Intelligent enterprise evaluation method, system and equipment based on large language model and medium

Abstract

The invention discloses an intelligent enterprise evaluation method, system and device based on a large language model and a medium, and mainly relates to the technical field of enterprise evaluation. The method is used for solving the problems that an existing scheme lacks a method for deeply fusing structured evaluation data and a large language model, lacks a multi-dimensional index correlation analysis technology based on the large language model and lacks an intelligent evaluation result interpretation and suggestion generation system. Comprising the following steps: acquiring multi-dimensional index analysis prompt data, and inputting the multi-dimensional index analysis prompt data, industrial standard ranges of indexes, influence factors and calculation methods into a large language model to obtain analysis results of the indexes in structured format data; when the analysis result relates to an abnormal result, taking the abnormal analysis result as filling data of a preset abnormal analysis prompt template, and obtaining an abnormal analysis prompt word; and inputting the exception analysis prompt word and the exception result into a large language model to obtain an exception analysis suggestion.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of enterprise evaluation, in particular to an intelligent enterprise evaluation method, system, device and medium based on a large language model. BACKGROUND

[0002] At present, the enterprise evaluation system mainly relies on traditional data analysis methods. By collecting index data of each dimension of the enterprise (such as profitability, asset quality, debt safety, business growth, industrial layout, etc.), a preset weight and scoring standard are used to calculate the score of the enterprise in each evaluation dimension. These evaluation results are usually presented in the form of tables, charts, etc., containing a large amount of digitized index data, weights and score information. Evaluation experts need to manually analyze the strengths and weaknesses of the enterprise based on these data and propose improvement suggestions.

[0003] However, the existing data analysis method lacks a method for deeply integrating structured evaluation data with a large language model. The existing solution has not effectively solved the problem of how to convert structured index data of enterprise evaluation into a form that can be understood and analyzed by a large language model. There is a lack of multi-dimensional index correlation analysis technology based on a large language model. The existing technology is difficult to automatically find the implicit correlation between cross-dimensional indicators in enterprise evaluation, such as the mutual influence between performance evaluation and management evaluation indicators. There is a lack of an intelligent evaluation result interpretation and suggestion generation system. The existing technology mainly provides numerical results and basic analysis, and lacks the ability to convert evaluation results into targeted and operable improvement suggestions. SUMMARY

[0004] The application provides an intelligent enterprise evaluation method, system, device and medium based on a large language model to solve the problems of lack of a method for deeply integrating structured evaluation data with a large language model, lack of multi-dimensional index correlation analysis technology based on a large language model, and lack of an intelligent evaluation result interpretation and suggestion generation system.

[0005] In a first aspect, the application provides an intelligent enterprise evaluation method based on a large language model, the method comprising: The process involves cleaning the enterprise evaluation form data to obtain cleaned form data; extracting structured data from the cleaned form data based on preset retention types; mapping the indicator data in the structured data to a preset unified dimension range; obtaining supplementary explanatory information and adding contextual information to the structured data; using the contextual information to determine the specific enterprise evaluation domain knowledge database and obtaining the industry standard range, influencing factors, and calculation methods for each indicator in the structured data; obtaining multi-dimensional indicator analysis prompt data; inputting the multi-dimensional indicator analysis prompt data, the industry standard range, influencing factors, and calculation methods of the indicators into a large language model to obtain the analysis results for each indicator in the structured data; when the analysis results involve abnormal results, using the abnormal analysis results as input data for a preset abnormal analysis prompt template to obtain abnormal analysis prompt words; inputting the abnormal analysis prompt words and abnormal results into the large language model to obtain abnormal analysis suggestions.

[0006] In one implementation of this application, the enterprise evaluation form data is cleaned to obtain cleaned form data, specifically including: Read the missing values ​​from the enterprise evaluation form data and determine the proportion of missing values; When the proportion is lower than the preset minimum value, delete the data corresponding to the missing values; When the proportion is not lower than the preset minimum value and is lower than the preset maximum value, determine the type of missing value; When the missing value is time series data, the average value of all data within the preset time window corresponding to the missing value is used as the supplementary data. When the missing value is non-time series data, the previous collected value of the data corresponding to the missing value is used as the supplementary data. At the same time, the weight A of the previous collected value is obtained, and the weight of the current supplementary data is adjusted to A / 2. When the ratio is not lower than the preset maximum value, an error is reported and the subsequent process is stopped. Based on 3 σ The principle is to determine whether the time-series data in the enterprise evaluation form is abnormal; Based on a pre-set normal dataset, determine whether the non-time-series data in the enterprise evaluation table is abnormal data; Remove outlier data, add supplementary data, and obtain the cleaned table data.

[0007] In one implementation of this application, structured data is obtained from the cleaned tabular data based on a preset retention type, specifically including: The preset retention type can be obtained through the preset interface; the initial preset retention type includes: evaluation dimension, indicator category, weight, indicator name, actual value, score and category score; The specific data corresponding to the preset reserved type is extracted from the table data after cleaning through a keyword advance algorithm to form the structured format data.

[0008] In an implementation manner of the present application, the supplementary explanation information is obtained, and the context information is supplemented in the structured format data, specifically including: The supplementary explanation information is obtained through a preset supplementary interface; The specific meanings corresponding to the industry background, the enterprise type, the evaluation time and the evaluation standard value are extracted from the supplementary explanation information by using a large language model; The specific meanings are added to the structured format data as the context information.

[0009] In an implementation manner of the present application, before the context information is used to determine the specific enterprise evaluation field knowledge database and obtain the industry standard range, the influence factor and the calculation method of each index in the structured format data, the method further includes: An enterprise total evaluation field knowledge database is obtained, and the enterprise total evaluation field knowledge database is split into a plurality of enterprise evaluation field knowledge databases based on the industry background, the enterprise type, the evaluation time and the evaluation standard value; wherein the enterprise evaluation field knowledge database contains the industry standard range, the influence factor and the calculation method of a plurality of indexes.

[0010] In an implementation manner of the present application, the multi-dimensional index analysis prompt data, the industry standard range, the influence factor and the calculation method of the index are input into the large language model to obtain the analysis result of each index in the structured format data, specifically including: The index and the analysis requirement for which the index analysis is performed this time are obtained through a preset interface; Whether the analysis requirement involves the industry standard range, the influence factor and the calculation method is detected by the large language model; When the industry standard range is involved, the corresponding analysis result is generated according to whether the first specific value of the index falls within the industry standard range; When the influence factor is involved, the first specific value of the index and the second specific value corresponding to the influence factor are obtained; The specific correlation between the first specific value and the second specific value is analyzed by using the large language model, and the specific correlation is taken as the analysis result; When the calculation method is involved, the first specific value is input into the corresponding calculation method, and the calculation result is taken as the analysis result; When the industry standard range, the influence factor and the calculation method are not involved, the multi-dimensional index analysis prompt data and the structured format data are input into the large language model to obtain the analysis result.

[0011] In a second aspect, the present application provides an intelligent enterprise evaluation system based on a large language model, the system comprising: The processing module is used to clean the enterprise evaluation form data to obtain cleaned form data; based on the preset retention type, it obtains structured format data from the cleaned form data; it maps the indicator data in the structured format data to a preset unified dimension range; and it obtains supplementary explanatory information to supplement contextual information in the structured format data. The acquisition module is used to utilize contextual information to determine the specific enterprise evaluation domain knowledge database, acquire the industry standard range, influencing factors, and calculation methods of each indicator in the structured data; acquire multi-dimensional indicator analysis prompt data, input the multi-dimensional indicator analysis prompt data, the industry standard range, influencing factors, and calculation methods of the indicators into the large language model, and obtain the analysis results of each indicator in the structured data; The module is used to obtain anomaly analysis prompts by using the anomaly analysis results as input data for a preset anomaly analysis prompt template when the analysis results involve anomalies. The anomaly analysis prompts and anomalies are then input into the large language model to obtain anomaly analysis suggestions.

[0012] In one implementation of this application, the processing module includes a cleaning unit. Used to read missing values ​​from enterprise evaluation form data and determine the proportion of missing values; When the proportion is lower than the preset minimum value, delete the data corresponding to the missing values; When the proportion is not lower than the preset minimum value and is lower than the preset maximum value, determine the type of missing value; When the missing value is time series data, the average value of all data within the preset time window corresponding to the missing value is used as the supplementary data. When the missing value is non-time series data, the previous collected value of the data corresponding to the missing value is used as the supplementary data. At the same time, the weight A of the previous collected value is obtained, and the weight of the current supplementary data is adjusted to A / 2. When the ratio is not lower than the preset maximum value, an error is reported and the subsequent process is stopped. Based on 3 σ The principle is to determine whether the time-series data in the enterprise evaluation form is abnormal; Based on a pre-set normal dataset, determine whether the non-time-series data in the enterprise evaluation table is abnormal data; Remove outlier data, add supplementary data, and obtain the cleaned table data.

[0013] Thirdly, this application provides an intelligent enterprise evaluation device based on a large language model, the device comprising: processor; and a memory having stored thereon executable code that, when executed, cause the processor to perform any one of the above described methods of intelligent enterprise evaluation based on large language model.

[0014] In a fourth aspect, the present application provides a non-volatile computer storage medium having stored thereon computer instructions that, when executed, implement any one of the above described methods of intelligent enterprise evaluation based on large language model.

[0015] From the above technical solutions, the present application has the following advantages: I. Deep collaboration between structured data and language model: Through the three-layer preprocessing mechanism of data cleaning, format unification and dimension mapping, the standardized interaction channel between structured data in the enterprise evaluation field and large language model is first constructed. This design breaks through the limitations of unstructured text analysis in traditional solutions, so that numerical data such as financial ratios and operation indicators can directly participate in model reasoning (such as industry standard range comparison, influence factor attribution, etc.), while retaining the semantic association of supplementary explanation information. This fusion processing of structured and unstructured data makes the analysis results output by the model not only rigorous in data, but also contain business interpretation, which improves the efficiency compared with traditional manual analysis.

[0016] II. Establish a dynamic multi-dimensional index correlation analysis system: Based on the preset multi-dimensional index analysis prompt template, the association reasoning ability of large language model is combined with the industry knowledge base (standard range / computation method). When the model identifies that a certain index deviates from the industry benchmark, it can automatically trigger the influence factor tracing mechanism (such as "whether the decrease in inventory turnover rate is related to the extension of accounts receivable period"), forming a dynamic causal network between indexes. This design solves the problem of isolated analysis of indexes in traditional evaluation systems, and is especially suitable for complex scenarios such as supply chain and financial health that require cross-dimensional diagnosis.

[0017] III. Generate executable abnormal diagnosis suggestion chain: Through the standardized conversion technology of abnormal analysis prompt word templates, the abnormal markers output by the model (such as "low liquidity ratio below warning value") are converted into decision suggestions containing repair priority and implementation path. The system will automatically associate remedial measures (such as optimizing short-term debt structure) based on the industry knowledge base, and generate step-by-step execution schemes (including calculating the expected improvement range, departments to be coordinated, etc.) through the large language model. This mechanism fills the gap of existing evaluation systems that "only find problems but not solve them", so that the response time for abnormal processing is shortened. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the present application, the drawings required to be used in the description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0019] Figure 1 is a flow chart of an intelligent enterprise evaluation method based on a large language model provided by an embodiment of the present application.

[0020] Figure 2 is a schematic diagram of the internal structure of an intelligent enterprise evaluation system based on a large language model provided by an embodiment of the present application.

[0021] Figure 3 is a schematic diagram of the internal structure of an intelligent enterprise evaluation device based on a large language model provided by an embodiment of the present application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort fall within the protection scope of the present application.

[0023] Those skilled in the art should understand that the embodiments described below are only preferred embodiments of the present disclosure, and do not mean that the present disclosure can only be implemented by the preferred embodiments. The preferred embodiments are only used to explain the technical principles of the present disclosure, and are not used to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those skilled in the art without creative effort still fall within the protection scope of the present disclosure.

[0024] It should also be noted that the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or apparatus including a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such a process, method, article, or apparatus. Without more limitations, the element defined by the statement "including a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus including the element.

[0025] The technical solutions provided by the embodiments of the present application will be described in detail below with reference to the drawings.

[0026] The embodiment provides a large language model-based intelligent enterprise evaluation method, as shown in Figure 1 The method provided by the embodiment of the application mainly comprises the following steps: Step 110, data cleaning is performed on the enterprise evaluation form data to obtain cleaned form data; based on a preset retention type, structured format data is obtained from the cleaned form data; index data in the structured format data is mapped to a preset unified dimension range; supplementary explanation information is obtained, and context information is supplemented in the structured format data.

[0027] In some embodiments, the enterprise evaluation form data is subjected to data cleaning to obtain cleaned form data, which can be specifically as follows: Read the missing values in the enterprise evaluation form data, and determine the proportion of the missing values; When the proportion is lower than a preset minimum value, the data corresponding to the missing values is deleted; When the proportion is not lower than the preset minimum value and is lower than a preset maximum value, the type of the missing values is determined; When the type of the missing values is time series data, the average value of all data in a preset time window corresponding to the missing values is taken as the supplementary data; When the type of the missing values is non-time series data, the last acquisition value of the data corresponding to the missing values is taken as the supplementary data, and the weight A of the last acquisition value is obtained, and the weight of the current supplementary data is adjusted to A / 2; When the proportion is not lower than the preset maximum value, error processing is performed, and the subsequent process is stopped; Based on the principle of 3 σ Determine whether the time series data in the enterprise evaluation form data is abnormal data; Based on a preset normal data set, determine whether the non-time series data in the enterprise evaluation form data is abnormal data; Remove the abnormal data, add the supplementary data, and obtain the cleaned form data.

[0028] The structured format data is obtained from the cleaned form data based on the preset retention type, which can be specifically as follows: The preset retention type is obtained through a preset interface; wherein, the initial preset retention type includes: evaluation dimension, index classification, weight, index name, actual value, score and classification score; Through a keyword advance algorithm, specific data corresponding to the preset retention type is extracted from the cleaned form data to form the structured format data.

[0029] The structured format data is obtained from the cleaned form data based on the preset retention type, which can be specifically as follows: Supplementary explanation information is obtained through a preset supplementary interface; The large language model is used to extract the industry background, enterprise type, evaluation time, and specific meanings corresponding to the evaluation standard value from the supplementary explanation information. The specific meanings are added to the structured format data as context information.

[0030] In step 120, the context information is used to determine a specific enterprise evaluation field knowledge database, and the industry standard range, influencing factors, and calculation method of each index in the structured format data are obtained.

[0031] Before the context information is used to determine the specific enterprise evaluation field knowledge database and obtain the industry standard range, influencing factors, and calculation method of each index in the structured format data, the method further includes: An enterprise general evaluation field knowledge database is obtained, and the enterprise general evaluation field knowledge database is split into several enterprise evaluation field knowledge databases based on the industry background, enterprise type, evaluation time, and evaluation standard value. The enterprise evaluation field knowledge database contains the industry standard range, influencing factors, and calculation method of several indexes.

[0032] In step 130, multi-dimensional index analysis prompt data is obtained, and the multi-dimensional index analysis prompt data, the industry standard range, influencing factors, and calculation method of the index are input into the large language model to obtain the analysis result of each index in the structured format data.

[0033] The multi-dimensional index analysis prompt data, the industry standard range, influencing factors, and calculation method of the index are input into the large language model to obtain the analysis result of each index in the structured format data, which can be specifically: The index and analysis requirement for this time of index analysis are obtained through a preset interface; The analysis requirement is detected by the large language model to determine whether it involves the industry standard range, influencing factors, and calculation method; When the industry standard range is involved, the corresponding analysis result is generated according to whether the first specific value of the index falls within the industry standard range; When the influencing factors are involved, the first specific value of the index and the second specific value corresponding to the influencing factors are obtained; The specific correlation between the first specific value and the second specific value is analyzed by the large language model, and the specific correlation is taken as the analysis result; When the calculation method is involved, the first specific value is input into the corresponding calculation method, and the calculation result is taken as the analysis result; When the industry standard range, influencing factors, and calculation method are not involved, the multi-dimensional index analysis prompt data and the structured format data are input into the large language model to obtain the analysis result.

[0034] As an example, the multi-dimensional indicator analysis prompt data can be specific to: Please analyze the following enterprise evaluation data based on the following enterprise evaluation data: Indicator level: Analyze the performance of each indicator, paying special attention to indicators with extremely high and extremely low scores. Classification level: Comprehensive evaluation of the overall performance of each indicator classification; Dimension level: Evaluate the overall situation of the two dimensions of 'performance evaluation' and'management evaluation'; Overall level: Based on all evaluation dimensions, make a comprehensive assessment of the overall situation of the enterprise.

[0035] Step 140, when the analysis result involves an abnormal result, the abnormal analysis result is filled in as the preset abnormal analysis prompt template to obtain an abnormal analysis prompt word; the abnormal analysis prompt word and the abnormal result are input into the large language model to obtain an abnormal analysis suggestion.

[0036] It should be noted that the preset abnormal analysis prompt template has already preset the filling position of the abnormal analysis result, and the abnormal analysis result can be directly brought into the preset abnormal analysis prompt template to obtain the abnormal analysis prompt word.

[0037] In addition, after obtaining the analysis result, the preset prompt data is directly input into the large language model, and the preset prompt data can be specific to: Based on the aforementioned enterprise evaluation analysis result, please generate the following content: 1. 3-5 core problems faced by the enterprise, ranked by severity; 2. Specific improvement suggestions for each core problem, including: Problem analysis and causes; Improvement measures and implementation path; Expected effect and verification method; 2-3 main advantages of the enterprise and further improvement suggestions. In addition, the analysis result and the suggestion can be presented in various forms to meet the needs of different scenarios: Text report generation: Use the large language model to generate a structured evaluation analysis report, including overall evaluation, indicator analysis, problem diagnosis, and improvement suggestions.

[0038] Visual content generation guidance: Provide guidance for data visualization, including key indicator dashboards, indicator correlation network diagrams, and problem-suggestion correspondence diagrams.

[0039] Key point summary generation: Extract the core points of the evaluation analysis to form a concise summary for quick understanding.

[0040] Interactive Q&A Support: Based on evaluation data and analysis results, provide interactive Q&A capabilities to answer users' specific questions about enterprise evaluations.

[0041] Multi-modal Content Generation Prompt Examples: Based on the analysis results of enterprise evaluations, please generate: 1. An evaluation analysis report of no more than 1500 words, including overall evaluation, core problem diagnosis, and improvement suggestions; 2. A 200-word summary of evaluation highlights; 3. Content guidance for the following visualizations: Key point interpretation of enterprise dimension score radar chart; Node design and relationship strength representation method of index correlation network graph; Organization structure of problem-suggestion correspondence graph.

[0042] Preferred solution: Construction of enterprise evaluation knowledge base based on knowledge graph; Evaluation knowledge extraction: Extract index correlation rules, problem diagnosis experience, and effective improvement measures from historical evaluation cases.

[0043] Knowledge graph construction: Construct the extracted knowledge into an enterprise evaluation knowledge graph, including index nodes, relationship types, problem-cause-solution path, etc.

[0044] Knowledge-enhanced analysis: Use the knowledge graph to enhance the analysis capabilities of large language models to provide more accurate index correlation analysis and problem diagnosis.

[0045] Continuous updating of knowledge graph: Continuously update and optimize the knowledge graph based on new evaluation cases and feedback.

[0046] In addition, the present application Figure 2 provides an intelligent enterprise evaluation system based on a large language model. As Figure 2 shown, the system provided by the present application mainly includes: The processing module 210 is used for data cleaning of enterprise evaluation table data to obtain cleaned table data; based on a pre-set retention type, structured format data is obtained from the cleaned table data; index data in the structured format data is mapped to a pre-set unified dimension range; supplementary explanation information is obtained and context information is supplemented in the structured format data.

[0047] The processing module 210 includes a cleaning unit, for reading missing values in enterprise evaluation table data, determining the proportion of missing values; When the proportion is lower than the pre-set minimum value, delete the data corresponding to the missing values; When the proportion is not lower than the preset minimum value and is lower than the preset maximum value, determine the type of missing value; When the missing value is time series data, the average value of all data within the preset time window corresponding to the missing value is used as the supplementary data. When the missing value is non-time series data, the previous collected value of the data corresponding to the missing value is used as the supplementary data. At the same time, the weight A of the previous collected value is obtained, and the weight of the current supplementary data is adjusted to A / 2. When the ratio is not lower than the preset maximum value, an error is reported and the subsequent process is stopped. Based on 3 σ The principle is to determine whether the time-series data in the enterprise evaluation form is abnormal; Based on a pre-set normal dataset, determine whether the non-time-series data in the enterprise evaluation table is abnormal data; Remove outlier data, add supplementary data, and obtain the cleaned table data.

[0048] The acquisition module 220 is used to utilize contextual information to determine the specific enterprise evaluation domain knowledge database, acquire the industry standard range, influencing factors, and calculation methods of each indicator in the structured data; acquire multi-dimensional indicator analysis prompt data, input the multi-dimensional indicator analysis prompt data, the industry standard range, influencing factors, and calculation methods of the indicators into the large language model, and obtain the analysis results of each indicator in the structured data.

[0049] The module 230 is used to obtain abnormal analysis prompts by using the abnormal analysis results as input data for a preset abnormal analysis prompt template when the analysis results involve abnormal results; and inputs the abnormal analysis prompts and abnormal results into the large language model to obtain abnormal analysis suggestions.

[0050] The above are method embodiments of this application. Based on the same inventive concept, this application also provides an intelligent enterprise evaluation device based on a large language model. Figure 3 As shown, the device includes: a processor; and a memory storing executable code thereon, which, when executed, causes the processor to perform an intelligent enterprise evaluation method based on a large language model as described in the above embodiments.

[0051] Specifically, the server side performs data cleaning on enterprise evaluation form data to obtain cleaned form data; obtains structured format data from the cleaned form data based on a preset retention type; maps index data in the structured format data to a preset unified dimension range; obtains supplementary description information and supplements context information in the structured format data; determines a specific enterprise evaluation field knowledge database using the context information, and obtains an industry standard range, an influence factor, and a calculation method of each index in the structured format data; obtains multi-dimensional index analysis prompt data, inputs the multi-dimensional index analysis prompt data, the industry standard range, the influence factor, and the calculation method of the index into a large language model, and obtains an analysis result of each index in the structured format data; when the analysis result involves an abnormal result, takes the abnormal analysis result as filling data of a preset abnormal analysis prompt template to obtain an abnormal analysis prompt word; inputs the abnormal analysis prompt word and the abnormal result into the large language model to obtain an abnormal analysis suggestion.

[0052] In addition, the embodiment of the present application further provides a non-volatile computer storage medium, which has executable instructions stored thereon, and the executable instructions, when executed, realize an intelligent enterprise evaluation method based on a large language model.

[0053] The above description of disclosed embodiments enables one of ordinary skill in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those of ordinary skill in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Accordingly, the application is not to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating intelligent enterprises based on a large language model, characterized in that, The method includes: The data from the enterprise evaluation forms is cleaned to obtain cleaned form data; based on the preset retention type, structured format data is obtained from the cleaned form data; the indicator data in the structured format data is mapped to a preset unified dimension range; supplementary explanatory information is obtained, and contextual information is added to the structured format data; By utilizing contextual information, we can identify specific knowledge databases for enterprise evaluation and obtain the industry standard range, influencing factors, and calculation methods for each indicator in the structured data. Obtain multi-dimensional indicator analysis prompts data, input the multi-dimensional indicator analysis prompts data, the industry standard range of the indicators, influencing factors, and calculation methods into the large language model to obtain the analysis results of each indicator in the structured data; When the analysis results involve abnormal results, the abnormal analysis results are used as input data for the preset abnormal analysis prompt template to obtain abnormal analysis prompt words; the abnormal analysis prompt words and abnormal results are input into the large language model to obtain abnormal analysis suggestions.

2. The intelligent enterprise evaluation method based on a large language model according to claim 1, characterized in that, The enterprise evaluation form data is cleaned to obtain the cleaned form data, specifically including: Read the missing values ​​from the enterprise evaluation form data and determine the proportion of missing values; When the proportion is lower than the preset minimum value, delete the data corresponding to the missing values; When the proportion is not lower than the preset minimum value and is lower than the preset maximum value, determine the type of missing value; When the missing value is time series data, the average value of all data within the preset time window corresponding to the missing value is used as the supplementary data. When the missing value is non-time series data, the previous collected value of the data corresponding to the missing value is used as the supplementary data. At the same time, the weight A of the previous collected value is obtained, and the weight of the current supplementary data is adjusted to A / 2. When the ratio is not lower than the preset maximum value, an error is reported and the subsequent process is stopped. Based on 3 σ The principle is to determine whether the time-series data in the enterprise evaluation form is abnormal; Based on a pre-set normal dataset, determine whether the non-time-series data in the enterprise evaluation table is abnormal data; Remove outlier data, add supplementary data, and obtain the cleaned table data.

3. The intelligent enterprise evaluation method based on a large language model according to claim 1, characterized in that, Based on preset retention types, structured data is obtained from the cleaned tabular data, specifically including: The preset retention type can be obtained through the preset interface; the initial preset retention type includes: evaluation dimension, indicator category, weight, indicator name, actual value, score and category score; The keyword advance algorithm extracts specific data corresponding to the preset retention type from the cleaned table data to form structured format data.

4. The intelligent enterprise evaluation method based on a large language model according to claim 1, characterized in that, Obtain supplementary information and add contextual information to the structured data, specifically including: Obtain supplementary information through the preset supplementary interface; Using a large language model, we extract the specific meanings of industry background, company type, evaluation time, and evaluation standard values ​​from the supplementary information. Add specific meanings as contextual information to structured data.

5. The intelligent enterprise evaluation method based on a large language model according to claim 4, characterized in that, Before utilizing contextual information to determine the specific enterprise evaluation domain knowledge database and obtain the industry standard range, influencing factors, and calculation methods for each indicator in the structured data, the method further includes: The overall enterprise evaluation domain knowledge database is obtained and then divided into several enterprise evaluation domain knowledge databases based on industry background, enterprise type, evaluation time, and evaluation standard values. Each enterprise evaluation domain knowledge database contains the industry standard range, influencing factors, and calculation methods for several indicators.

6. The intelligent enterprise evaluation method based on a large language model according to claim 1, characterized in that, By inputting multi-dimensional indicator analysis data, industry standard ranges of indicators, influencing factors, and calculation methods into a large language model, the analysis results of each indicator in the structured data are obtained, specifically including: The indicators and analysis requirements for this analysis can be obtained through the preset interface. By using a large language model, we can detect and analyze whether the requirements involve industry standards, influencing factors, and calculation methods. When industry standard ranges are involved, corresponding analysis results are generated based on whether the first specific value of the indicator falls within the industry standard range. When influencing factors are involved, obtain the first specific value of the indicator and the second specific value of the influencing factor; The specific relationship between the first and second specific values ​​is analyzed using a large language model, and the specific relationship is used as the analysis result. When calculation methods are involved, the first specific value is input into the corresponding calculation method, and the calculation result is used as the analysis result. When industry standards, influencing factors, and calculation methods are not involved, multi-dimensional indicator analysis prompts and structured format data are input into a large language model to obtain analysis results.

7. An intelligent enterprise evaluation system based on a large language model, characterized in that, The system includes: The processing module is used to clean the enterprise evaluation form data to obtain cleaned form data; based on the preset retention type, it obtains structured format data from the cleaned form data; it maps the indicator data in the structured format data to a preset unified dimension range; and it obtains supplementary explanatory information to supplement contextual information in the structured format data. The acquisition module is used to utilize contextual information to determine the specific enterprise evaluation domain knowledge database, acquire the industry standard range, influencing factors, and calculation methods of each indicator in the structured data; acquire multi-dimensional indicator analysis prompt data, input the multi-dimensional indicator analysis prompt data, the industry standard range, influencing factors, and calculation methods of the indicators into the large language model, and obtain the analysis results of each indicator in the structured data; The module is used to obtain anomaly analysis prompts by using the anomaly analysis results as input data for a preset anomaly analysis prompt template when the analysis results involve anomalies. The anomaly analysis prompts and anomalies are then input into the large language model to obtain anomaly analysis suggestions.

8. The intelligent enterprise evaluation system based on a large language model according to claim 7, characterized in that, The processing module includes a cleaning unit. Used to read missing values ​​from enterprise evaluation form data and determine the proportion of missing values; When the proportion is lower than the preset minimum value, delete the data corresponding to the missing values; When the proportion is not lower than the preset minimum value and is lower than the preset maximum value, determine the type of missing value; When the missing value is time series data, the average value of all data within the preset time window corresponding to the missing value is used as the supplementary data. When the missing value is non-time series data, the previous collected value of the data corresponding to the missing value is used as the supplementary data. At the same time, the weight A of the previous collected value is obtained, and the weight of the current supplementary data is adjusted to A / 2. When the ratio is not lower than the preset maximum value, an error is reported and the subsequent process is stopped. Based on 3 σ The principle is to determine whether the time-series data in the enterprise evaluation form is abnormal; Based on a pre-set normal dataset, determine whether the non-time-series data in the enterprise evaluation table is abnormal data; Remove outlier data, add supplementary data, and obtain the cleaned table data.

9. An intelligent enterprise evaluation device based on a large language model, characterized in that, The device includes: processor; And a memory having executable code stored thereon, which, when executed, causes the processor to perform an intelligent enterprise evaluation method based on a large language model as described in any one of claims 1-6.

10. A non-volatile computer storage medium, characterized in that, It stores computer instructions, which, when executed, implement an intelligent enterprise evaluation method based on a large language model as described in any one of claims 1-6.

Images