieving accurate quantification and ranking of comprehensive capabilities. This is fed back to the maximum entropy weight configuration module, dynamically correcting the constraint expressions regarding relative capability ranking, and enhancing the model's adaptability and consistency. The evaluation index weight optimization module employs a particle swarm optimization algorithm to simulate swarm intelligence behavior, introducing a linearly decreasing inertial weight strategy, efficiently iterating in the search space, gradually approaching the globally optimal weight combination, and finally outputting a scientifically reasonable weight configuration scheme.

[0025] As a second aspect of the present invention, based on the above system, a method for fast configuration of maximum entropy weights of indicators based on cloud models is provided, the specific implementation steps of which are as follows: Step 1: The prior cognition fusion and constraint modeling module based on the maximum entropy principle constructs a weight modeling framework based on the maximum entropy principle. It transforms the subjective preferences of experts for each evaluation indicator and the relative ranking relationship between evaluation objects into quantifiable mathematical constraints, thereby establishing a highly adaptive weight optimization model. The specific process is as follows.

[0026] (1) Construct the objective function For a hierarchical set of capability assessment indicators, it is necessary to aggregate capabilities using the quantified values ​​and absolute weights of the underlying capability assessment indicators. Therefore, the absolute weights of the capability assessment indicators are used as input parameters. It is assumed that the number of underlying capability indicators in the constructed assessment indicator set is... Then, the objective function for constructing the weight allocation model under the maximum entropy criterion is:

[0027] in, Let be the objective function. For the first The absolute weight of each evaluation indicator, for The absolute weight vector of the evaluation index of dimension, i.e. .

[0028] (2) Constructing constraints based on prior knowledge ① When addressing evaluation issues, the indicator weights should be the normalized absolute weights of each indicator; therefore, their sum should be 1.

[0029] ② The absolute weights of each indicator after normalization are not allowed to be less than zero and must all be less than 1, that is...

[0030] ③ Based on industry experts' understanding of this type of assessment object, some accumulated and generally accepted knowledge can be obtained. That is, among the assessment objects in this category, the comprehensive ability of one object is better than that of another object. This knowledge, after being synthesized, can be converted into a mathematical expression as follows:

[0031] in, For the first The relative ability value of each assessment subject. The number of objects in this recognized knowledge.

[0032] ④ Based on the subjective preferences and prior knowledge of industry experts regarding each evaluation indicator, some indicators can be defined as "highly important" indicators, meaning that certain indicators are far more important than others in the evaluation process. This prior knowledge, after comprehensive quantification, can be expressed as a mathematical formula:

[0033] in, This refers to the set of indicators that experts have identified as "highly important" and "lowly important." This is the indicator importance ratio coefficient, with a value greater than 1.

[0034] The final optimization problem, transformed from the weight allocation problem, is expressed mathematically as follows:

[0035] Step 2: The relative comprehensive capability quantification module uses a cloud model to transform quantitative data with randomness and fuzziness into qualitative concepts. Under the TOPSIS method framework, through sub-functional modules such as constructing positive and negative ideal solutions and evaluating relative comprehensive capabilities based on positive and negative distances, a relative comprehensive capability quantification model for multiple evaluation objects is constructed. This model is then used as a constraint to update the quantification formula for the relative capability ranking constraint of multiple evaluation objects in the maximum entropy weight configuration construction module. The specific process is shown below.

[0036] (1) Construction of positive and negative ideal solutions Assume the first The first assessment subject The number of quantitative sample data for each underlying capability indicator is: Based on cloud model theory, a reverse cloud generator is used to transform multiple sets of sample data into qualitative concepts with digital characteristics. ,in , and The first The first of the assessment subjects The performance decision cloud matrix of the evaluated object is determined by the mean, entropy, and hyperentropy of each evaluation metric. .

[0037]

[0038] The mean describes the basic certainty of a qualitative concept; a higher value indicates a higher basic certainty of the quantified data. Entropy and hyperentropy describe the degree of uncertainty of a qualitative concept; lower values ​​indicate lower fuzziness and randomness in the quantified data. Based on the definitions of these three numerical characteristics and following the construction principles of the positive ideal solution in the TOPSIS algorithm, the constructed positive ideal solution is as follows:

[0039] Among them, the positive ideal mean solution positive ideal entropy solution and positive ideal superentropy solution The representation is as follows:

[0040] Similarly, following the principle that the construction method of the negative ideal solution is the opposite of that of the positive ideal solution, the expression for the constructed negative ideal solution is as follows:

[0041] In the above formula, the negative ideal mean solution Negative ideal entropy solution and negative ideal hyperentropy solution The representation is as follows:

[0042] (2) Relative comprehensive ability assessment based on positive and negative cosine distance ① Positive distance

[0043] positive distance Used to characterize the Weighted evaluation object decision cloud matrix of each evaluation object Weighted positive ideal cloud The absolute difference, the weighted evaluation object decision cloud matrix Decision cloud matrix for evaluating object performance Based on this, absolute weights are introduced for each underlying capability indicator, which can be specifically expressed as:

[0044] Similarly, weighted positive ideal cloud In the positive ideal solution Based on this, the absolute weights of each underlying capability indicator are introduced, expressed as follows:

[0045] positive distance Based on this, we introduce cosine distance, which is expressed as:

[0046] in, For vectors, 1-norm For the first Weighted positive difference matrix of each assessment subject The Middle Column vector, and the first Positive distance of each performance metric Correspondingly, the weighted positive difference matrix of the evaluation object The specific form is as follows:

[0047] Since entropy and hyperentropy are smaller than the mean, and cosine distance itself is not sensitive to vector length, using cosine distance can better measure the coordinated change trend of these three dimensions simultaneously. It characterizes the similarity and distance difference between the numerical characteristics (mean, entropy, and hyperentropy) of each indicator of the evaluation object and the positive ideal solution, as follows:

[0048] ② Negative distance

[0049] Similarly, negative distance Decision cloud matrix representing weighted evaluation objects Weighted Negative Ideal Cloud The absolute difference, negative ideal cloud In negative ideal solution Based on this, absolute weights are introduced for each underlying capability indicator, specifically in the following form:

[0050] negative distance Based on this, we introduce cosine distance, as shown below:

[0051] in, For the first Weighted negative difference matrix of each assessment subject The Middle Column vector, and the first Negative distance of each performance metric Correspondingly, the weighted negative difference matrix of the evaluation object The specific form is as follows:

[0052] Among them, the cosine distance difference between the numerical characteristics (mean, entropy, and hyperentropy) of each indicator of the evaluation object and the negative ideal solution. Specifically, it takes the following form:

[0053] ③ The relative comprehensive ability of the assessment object

[0054] According to the TOPSIS method for calculating relative ability, then the first The relative comprehensive ability of each assessment subject is:

[0055] The above formula is used to calculate the relative comprehensive ability of multiple assessment objects, and the ability ranking can be completed through the calculation results, realizing the quantitative transformation of expert cognition. This is used as a constraint to update the quantitative formula of the relative ability ranking constraint of multiple assessment objects in the maximum entropy weight configuration construction module. The final weight configuration model under uncertain information is as follows.

[0056]

[0057] Step 3: Evaluation of index weights. The global optimization and convergence module is based on an adaptive particle swarm optimization mechanism, following... Figure 1 The process involves global random optimization to ultimately obtain the optimal weight solution. The specific process is as follows.

[0058] (1) Initialize the particle swarm, including the swarm size. Number of iterations T The position of each particle and speed and individual extreme values and global extrema Among them, the group size Based on the number of indicators M Determined, usually taken For each particle Generate random vectors ,in The initial position of the particle is determined by performing normalization on the random vector. The specific formula is as follows:

[0059] initial velocity ,in , Take a smaller random initial value and determine whether the initial particle swarm satisfies constraints ①, ②, ③, and ④. If it does, proceed with the subsequent process operations. If it does not, re-initialize the particle swarm until a particle swarm that meets the constraints is generated. (2) Calculate the fitness value of each particle. The specific formula is as follows:

[0060] (3) For each particle, use its fitness value and individual extreme values Comparison, if Then use Replace ; (4) For each particle, use its fitness and global extrema Comparison, if Then use Replace ; (5) Perform T In the next iteration, the particle velocity is updated. and location The speed update formula is expressed as follows:

[0061] in Indicates the first i The particle in the first t The speed of the next iteration Indicates the first i The particle in the first t The position of the next iteration. Indicates the first i The historical best position of each particle This represents the current position of the globally optimal particle. Two independent random numbers are used to control the randomness. , These represent two learning factors. This represents dynamically adjusted inertia weights, using a linearly decreasing inertia weight strategy, which decreases with the number of iterations. Down to The formula is expressed as follows:

[0062] The position update formula is expressed as follows: Represents the 1-norm

[0063] (6) After updating, check whether the position of the particle satisfies the constraint boundary conditions ①, ②, ③, and ④. If it does, continue to execute the subsequent steps (7); otherwise, correct the update process. (7) Determine whether the algorithm termination condition is met. If so, end the algorithm and output the optimal weight solution. Otherwise, return to step (2).

[0064] The units described in the embodiments of the present invention can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0065] According to one aspect of the present invention, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described above.

[0066] In another aspect, embodiments of the present invention also provide a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.

Claims

1. A rapid configuration system for maximum entropy weights of indicators based on a cloud model, characterized in that, include: Maximum entropy weight configuration module, relative comprehensive capability quantification module, evaluation index weight optimization module; The maximum entropy weight configuration module is used to write a first program using a computer and run the first program on the processor. Based on the principle of maximum entropy, it constructs an objective function and transforms the prior weight information of the evaluation index and the relative ability ranking among multiple evaluation objects into quantifiable numerical constraints, forming a weight configuration optimization model that combines objectivity and subjective experience. The relative comprehensive capability quantification module is used to write a second program on a computer and run the second program on the processor. Based on the TOPSIS approximation ideal solution ranking method and the cloud model, it integrates the cosine distance metric to convert the mean, entropy and hyperentropy of each evaluation index of each evaluation object into the distance between each evaluation object and the positive and negative ideal solutions, so as to achieve accurate quantification and ranking of comprehensive capabilities. The results are fed back to the maximum entropy weight configuration module to dynamically correct the constraint expression of relative capability ranking and enhance the adaptability and consistency of the model. The evaluation index weight optimization module is used to write a third program on a computer and run the third program on the processor. It adopts the particle swarm optimization algorithm to simulate the behavior of swarm intelligence, introduces a linear decreasing inertial weight strategy, and efficiently iterates in the search space to gradually approach the global optimal weight combination, and finally outputs a scientific and reasonable weight configuration scheme.

2. A method for fast configuration of maximum entropy weights for metrics based on a cloud model, characterized in that, The cloud model-based index maximum entropy weight fast configuration system according to claim 1 includes the following steps: Step 1: Based on the principle of maximum entropy, a priori cognition is fused and constraint modeling is performed. A weight modeling framework based on the principle of maximum entropy is constructed, which transforms the subjective preferences of experts for each evaluation index and the relative ranking relationship between evaluation objects into quantifiable mathematical constraints, thereby establishing a highly adaptive weight optimization model. Step 2: The relative comprehensive capability quantification module uses the cloud model to transform quantitative data with randomness and fuzziness into qualitative concepts. Under the TOPSIS method framework, it constructs a relative comprehensive capability quantification model for multiple assessment objects through positive and negative ideal solutions and a sub-functional module based on positive and negative distances. This model is then used as a constraint to update the quantification formula for the relative capability ranking constraint of multiple assessment objects in the maximum entropy weight configuration construction module. Step 3: Evaluation index weight global optimization and convergence module. Based on the adaptive particle swarm optimization mechanism, it performs global random optimization and finally obtains the optimal weight solution.

3. The method for fast configuration of maximum entropy weights for indicators based on cloud models according to claim 2, characterized in that, In step 1, the prior knowledge fusion and constraint modeling based on the maximum entropy principle constructs a weight modeling framework based on the maximum entropy principle. This transforms the experts' subjective preferences for each evaluation indicator and the relative ranking relationship between evaluation objects into quantifiable mathematical constraints, thereby establishing a highly adaptive weight optimization model. Specifically, this includes the following sub-steps: (1) Construct the objective function: For a hierarchical set of capability assessment indicators, it is necessary to aggregate capabilities using the quantified values ​​and absolute weights of the underlying capability assessment indicators, with the absolute weights of the capability assessment indicators as input parameters. Assume the number of underlying capability indicators in the constructed assessment indicator set is... Then, the objective function for constructing the weight allocation model under the maximum entropy criterion is: ； in, Let be the objective function. For the first The absolute weight of each evaluation indicator for The absolute weight vector of the evaluation index of dimension, i.e. . (2) Constructing constraints based on prior knowledge: ① When addressing evaluation issues, the indicator weights should be the normalized absolute weights of each indicator; therefore, their sum should be 1. ； ② The absolute weights of each indicator after normalization are not allowed to be less than zero and must all be less than 1, that is... ; ③ Based on industry experts' understanding of the evaluation objects, and acquiring accumulated generally accepted knowledge—that is, if the overall ability of one evaluation object is better than another—then the knowledge, after integration, can be converted into a mathematical expression as follows: ; in, For the first The relative ability value of each assessment subject. The number of objects in this recognized knowledge; ④ Based on the subjective preferences and prior knowledge of industry experts regarding each evaluation indicator, some indicators are defined as highly important indicators, meaning that certain indicators are far more important than others in the evaluation process. This prior knowledge is comprehensively quantified and expressed as a mathematical formula: ； in, This refers to the set of indicators that experts have identified as either highly important or low in importance. This is the indicator importance coefficient, with a value greater than 1. The final optimization problem, transformed from the weight allocation problem, is expressed mathematically as follows: 。 4. The method for fast configuration of maximum entropy weights for indicators based on cloud models according to claim 2, characterized in that, In step 2, the relative comprehensive capability quantification module uses a cloud model to transform quantitative data with randomness and fuzziness into qualitative concepts. Within the TOPSIS method framework, it constructs a relative comprehensive capability quantification model for multiple assessment objects through a sub-module that builds positive and negative ideal solutions and assesses relative comprehensive capabilities based on positive and negative distances. This model serves as a constraint to update the relative capability ranking constraint quantification formula for multiple assessment objects in the maximum entropy weight configuration construction module. Specifically, this includes the following sub-steps: (1) Construction of positive and negative ideal solutions: Assume the first The first assessment subject The number of quantitative sample data for each underlying capability indicator is: Based on cloud model theory, a reverse cloud generator is used to transform multiple sets of sample data into qualitative concepts with digital characteristics. ,in , and The first The first of the assessment subjects The performance decision cloud matrix of the evaluated object is determined by the mean, entropy, and hyperentropy of each evaluation metric. , ； The mean describes the basic certainty of a qualitative concept; a higher value indicates a higher basic certainty of the quantified data. Entropy and hyperentropy describe the degree of uncertainty of a qualitative concept; lower values ​​indicate lower fuzziness and randomness in the quantified data. Based on the definitions of these three numerical characteristics and following the construction principles of the positive ideal solution in the TOPSIS algorithm, the constructed positive ideal solution is as follows: ； Among them, the positive ideal mean solution positive ideal entropy solution and positive ideal superentropy solution The representation is as follows: ； Following the principle that the construction method for negative ideal solutions is the opposite of that for positive ideal solutions, the expression for negative ideal solutions is constructed as follows: ； In the above formula, the negative ideal mean solution Negative ideal entropy solution and negative ideal hyperentropy solution The representation is as follows: ； (2) Relative comprehensive ability assessment based on positive and negative cosine distance: ① Positive distance positive distance Used to characterize the Weighted evaluation object decision cloud matrix of each evaluation object Weighted positive ideal cloud The absolute difference, the weighted evaluation object decision cloud matrix Decision cloud matrix for evaluating object performance Based on this, absolute weights are introduced for each underlying capability indicator, specifically expressed as follows: ； Weighted Positive Ideal Cloud In the positive ideal solution Based on this, the absolute weights of each underlying capability indicator are introduced, expressed as follows: ； positive distance Based on this, we introduce cosine distance, which is expressed as: ； in, For vectors, 1-norm For the first Weighted positive difference matrix of each assessment subject The Middle Column vector, and the first Positive distance of each performance metric Correspondingly, the weighted positive difference matrix of the evaluation object The specific form is as follows: ； The similarity and distance difference between the numerical characteristics of each indicator of the object and the positive ideal solution are evaluated using cosine distance characteristics, as detailed below: ； ② Negative distance Negative distance Decision cloud matrix representing weighted evaluation objects Weighted Negative Ideal Cloud The absolute difference, negative ideal cloud In negative ideal solution Based on this, absolute weights are introduced for each underlying capability indicator, specifically in the following form: ； Negative distance Based on this, we introduce cosine distance, which takes the following form: ； in, For the first Weighted negative difference matrix of each assessment subject The Middle Column vector, and the first Negative distance of each performance metric Correspondingly, the weighted negative difference matrix of the evaluation object The specific form is as follows: ； Among them, the cosine distance difference between the numerical characteristics of each indicator of the evaluation object and the negative ideal solution. Specifically, it takes the following form: ； ③ The relative comprehensive ability of the assessment object According to the TOPSIS method for calculating relative ability, then the first The relative comprehensive ability of each assessment subject is: ； The above formula is used to calculate the relative comprehensive capabilities of multiple assessment objects, and the results are used to rank the capabilities, achieving a quantitative transformation of expert perception. This is then used as a constraint to update the quantitative formula for the relative capability ranking constraint of multiple assessment objects in the maximum entropy weight configuration construction module. The final weight configuration model under uncertain information is as follows: 。 5. The method for rapid configuration of maximum entropy weights for indicators based on cloud models according to claim 3, characterized in that, In step 3, the global optimization and convergence module for the evaluation index weights performs global random optimization based on an adaptive particle swarm optimization mechanism to finally obtain the optimal weight solution, specifically including the following sub-steps: (1) Initialize the particle swarm, including the swarm size. Number of iterations T The position of each particle and speed and individual extreme values and global extrema Among them, the group size Based on the number of indicators M Determined, for each particle Generate random vectors ,in The initial position of the particle is determined by performing normalization on the random vector. The specific formula is as follows: ； initial velocity ,in , Take a smaller random initial value and determine whether the initial particle swarm satisfies constraints ①, ②, ③, and ④. If it does, proceed with the subsequent process operations. If it does not, re-initialize the particle swarm until a particle swarm that meets the constraints is generated. (2) Calculate the fitness value of each particle. The specific formula is as follows: ； (3) For each particle, use its fitness value and individual extreme values In comparison, In the case of, use Replace ; (4) For each particle, use its fitness and global extrema In comparison, In the case of, use Replace ; (5) Perform T In the next iteration, the particle velocity is updated. and location The speed update formula is expressed as follows: ； in, Indicates the first i The particle in the first t The speed of each iteration Indicates the first i The particle in the first t The position of the next iteration. Indicates the first i The historical best position of each particle This represents the current position of the globally optimal particle. Two independent random numbers are used to control the randomness. , These represent two learning factors. This represents dynamically adjusted inertia weights, using a linearly decreasing inertia weight strategy, which decreases with the number of iterations. Down to The formula is expressed as follows: ； The position update formula is expressed as follows: Represents the 1-norm ； (6) After updating, check whether the position of the particle satisfies the constraint boundary conditions ①, ②, ③, and ④. If it does, continue to execute the subsequent steps (7); otherwise, correct the update process. (7) Determine whether the algorithm termination condition is met. If so, end the algorithm and output the optimal weight solution. Otherwise, return to step (2).

6. The method for fast configuration of maximum entropy weights for indicators based on cloud models according to claim 5, characterized in that, Group size .

7. A device for rapid configuration of maximum entropy weights for indicators based on a cloud model, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when loaded by the processor, executes the method as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, A computer program is stored in a readable storage medium, which, when loaded by a processor, executes the method as described in any one of claims 1 to 6.

Images