Type II fuzzy model and LLM-based double-creation event scoring method and system
By combining a type II fuzzy model and a large language model in the scoring method, the problem of inconsistent scoring standards in innovation and entrepreneurship competitions is solved, resulting in more objective and stable scoring results that can adapt to the scoring needs of different competitions and fields.
Patent Information
- Application Number
- CN202511598142.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-04
- Publication Date
- 2026-01-30
AI Technical Summary
Existing scoring methods for innovation and entrepreneurship competitions lack uniformity and transparency, are greatly influenced by the subjective factors of experts, and are difficult to accurately reflect the core competitiveness of projects in a dynamic and changing environment. Traditional models cannot effectively handle ambiguity and uncertainty.
A scoring method based on Type 2 fuzzy model and Large Language Model (LLM) is adopted. The fuzziness and uncertainty are handled by the Type 2-Fuzzy model, and the cumulative prospect theory (CPT) is combined to dynamically adjust the weights, reduce subjective bias, and improve the robustness and fairness of the scoring.
It significantly improves the objectivity and consistency of scoring, reduces subjective bias, enhances scoring efficiency, adapts to the scoring needs of different competitions and fields, and ensures the stability and fairness of scoring results.
Smart Images
Figure CN121436780A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a scoring method and system for innovation and entrepreneurship competitions based on type II fuzzy models and LLM. Background Technology
[0002] Various innovation and entrepreneurship competitions serve as important platforms for selecting high-quality projects and cultivating innovative entities. These competitions require weighting for dimensions such as innovativeness, feasibility, and market potential. However, differences in field background and goal orientation among different competitions lead to a lack of uniformity in scoring standards. For example, competitions emphasizing technological innovation may assign higher weight to the advancement and breakthrough of technology, while competitions focusing on marketability tend to focus more on the project's business model and market adaptability. This significant difference in weighting makes it difficult for participants to accurately understand the judging criteria, affecting the fairness and transparency of the competitions. Current competition scoring relies heavily on expert review, and the results are easily influenced by subjective factors such as the background, experience, emotions, and preferences of the reviewers. For example, some experts may assign higher scores to projects in a particular technological field due to their familiarity with that field, while underestimating the innovativeness or value of projects in other fields. Furthermore, inconsistencies in scoring standards among experts also lead to significant fluctuations in scoring results; the cross-review consistency coefficient is often below 0.7, making it difficult to guarantee the stability of the scores. These subjective biases not only affect the fairness of the scoring but may also increase the controversy surrounding project ranking results. Traditional scoring methods typically use fixed standards or single indicators, making it difficult to comprehensively depict the multidimensional value of a project. Especially in a dynamically changing environment, the scoring results often fail to accurately reflect the project's core competitiveness.
[0003] Type 1 fuzzy models are a fundamental tool for handling fuzzy information and uncertainty, assigning a definite membership value to each element through a single membership function. However, in complex scoring scenarios, evaluation indicators often exhibit high uncertainty and dynamic changes. For example, differences in the understanding of "innovation" among experts in different fields can lead to significant fluctuations in the membership degrees of evaluation criteria. Type 1 fuzzy models cannot effectively handle these multiple uncertainties because their membership functions are fixed and cannot be dynamically adjusted to adapt to the complexity of the scoring scenario. This limitation results in insufficient robustness and adaptability of the scoring results.
[0004] Traditional machine learning models, such as random forests and support vector machines, can automate scoring to some extent, but their ability to extract semantic information from text is limited, relying primarily on the analysis of explicit indicators. For example, machine learning models can handle numerical data in fixed formats (such as sales revenue or number of patents), but struggle to capture features implicit in project proposals or presentation texts, such as innovative points and teamwork capabilities. Furthermore, these models typically require substantial prior knowledge and manual feature engineering, resulting in low efficiency and an inability to dynamically adapt to complex and ever-changing scoring scenarios. Summary of the Invention
[0005] This invention aims to address at least one of the technical problems existing in related technologies. To this end, this invention provides a scoring method and system for innovation and entrepreneurship competitions based on a Type 2-Fuzzy model and a Large Language Model (LLM). The Type 2-Fuzzy model handles fuzziness and uncertainty, making the scoring results more objective and significantly improving the robustness of the scoring. The Large Language Model dynamically adjusts weights to adapt to the scoring needs of different competitions and fields. The collaborative scoring mechanism combining the Type 2-Fuzzy model, the Large Language Model, and Cumulative Prospect Theory (CPT) reduces subjective bias, improves scoring efficiency, and ensures fairness and consistency.
[0006] This invention provides a scoring method and system for innovation and entrepreneurship competitions based on type-two fuzzy models and LLM, comprising: S1: Construct a scoring index system for innovation and entrepreneurship competitions and obtain project evaluation criteria; S2: Calculate the importance of experts using linguistic variables represented by intuitionistic fuzzy numbers, obtain the weights of each expert in the decision-making group, and calculate the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria based on the weights of each expert. S3: Based on the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria, construct an interval type II fuzzy number fuzzy decision matrix, and calculate the positive and negative prospect values of each project based on the interval type II fuzzy number fuzzy decision matrix; S4: Calculate the Type2-Fuzzy model weights of each project evaluation criterion and the large language model integration weights of each project evaluation criterion, and fuse the Type2-Fuzzy model weights and the large language model integration weights of the project evaluation criterion to obtain the fusion weights of each project evaluation criterion. S5: Calculate the cumulative prospect weight of each project's evaluation criteria based on the fusion weight of each project's evaluation criteria; S6: Calculate the comprehensive prospect value of each project based on the cumulative prospect weight of each project's evaluation criteria, the positive prospect value and the negative prospect value of each project, and rank the projects based on their comprehensive prospect values.
[0007] Furthermore, the scoring index system for the innovation and entrepreneurship competition includes innovation ability, creativity ability, entrepreneurial ability, presentation performance, and copywriting quality; the project evaluation criteria include dimensions of innovation ability, creativity ability, entrepreneurial ability, presentation performance, and copywriting quality. The dimensions of innovation capability include criteria for product innovation, technological innovation, business model innovation, and service innovation. The dimensions of creative ability include the criteria for the rationality of planning and the criteria for project feasibility; Entrepreneurial capability dimensions include the practical effectiveness criterion and the resource integration criterion; The dimensions of presentation performance include the principles of teamwork, clarity of expression, and problem-solving. The dimensions of copywriting quality include the criteria of logical rigor and the criteria of content completeness.
[0008] Furthermore, the importance of experts is reflected through linguistic variables represented by intuitionistic fuzzy numbers, calculated as follows: in, For the first The weight of each expert, For the first The degree of affiliation of an expert For the first The degree of non-membership of each expert For the first The degree of hesitation of an expert The number of experts.
[0009] Furthermore, step S2 includes: Convert linguistic variables into interval type II fuzzy numbers; The fuzzy scores of each expert for the project evaluation criteria are calculated based on the interval type II fuzzy numbers. The comprehensive weight of the project evaluation criteria is calculated based on the fuzzy scores given by each expert to the project evaluation criteria and the weights of each expert. The project scores corresponding to the project evaluation criteria of each expert are obtained based on the interval type II fuzzy number. The average project score corresponding to the project evaluation criteria is calculated based on the project scores given by each expert according to the evaluation criteria and the weight of each expert.
[0010] Furthermore, the positive and negative foreground values for each item are calculated based on the fuzzy decision matrix of the interval type II fuzzy number, including: Calculate the positive and negative ideal solutions for the evaluation criteria of each project based on the fuzzy decision matrix of the interval type II fuzzy number; Calculate the positive fuzziness between each item and the positive ideal solution of each item's evaluation criteria, and the negative fuzziness between each item and the negative ideal solution of each item's evaluation criteria. Defuzzify the positive and negative ambiguities to obtain the profit and loss values for each project; The profit and loss values of each project are normalized, and the positive and negative prospect values of each project are calculated based on the normalized profit and loss values.
[0011] Furthermore, step S4 includes: The linguistic variables that reflect the importance of the evaluation criteria for each project are converted into interval type II fuzzy numbers; The Type 2-Fuzzy model weights of each project's evaluation criteria are obtained by defuzzifying the interval type 2 fuzzy numbers.
[0012] Furthermore, step S4 also includes training at least one large language model, inputting a preset scoring weight instruction into the large language model to obtain the item evaluation criterion weights of multiple large language models, and integrating the item evaluation criterion weights of multiple large language models to obtain the integrated weights of the large language models. The large language models include kimi2, doubao, and deepseek.
[0013] Furthermore, the formula for calculating the cumulative prospect weight is: in, For positive prospects, negative prospect weight, The fusion weights for project evaluation criteria, For the first The Type2-Fuzzy model weights of the evaluation criteria for each project. For the return decay parameter, This is the loss attenuation parameter.
[0014] Furthermore, the formula for calculating the overall foreground value is as follows: in, For the project The overall prospects value The number of project evaluation criteria. For the number of projects, For positive prospects, negative prospect weight, For the first Project No. Positive prospect values for each project evaluation criterion. For the first Project No. The negative prospect value of the project evaluation criteria.
[0015] This invention also provides a scoring system for innovation and entrepreneurship competitions based on a type-II fuzzy model and LLM, used to execute the aforementioned scoring method for innovation and entrepreneurship competitions based on a type-II fuzzy model and LLM, comprising: The construction module builds a scoring index system for innovation and entrepreneurship competitions and obtains project evaluation criteria. The first calculation module calculates the importance of experts using linguistic variables represented by intuitionistic fuzzy numbers, obtains the weights of each expert in the decision-making group, and calculates the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria based on the weights of each expert. The second calculation module constructs an interval type II fuzzy number fuzzy decision matrix based on the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria, and calculates the positive and negative foreground values of each project based on the interval type II fuzzy number fuzzy decision matrix. The weight fusion module calculates the Type2-Fuzzy model weights of each project evaluation criterion and the large language model integration weights of each project evaluation criterion, and merges the Type2-Fuzzy model weights and the large language model integration weights of the project evaluation criterion to obtain the fusion weights of each project evaluation criterion. The third calculation module calculates the cumulative prospect weight of each project evaluation criterion based on the fusion weight of each project evaluation criterion. The comprehensive calculation module calculates the comprehensive prospect value of each project based on the cumulative prospect weight of each project's evaluation criteria, the positive prospect value, and the negative prospect value of each project, and sorts the projects based on their comprehensive prospect values.
[0016] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: By using a Type2-Fuzzy model to handle fuzziness and uncertainty, scoring results are made more objective, significantly improving the robustness of the scoring. A large language model dynamically adjusts weights to adapt to the scoring needs of different competitions and fields. A collaborative scoring mechanism combining the Type2-Fuzzy model, the large language model, and Cumulative Prospect Theory (CPT) reduces subjective bias, improves scoring efficiency, and ensures fairness and consistency.
[0017] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0019] Figure 1 This is a flowchart illustrating a scoring method for innovation and entrepreneurship competitions based on a type-two fuzzy model and LLM provided by the present invention.
[0020] Figure 2 This is a schematic diagram of the structure of an innovation and entrepreneurship competition scoring system based on a type II fuzzy model and LLM provided by the present invention.
[0021] Figure label: 101. Construction Module; 102. First Calculation Module; 103. Second Calculation Module; 104. Weight Fusion Module; 105. Third Calculation Module; 106. Comprehensive Calculation Module. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but cannot be used to limit the scope of this invention.
[0023] In the description of the embodiments of the present invention, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0024] The following is combined Figures 1 to 2 This invention describes a scoring method and system for innovation and entrepreneurship competitions based on a type-two fuzzy model and LLM.
[0025] like Figure 1 As shown, a scoring method for innovation and entrepreneurship competitions based on type-II fuzzy models and LLM includes: S1: Construct a scoring index system for innovation and entrepreneurship competitions and obtain project evaluation criteria; The scoring system for the innovation and entrepreneurship competition includes innovation ability, creativity ability, entrepreneurial ability, presentation performance, and copywriting quality. Innovation capability includes product innovation, technological innovation, business model innovation, and service innovation; Product innovation refers to the novelty and uniqueness of a product in terms of its functions, design, and features, and whether it meets new needs or solves existing problems.
[0026] Technological innovation refers to the level of advancement of the technology used in a project and whether it has a technological advantage over similar projects.
[0027] Model innovation refers to the novelty and uniqueness of a business model or profit model, and whether it can bring new development opportunities to the project.
[0028] Service innovation refers to the novelty of the service's form, content, quality, etc., and whether it brings a better experience to the user.
[0029] Creative ability includes the rationality of planning and the feasibility of the project.
[0030] The rationality of the plan refers to whether the design of the business model, marketing model, technology model, and financial support is scientific, reasonable, and creative.
[0031] Project feasibility is the operability and feasibility of an idea, including feasibility analysis from economic, managerial, technical, and market perspectives.
[0032] Entrepreneurial ability includes practical effectiveness and resource integration.
[0033] Practical effectiveness refers to the actual results and outcomes of practical activities, including preparation for starting a business, company registration, e-commerce marketing, and business performance.
[0034] Resource integration capability refers to the ability to identify, acquire, integrate, and efficiently utilize internal and external resources such as capital, networks, supply chains, and policies.
[0035] Presentation performance includes teamwork, clarity of expression, and problem-solving skills. Teamwork refers to whether the division of labor among team members is reasonable and efficient.
[0036] Clarity of expression refers to whether the content of a speech is expressed accurately, clearly, and easily understood, and whether it highlights the key points and advantages of the project.
[0037] Coping skills refer to the ability to respond to questions from judges or unexpected situations, and whether the answers are reasonable, complete, and organized.
[0038] Copywriting quality includes logical rigor and content completeness.
[0039] Logical rigor refers to whether the logical structure of the text and presentation slides is reasonable and rigorous, and whether the content is organized clearly and systematically.
[0040] Content completeness refers to whether the written materials and presentation slides comprehensively and in detail introduce all aspects of the project, and whether there are any omissions or deficiencies.
[0041] The project evaluation criteria include dimensions of innovation ability, creativity, entrepreneurship, presentation performance, and copywriting quality. The dimensions of innovation capability include criteria for product innovation, technological innovation, business model innovation, and service innovation. The dimensions of creative ability include the criteria for the rationality of planning and the criteria for project feasibility; Entrepreneurial capability dimensions include the practical effectiveness criterion and the resource integration criterion; The dimensions of presentation performance include the principles of teamwork, clarity of expression, and problem-solving. The dimensions of copywriting quality include the criteria of logical rigor and the criteria of content completeness.
[0042] S2: Calculate the importance of experts using linguistic variables represented by intuitionistic fuzzy numbers, obtain the weights of each expert in the decision-making group, and calculate the comprehensive weight of the project evaluation criteria and the project score corresponding to the project evaluation criteria based on the weights of each expert. A decision-making panel is formed by inviting several knowledgeable and experienced experts. Assume the panel consists of K experts, and the importance of each expert is represented by linguistic variables expressed as intuitionistic fuzzy numbers (IFNs), in the form of... The expression for calculating expert weights is: in, For the first The weight of each expert, For the first The degree of affiliation of an expert For the first The degree of non-membership of each expert For the first The degree of hesitation of an expert The number of experts.
[0043] Experts used a linguistic variable representing the importance of sub-criterions to assess the importance of each sub-criterion, and a linguistic variable representing the scoring of sub-criterions to assess the item scores under each sub-criterion.
[0044] Experts used Interval Type II Fuzzy Numbers (IT2FNs) to weight and score the selected sub-criteria for the alternatives.
[0045] Experts used language variables to assess the importance of the project evaluation criteria and the corresponding project scores for those criteria. S21: Convert linguistic variables into interval type II fuzzy numbers; In some specific embodiments of the present invention, the correspondence between the linguistic variables of the importance of the sub-criteria and the interval type II fuzzy numbers is shown in Table 1, and the correspondence between the linguistic variables of the project performance rating in the sub-criteria and the interval type II fuzzy numbers is shown in Table 2.
[0046] Table 1. Correspondence between linguistic variables and interval type II fuzzy numbers in relation to the importance of sub-criterions. Table 2. Correspondence between linguistic variables and interval type II fuzzy numbers in the sub-criteria for project performance rating. The fuzzy scores of each expert for the project evaluation criteria are calculated based on the interval type II fuzzy numbers. The comprehensive weight of the project evaluation criteria is calculated based on the fuzzy scores given by each expert to the criteria and the weights of each expert. The calculation expression is as follows: in, For the first The overall weight of the evaluation criteria for each project For the first Project Evaluation Criteria The overall weight of each expert, For the first The weight of each expert, .
[0047] The project scores corresponding to the project evaluation criteria of each expert are obtained based on the interval type II fuzzy number. The average project score corresponding to the project evaluation criteria is calculated based on the project scores given by each expert according to the project evaluation criteria and the weight of each expert. The formula for calculating the average project score is: in, For the first Project about the The average score of each project according to the evaluation criteria. For the first The expert on the first Project about the Project scoring based on project evaluation criteria .
[0048] S3: Based on the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria, construct an interval type II fuzzy number fuzzy decision matrix, and calculate the positive and negative prospect values of each project using the interval type II fuzzy number fuzzy decision matrix; Assume the decision problem includes A startup project and Project evaluation criteria. Experts used interval type-two fuzzy numbers (IT2FNs) to weight and score the selected sub-criteria for the alternatives. Using interval type-two fuzzy numbers, the upper and lower bounds of the matrix elements and the weights of the project evaluation criteria can be defined as an interval.
[0049] According to the Project about the The project average score is constructed based on the project evaluation criteria. The matrix is used to obtain the interval type II fuzzy number fuzzy decision matrix.
[0050] Based on the established interval type-2 fuzzy number (IT2FNs) fuzzy decision matrix, the estimated values of alternative solutions related to the evaluation criteria of each project are obtained. Then, the interval type-2 fuzzy sets of all projects relative to the sub-criteria are sorted.
[0051] Calculate the positive and negative ideal solutions for the evaluation criteria of each project based on the fuzzy decision matrix of the interval type II fuzzy number; The expression for calculating the ideal solution is: in, For the first The positive ideal solution of the evaluation criteria for each project For the first The lower bound fuzzy evaluation of the first dimension of the project. For the first The lower bound fuzzy evaluation of the second dimension of the project. For the first The lower bound fuzzy evaluation of the third dimension of the project. For the first The lower bound fuzzy evaluation of the fourth dimension of the project. The first fuzzy membership function, For the first Project No. Fuzzy evaluation of the lower limit of the evaluation criteria for each project. This is the second fuzzy membership function; For the first The upper limit of the first dimension of the project is fuzzy. For the first The upper limit of the second dimension of the project is fuzzy. For the first The upper limit of the third dimension of the project is fuzzy. For the first The upper limit of the fourth dimension of the project is fuzzy. For the first Project No. The upper limit of the evaluation criteria for each project is fuzzy. To find the maximum value; The expression for calculating the negative ideal solution is: in, For the first The negative ideal solution of the evaluation criteria for each project To find the minimum value; Calculate the positive fuzziness between each item and the positive ideal solution of each item's evaluation criteria, and the negative fuzziness between each item and the negative ideal solution of each item's evaluation criteria. in, For the first fuzzy set Second fuzzy set Distance function between The lower bound value for quantization of the first dimension interval. The upper limit of the quantization for the first dimension interval. The lower limit of the second dimension interval quantization. The upper limit of the second dimension interval quantization. The lower limit value for quantization of the third dimension interval. The upper limit of the quantization for the third dimension interval. The lower limit value for the fourth dimension interval quantization. The upper limit of the quantization for the fourth dimension interval. for The lower bound fuzzy set, for The lower bound fuzzy set, for The upper limit of the fuzzy set, for The upper limit of the fuzzy set; The following formula can be used to obtain: in, For the first The upper limit of the interval quantization of the dimension or the first The lower bound of the interval quantization of a dimension. It can be the upper or lower limit value; For the first project Upper or lower bound fuzzy evaluation of each dimension. For the second project Upper or lower bound fuzzy evaluation of each dimension. This is the upper limit value. This is the upper limit value; The following formula can be used to obtain: in, For the first The upper limit of the interval quantization of the dimension or the first interval quantization. The lower bound of the interval quantization of a dimension. For the first Fuzzy membership function for upper bound fuzzy set or lower bound fuzzy set For the first project Upper or lower bound fuzzy evaluation of each dimension. For the second project Upper or lower bound fuzzy evaluation of each dimension; The positive fuzziness between each item and the positive ideal solution of each item's evaluation criteria is quantified using a distance function; the negative fuzziness between each item and the negative ideal solution of each item's evaluation criteria is quantified using a distance function.
[0052] Defuzzify the positive and negative ambiguities to obtain the profit and loss values for each project; The calculation expression is: in, The profit and loss value after deblurring. For the first A fuzzy set of projects.
[0053] The profit and loss values of each project are normalized, and the positive and negative prospect values of each project are calculated based on the normalized profit and loss values.
[0054] The profit and loss values are normalized, and the calculation expression is as follows: in, For the first Project No. Normalized return values based on project evaluation criteria For the first Project about the Project scoring based on project evaluation criteria For the first Project about the The distance function between the average project score and the negative ideal solution of each project evaluation criterion; in, For the first Project No. Normalized loss values for each project evaluation criterion. For the first Project about the The distance function between the average project score and the positive ideal solution of the project evaluation criteria.
[0055] The formula for calculating the foreground value is: in, For the first Project No. Positive prospect values for each project evaluation criterion. For the first Project No. Negative prospective value of each project evaluation criterion This is the marginal sensitivity coefficient for revenue. The marginal sensitivity coefficient is the loss coefficient. This is the loss aversion coefficient; In some specific embodiments of the present invention, reference is made to Tversky's empirical research, , as well as The possible values are: , .
[0056] S4: Calculate the Type2-Fuzzy model weights of each project evaluation criterion and the large language model integration weights of each project evaluation criterion, and fuse the Type2-Fuzzy model weights and the large language model integration weights of the project evaluation criterion to obtain the fusion weights of each project evaluation criterion. The linguistic variables that reflect the importance of the evaluation criteria for each project are converted into interval type II fuzzy numbers; The Type 2-Fuzzy model weights of each project evaluation criterion are obtained by defuzzifying the interval type 2 fuzzy numbers. Since the rationality of the project evaluation criterion weights directly affects the final result during the decision-making process, this invention treats the importance of each project evaluation criterion as a linguistic variable, which is further transformed into interval type 2 fuzzy numbers. The defuzzified values obtained according to step S3 are the precise weights of the type 2-fuzzy model of each project evaluation criterion.
[0057] Train at least one large language model, input a preset scoring weight instruction into the large language model, obtain the item evaluation criterion weights of multiple large language models, and integrate the item evaluation criterion weights of multiple large language models to obtain the integrated weight of the large language model.
[0058] In some specific embodiments of the present invention, the instruction is: "You are a judge of an innovation and entrepreneurship competition. You are designing a scoring index system for the competition. Please refer to the scoring dimensions, project evaluation criteria, and index descriptions of the competition in the document, and assign weights to each project evaluation criterion. Requirements: Directly output the weight of each project evaluation criterion in tabular form. The sum of the weights must be 100%."
[0059] The large language models include kimi2, doubao, and deepseek.
[0060] To improve the representativeness and reliability of the output weights of large language models, this study systematically integrates the output weights of three large language models to construct an integrated weight for the large language model. This ensures that it can more comprehensively capture the core information of the target evaluation dimensions and effectively avoid the biases and limitations that may exist in the output weights of a single model. The expression for calculating the integration weights of the large language model is: in, For the first The evaluation criteria for each project are integrated into a large language model with weights. The number of project evaluation criteria. For the first model, the first The weights of the evaluation criteria for each project. For the second model The weights of the evaluation criteria for each project. For the third model The weights of the evaluation criteria for each project. For the first model, the first The weights of the evaluation criteria for each project. For the second model The weights of the evaluation criteria for each project. For the third model The weight of each project's evaluation criteria.
[0061] The formula for calculating the fusion weight of the project evaluation criteria is as follows: in, The fusion weights for project evaluation criteria, For the first The Type2-Fuzzy model weights of the evaluation criteria for each project.
[0062] S5: Calculate the cumulative prospect weight of each project's evaluation criteria based on the fusion weight of each project's evaluation criteria. The calculation expression is as follows: in, For positive prospects, negative prospect weight, The fusion weights for project evaluation criteria, For the first The Type2-Fuzzy model weights of the evaluation criteria for each project. For the return decay parameter, This is the loss attenuation parameter.
[0063] In some specific embodiments of the present invention , .
[0064] S6: Calculate the comprehensive prospect value based on the cumulative prospect weight and the positive and negative prospect values of each project, and sort the projects according to the comprehensive prospect value.
[0065] The formula for calculating the overall foreground value is: in, For the project The overall prospects value The number of project evaluation criteria. For the number of projects, For positive prospects, negative prospect weight, For the first Project No. Positive prospect values for each project evaluation criterion. For the first Project No. The negative prospect value of the project evaluation criteria.
[0066] The projects are ranked based on their overall prospects.
[0067] like Figure 2As shown, an innovation and entrepreneurship competition scoring system based on a type-II fuzzy model and LLM is used to implement an innovation and entrepreneurship competition scoring method based on a type-II fuzzy model and LLM, including: Module 101 constructs a scoring index system for innovation and entrepreneurship competitions to obtain project evaluation criteria; The first calculation module 102 calculates the importance of experts using linguistic variables represented by intuitionistic fuzzy numbers, obtains the weights of each expert in the decision-making group, and calculates the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria based on the weights of each expert. The second calculation module 103 constructs an interval type II fuzzy number fuzzy decision matrix based on the comprehensive weight of the project evaluation criteria and the average project score corresponding to the project evaluation criteria, and calculates the positive and negative foreground values of each project based on the interval type II fuzzy number fuzzy decision matrix. The weight fusion module 104 calculates the Type2-Fuzzy model weights of each project evaluation criterion and the large language model integration weights of each project evaluation criterion, and merges the Type2-Fuzzy model weights and the large language model integration weights of the project evaluation criterion to obtain the fusion weights of each project evaluation criterion. The third calculation module 105 calculates the cumulative prospect weight of each project evaluation criterion based on the fusion weight of each project evaluation criterion; The comprehensive calculation module 106 calculates the comprehensive prospect value of each project based on the cumulative prospect weight of each project's evaluation criteria and the positive and negative prospect values of each project, and sorts the projects according to their comprehensive prospect values.
[0068] Through the collaborative work of the above modules, the Type2-Fuzzy model handles fuzziness and uncertainty, making the scoring results more objective and significantly improving the robustness of the scoring. The large language model dynamically adjusts weights to adapt to the scoring needs of different competitions and fields. The collaborative scoring mechanism combining the Type2-Fuzzy model, the large language model, and Cumulative Prospect Theory (CPT) reduces subjective bias, improves scoring efficiency, and ensures fairness and consistency.
[0069] This invention demonstrates significant advantages in multiple dimensions for evaluating innovation and entrepreneurship competition projects. It employs a Type 2-Fuzzy model with interval type II fuzzy sets (IT2FNs), constructing an uncertainty footprint through upper membership functions (UMF) and lower membership functions (LMF). This allows for the simultaneous quantification of three core uncertainties: "disagreement among judges," "ambiguity in linguistic evaluation," and "drift of indicator boundaries." By transforming fuzzy linguistic evaluations such as "moderate innovation" and "good presentation performance" into trapezoidal type II fuzzy numbers, the accuracy of resolving evaluation fuzziness is significantly improved compared to the traditional Type-1 fuzzy model. Furthermore, when calculating scoring differences using the improved interval distance formula, it simultaneously reflects both numerical deviation and differences in fuzziness, making the scoring results more closely reflect actual evaluation scenarios.
[0070] By combining the value function and weight function of Cumulative Prospect Theory (CPT), the judges' "loss aversion" and "probability distortion" behaviors are corrected. Among these, , , , The system quantifies the judges' oversensitivity to "project weaknesses" (such as the dragging effect of insufficient technological innovation on the overall score) and their overestimation bias of "low-probability risks" (such as the interference of extreme cases on market feasibility). The overall prospect value ranking remained stable after multiple rounds of parameter sensitivity testing, verifying the robustness of the scoring.
[0071] Traditional scoring processes, especially complex review tasks involving multiple judges, are often inefficient and susceptible to subjective biases. The personal preferences, knowledge backgrounds, emotional states, and even fatigue levels of different judges can have unpredictable effects on the scoring results, leading to a lack of consistency and comparability. The Type2-Fuzzy model, through expert weight calculation and fuzzy decision matrix aggregation, significantly reduces the review time of eight cross-disciplinary experts (including corporate executives and university professors), resulting in a substantial improvement in efficiency. Simultaneously, LLM, leveraging its semantic understanding capabilities developed through training on massive amounts of text, can automatically extract implicit features such as innovation points and team capabilities from business plans, significantly reducing reliance on manual feature engineering and improving scoring decoding efficiency. Furthermore, the generation of the "Scoring Basis Explanation" module (such as feedback on the project's strengths and weaknesses in the "Practical Effectiveness" dimension) requires no additional manual writing.
[0072] Leveraging its powerful natural language understanding, generation, and reasoning capabilities, LLM can dynamically parse and understand rule text, background information, and user feedback in different scoring scenarios. Based on these dynamic inputs, LLM can adjust the weights of various scoring indicators in real time, thereby constructing a highly customized scoring model. The combination of LLM's dynamic weight generation and Type2-Fuzzy's dynamic membership function allows for flexible adaptation to the evaluation emphases of different competitions. In competitions emphasizing technological breakthroughs, the overall weight of "technological innovation" can be dynamically increased by adjusting input instructions compared to competitions emphasizing marketability, resulting in a lower domain adaptation error compared to fixed-standard scoring. LLM can identify implicit service innovations from business plans, while Type2-Fuzzy integrates fuzzy expert evaluations, ultimately achieving a score 15% higher than traditional methods.
[0073] This invention is not only applicable to scenarios involving a large number of innovative and fuzzy evaluation factors, such as innovation and entrepreneurship competitions, but can also be extended to many fields that require handling fuzzy multi-criteria decision-making, such as product design evaluation, investment project decision-making, and talent selection. It has broad practicality and promotional value.
[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A two-creative event scoring method based on a two-type fuzzy model and LLM, characterized in that, The method comprises the following steps: S1: constructing a double-creation event scoring index system to obtain project evaluation criteria; S2: calculating the importance of experts through language variables represented by intuitionistic fuzzy numbers, obtaining the weights of experts in the decision-making group, and calculating the comprehensive weight of the project evaluation criteria and the average score of the project corresponding to the project evaluation criteria according to the weights of experts; S3: constructing an interval type-2 fuzzy number fuzzy decision matrix according to the comprehensive weight of the project evaluation criteria and the average score of the project corresponding to the project evaluation criteria, and calculating the positive prospect value and the negative prospect value of each project according to the interval type-2 fuzzy number fuzzy decision matrix; S4: calculating the Type2-Fuzzy model weight of each project evaluation criterion and the large language model integration weight of each project evaluation criterion, fusing the Type2-Fuzzy model weight and the large language model integration weight of the project evaluation criteria to obtain the fusion weight of each project evaluation criterion; S5: calculating the cumulative prospect weight of each project evaluation criterion according to the fusion weight of each project evaluation criterion; S6: calculating the comprehensive prospect value of each project according to the cumulative prospect weight of each project evaluation criterion, the positive prospect value and the negative prospect value of each project, and ranking the projects according to the comprehensive prospect value of each project.
2. The method according to claim 1, wherein, The double-creation event scoring index system comprises innovation ability, creativity ability, entrepreneurship ability, speech performance and copy quality; the project evaluation criteria comprise innovation ability dimension, creativity ability dimension, entrepreneurship ability dimension, speech performance dimension and copy quality dimension; The innovation ability dimension comprises product innovation criterion, technical innovation criterion, mode innovation criterion and service innovation criterion; The creativity ability dimension comprises planning rationality criterion and project feasibility criterion; The entrepreneurship ability dimension comprises practical effectiveness criterion and resource integration criterion; The speech performance dimension comprises team cooperation criterion, expression clarity criterion and coping processing criterion; The copy quality dimension comprises logic rigor criterion and content integrity criterion.
3. The method according to claim 1, wherein, The importance of experts is embodied by language variables represented by intuitionistic fuzzy numbers, and the calculation expression is: wherein, is the weight of the th expert, is the membership of the th expert, is the non-membership of the th expert, is the hesitancy of the th expert, is the number of experts.
4. The method according to claim 1, wherein, The step S2 comprises: converting the language variables into interval type-2 fuzzy numbers; calculating the fuzzy scores of the project evaluation criteria by each expert according to the interval type-2 fuzzy numbers; calculating the comprehensive weight of the project evaluation criteria according to the fuzzy scores of the project evaluation criteria by each expert and the weights of experts; obtaining the scores of the project corresponding to the project evaluation criteria by each expert according to the interval type-2 fuzzy numbers; calculating the average scores of the project corresponding to the project evaluation criteria according to the scores of the project corresponding to the project evaluation criteria by each expert and the weights of experts.
5. The method according to claim 1, wherein, The calculation of the positive prospect value and the negative prospect value of each project according to the interval type-2 fuzzy number fuzzy decision matrix comprises: calculating the positive ideal solution and the negative ideal solution of each project evaluation criterion according to the interval type-2 fuzzy number fuzzy decision matrix; calculating the positive fuzzy degree between each project and the positive ideal solution of each project evaluation criterion and the negative fuzzy degree between each project and the negative ideal solution of each project evaluation criterion; defuzzifying the positive fuzzy degree and the negative fuzzy degree to obtain the benefit value and the loss value of each project; The benefit value and loss value of each item are normalized, and the positive prospect value and negative prospect value of each item are calculated according to the normalized benefit value and loss value.
6. The method according to claim 1, wherein, The S4 step comprises: Converting the language variable representing the importance of each item evaluation criterion into an interval Type2 fuzzy number; De-fuzzification processing is performed on the interval Type2 fuzzy number to obtain the Type2-Fuzzy model weight of each item evaluation criterion.
7. The method according to claim 6, wherein, The S4 step further comprises training at least one large language model, inputting a preset scoring weight instruction to the large language model, obtaining the item evaluation criterion weight of multiple large language models, and integrating the item evaluation criterion weight of multiple large language models to obtain the large language model integrated weight; The large language model comprises kimi2, doubao and deepseek.
8. The double-creation event scoring method based on the Type2 fuzzy model and the LLM according to claim 1, wherein, The calculation expression of the cumulative prospect weight is: wherein, is a positive foreground weight, is a negative foreground weight, is a fusion weight of the project evaluation criteria, is a Type2-Fuzzy model weight of the th project evaluation criterion, is a gain decay parameter, is a loss decay parameter.
9. The method according to claim 1, wherein, The calculation expression of the comprehensive prospect value is: wherein, is the overall prospect value for the project is the project evaluation criterion number, is the project number, is the positive prospect weight, is the negative prospect weight, is the prospect value for the is the prospect value for the is the positive prospect value for the project evaluation criterion number for the project number, is the negative prospect value for the project evaluation criterion number for the project number. 10. A two-creative event scoring system based on a two-type fuzzy model and LLM, characterized in that, A device for executing the double-creation event scoring method based on the Type2 fuzzy model and the LLM according to any one of claims 1 to 9 comprises: A construction module constructs a double-creation event scoring index system to obtain item evaluation criteria; A first calculation module calculates the importance of experts through language variables represented by intuitionistic fuzzy numbers, obtains the weight of each expert in the decision-making group, calculates the comprehensive weight of the item evaluation criteria and the average score of the items corresponding to the item evaluation criteria according to the weight of each expert; A second calculation module constructs an interval Type2 fuzzy number fuzzy decision matrix according to the comprehensive weight of the item evaluation criteria and the average score of the items corresponding to the item evaluation criteria, and calculates the positive prospect value and negative prospect value of each item according to the interval Type2 fuzzy number fuzzy decision matrix; A weight fusion module calculates the Type2-Fuzzy model weight of each item evaluation criterion and the large language model integrated weight of each item evaluation criterion, fuses the Type2-Fuzzy model weight and the large language model integrated weight of the item evaluation criteria, and obtains the fusion weight of each item evaluation criterion; A third calculation module calculates the cumulative prospect weight of each item evaluation criterion according to the fusion weight of each item evaluation criterion; A comprehensive calculation module calculates the comprehensive prospect value of each item according to the cumulative prospect weight of each item evaluation criterion, the positive prospect value and the negative prospect value of each item, and sorts the items according to the comprehensive prospect value of each item.