Enterprise rating method and device based on ESG index and computer equipment

Abstract

The invention relates to an enterprise rating method and device based on an ESG index and computer equipment. The method comprises the following steps: acquiring enterprise information of a target enterprise, and dividing the enterprise information into a reference data set and a real data set; according to each ESG index dimension, performing conjoint analysis on the reference data set and the real data set to obtain a risk coefficient corresponding to the target enterprise; obtaining a first rating threshold value in a current enterprise rating environment, and determining a risk interval corresponding to the target enterprise according to a numerical value difference between the risk coefficient and the first rating threshold value; and obtaining a second rating threshold value matched with the risk interval corresponding to the target enterprise, and determining the risk level corresponding to the target enterprise according to the numerical value difference between the risk coefficient and the second rating threshold value. By adopting the method, the layering and refinement of the enterprise rating result can be ensured.

Description

Technical Field

[0001] This application relates to the field of enterprise rating technology, and in particular to an enterprise rating method, apparatus and computer equipment based on ESG indicators. Background Technology

[0002] In the field of corporate rating technology, risk assessment of enterprises is conducted based on ESG indicators such as environment, society, and governance, thereby achieving quantitative analysis of the overall performance of enterprises.

[0003] The relevant assessment methods generate enterprise rating results by directly comparing the assessment results of ESG indicators. However, such methods have the disadvantage of lacking detailed assessment, which makes it difficult for enterprise rating results to accurately reflect the true status of enterprises at different risk levels. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, and computer-readable storage medium for enterprise rating based on ESG indicators to address the aforementioned technical issues. This method enables a systematic transformation from raw enterprise information to the final risk level through layer-by-layer mapping, ensuring the hierarchical and refined nature of enterprise rating results.

[0005] Firstly, this application provides a corporate rating method based on ESG indicators, including: Obtain the target company's enterprise information and divide the enterprise information into a reference dataset and a real dataset; Based on each ESG indicator dimension, the reference dataset and the real dataset are jointly analyzed to obtain the risk coefficient corresponding to the target enterprise; Obtain the first rating threshold under the current enterprise rating environment, and determine the risk range corresponding to the target enterprise based on the numerical difference between the risk coefficient and the first rating threshold; Obtain the second rating threshold that matches the risk range corresponding to the target company, and determine the risk level corresponding to the target company based on the numerical difference between the risk coefficient and the second rating threshold.

[0006] Secondly, this application also provides a corporate rating device based on ESG indicators, comprising: The acquisition module is used to acquire enterprise information of the target enterprise and divide the enterprise information into a reference dataset and a real dataset. The joint analysis module is used to perform joint analysis on the reference dataset and the real dataset according to various ESG indicator dimensions to obtain the risk coefficient corresponding to the target enterprise. The first assessment module is used to obtain the first rating threshold under the current enterprise rating environment, and determine the risk range corresponding to the target enterprise based on the numerical difference between the risk coefficient and the first rating threshold. The second assessment module is used to obtain the second rating threshold that matches the risk range corresponding to the target enterprise, and to determine the risk level corresponding to the target enterprise based on the numerical difference between the risk coefficient and the second rating threshold.

[0007] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the above steps.

[0008] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the above steps.

[0009] The aforementioned enterprise rating method, apparatus, computer equipment, and computer-readable storage medium based on ESG indicators first classify the enterprise information of the target enterprise and construct a reference dataset and a real dataset, thus providing a basis for comparison in subsequent difference analysis. Second, through joint analysis of the reference dataset and the real dataset across various ESG indicator dimensions, a risk coefficient capable of quantifying the overall risk level of the enterprise is obtained. Next, the risk range corresponding to the target enterprise is calculated based on the difference between the risk coefficient and the first rating threshold, thus transforming the risk performance of the target enterprise from a numerical form to a range form. Finally, the risk level corresponding to the target enterprise is calculated based on the difference between the risk coefficient and the second rating threshold, thus further refining the risk performance of the target enterprise from a range form to a grade form. Based on this, a systematic transformation from raw enterprise information to the final risk level is achieved through layer-by-layer mapping, ensuring the hierarchical and refined nature of the enterprise rating results. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating a corporate rating method based on ESG indicators in one embodiment; Figure 2 This is a system architecture diagram of a computer system based on a large AI model in one embodiment; Figure 3This is a functional flowchart of a computer system based on a large AI model in one embodiment; Figure 4 This is a structural block diagram of an enterprise rating device based on ESG indicators in one embodiment. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0013] In one embodiment, such as Figure 1 As shown, an enterprise rating method based on ESG indicators is provided. This embodiment illustrates the application of this method to a server. It is understood that this method can also be applied to terminals, and can also be applied to systems including terminals and servers, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps S101 to S104.

[0014] Step S101: Obtain the enterprise information of the target enterprise and divide the enterprise information into a reference dataset and a real dataset.

[0015] Among them, the target enterprise refers to the enterprise to be subject to risk rating, and the enterprise information refers to the multi-dimensional data formed by the target enterprise in the process of operation, management and development.

[0016] The reference dataset represents a set of comparative information compiled based on industry standards, historical levels, or benchmark rules; the real dataset represents a set of actual information about the target company in its current or recent state, used to reflect the target company's true operating performance and as input for joint analysis with the reference dataset.

[0017] For example, the target company's corporate information includes corporate environmental datasets, corporate social datasets, corporate financial datasets, and corporate governance datasets. Optionally, the corporate environmental dataset represents a collection of data reflecting the target company's relevant situation in the environmental dimension, covering indicators of the target company in terms of energy consumption, emissions, and environmental management, derived from the target company's own disclosed environmental reports and externally available environmental change data records.

[0018] Optionally, the corporate social dataset represents a collection of data used to reflect the performance of a target company in terms of social responsibility. The data covers indicators such as employee rights, community contributions, and social welfare, and is derived from publicly available third-party social responsibility reports and self-survey data generated within the company.

[0019] Optionally, the corporate financial dataset refers to a set of data used to reflect the relevant situation of the target company in the financial dimension. The content covers indicators of the target company in terms of financial performance, capital operation and debt repayment ability, and it comes from financial data and financial information disclosed by various public platforms.

[0020] Optionally, the corporate governance dataset represents a collection of data reflecting the organizational structure and management mechanisms of a target company in terms of governance. The data covers indicators such as equity structure, board composition, and internal control, and is derived from the company's annual reports and publicly disclosed information from stock exchanges.

[0021] For example, after acquiring enterprise information at various levels, the information is preprocessed, including data cleaning, noise reduction, and dimensionless processing. The preprocessed enterprise information is then divided into a reference dataset and a real dataset. The reference dataset serves as a basic reference, while the real dataset retains the actual performance of the target enterprise. Furthermore, after the division, a corresponding index relationship is established between the reference dataset and the real dataset, transforming the enterprise information from a previously scattered data state into a dual dataset structure. This provides a clear data foundation for subsequent joint analysis and risk calculation.

[0022] Step S102: Based on each ESG indicator dimension, perform joint analysis on the reference dataset and the real dataset to obtain the risk coefficient corresponding to the target enterprise.

[0023] Among them, the ESG indicator dimension represents the evaluation dimension that divides the target enterprise from three dimensions: environment, society, and governance, and is used to ensure the multi-dimensional quantification of the enterprise's sustainable development capabilities.

[0024] The risk coefficient is a comprehensive value calculated by comparing the differences between the reference dataset and the real dataset in various ESG indicator dimensions. It is used to quantify the overall risk level of the target enterprise.

[0025] For example, within the ESG indicator dimensions such as environment, society, and governance, the corresponding content of the same indicator in the same ESG indicator dimension is matched item by item between the reference dataset and the real dataset. This involves matching the reference value corresponding to the reference dataset with the real value corresponding to the real dataset for the same indicator and then performing differentiation processing. This yields the differentiation results at the corresponding indicator level, such as differences or ratios, thus reflecting the degree of deviation between the reference level and the actual level at a single indicator level. Subsequently, the differentiation results corresponding to each indicator within the same ESG indicator dimension need to be integrated to form the overall differentiation result for that ESG indicator dimension. This transforms complex indicator data into a single, calculable result, facilitating comparison between different ESG indicator dimensions. Next, the overall differentiation results for the environmental, social, and governance dimensions are further integrated, transforming the originally scattered multidimensional difference values ​​into a unified measurement indicator.

[0026] Finally, by comprehensively calculating the overall differences in multiple ESG indicator dimensions, a single value that reflects the overall deviation of the target company is obtained, which is the risk coefficient. The generation process of the risk coefficient reflects a layer-by-layer processing logic from raw data to difference comparison, then to dimension integration, and finally to result refinement, thereby transforming complex and scattered multidimensional corporate information into a quantitative indicator that can be directly used for rating judgment.

[0027] Optionally, the corporate environmental dataset reflects information such as emissions, resource utilization, and environmental management, serving as relevant information for comparing the content of the reference dataset with the real dataset in the environmental dimension; the corporate social dataset reflects information such as employee rights, social responsibility, and public welfare activities, serving as relevant information for comparing the content of the reference dataset with the real dataset in the social dimension; and the corporate governance dataset reflects information such as equity structure, board composition, and internal control mechanisms, serving as relevant information for comparing the content of the reference dataset with the real dataset in the governance dimension.

[0028] In addition, corporate financial datasets provide supplementary support for analysis in the three dimensions of environment, society, and governance. For example, financial information can reflect a target company's ability to invest in the environment, its expenditure on social responsibility, and its financial compliance in governance. Therefore, corporate financial datasets can be used as auxiliary indicators and, after decomposition, embedded into the analysis process of the three dimensions of environment, society, and governance.

[0029] Step S103: Obtain the first rating threshold under the current enterprise rating environment, and determine the risk range corresponding to the target enterprise based on the numerical difference between the risk coefficient and the first rating threshold.

[0030] The current corporate rating environment refers to the background conditions corresponding to the rating system constructed in a specific period or scenario, which is used to provide a unified reference standard for threshold setting and risk assessment.

[0031] The first rating threshold represents the benchmark value set under the current corporate rating environment for the initial risk classification, in order to distinguish the risk coefficients for the first time and map them to the corresponding risk range; the risk range represents the range of values ​​divided according to the first rating threshold, used to characterize the company's relative position in the overall risk level.

[0032] For example, the first step is to obtain the first rating threshold under the current corporate rating environment. This first rating threshold is not an isolated value, but is set based on the background conditions of the current rating system. It reflects the benchmark requirements for risk classification under the current corporate rating environment. The current corporate rating environment can be composed of current policy guidance, industry general standards, and the overall market acceptance of risk. Therefore, the first rating threshold is essentially to provide a calculable and unified reference point for risk classification under a given external environment.

[0033] Furthermore, after obtaining the first rating threshold, based on the numerical difference between the first rating threshold and the risk coefficient, and according to the range within which the numerical difference falls, the target company is assigned to a specific risk interval. For example, if the risk coefficient is less than the first rating threshold, the target company is classified into a low-risk interval, indicating that the target company is in a risk interval with a relatively low overall risk level, i.e., the target company is classified as a low-risk company. If the risk coefficient is greater than or equal to the first rating threshold, the target company is classified into a high-risk interval, indicating that the target company is in a risk interval with a relatively high overall risk level, i.e., the target company is classified as a high-risk company. Based on this, through the continuous operation of obtaining the first rating threshold, calculating the difference, and determining the risk interval, the risk coefficient transitions from a single numerical state to an interval positioning state, thereby realizing the transformation from quantitative results to structured expression.

[0034] Optionally, for example, during periods of stricter policy guidance, the first rating threshold is set lower to strengthen the constraints on enterprises in environmental and social dimensions, resulting in enterprises with slightly higher risk coefficients being classified as high-risk enterprises. Similarly, during periods of generally stringent industry standards, the first rating threshold is set lower to accommodate the general operating levels of the industry, again resulting in enterprises with slightly higher risk coefficients being classified as high-risk enterprises. Furthermore, during periods of declining market acceptance, the first rating threshold is set even lower to accommodate a smaller range of risk coefficient fluctuations, resulting in enterprises with slightly higher risk coefficients being classified as high-risk enterprises. Therefore, changes in policy, industry, and market conditions can directly affect the risk classification results by adjusting the first rating threshold in different scenarios.

[0035] Step S104: Obtain the second rating threshold that matches the risk range corresponding to the target company, and determine the risk level corresponding to the target company based on the numerical difference between the risk coefficient and the second rating threshold.

[0036] The second rating threshold represents a benchmark value set further within a specific risk range, used to further differentiate the risk coefficients within the specific risk range and map them to the corresponding risk level; the risk level represents the range of values ​​defined based on the second rating threshold, used to characterize the clear level of the enterprise in terms of overall risk.

[0037] For example, based on determining the risk range corresponding to the target company, a second rating threshold corresponding to that risk range is further determined. That is, the risk range obtained through the first rating threshold is still at a relatively coarse level. To make the risk assessment more specific, more refined reference points need to be introduced within the risk range. Furthermore, after obtaining the second rating threshold, based on the numerical difference between the second rating threshold and the risk coefficient, and according to the range within which the numerical difference falls, the target company is assigned to a specific risk level. For example, as shown in Table (1), if the risk coefficient Q of the target enterprise is less than the first rating threshold Y, then the target enterprise is a low-risk enterprise in the low-risk range. Therefore, the second rating threshold R corresponding to the low-risk range is selected, and the second rating threshold R is less than the first rating threshold Y. Specifically: if the risk coefficient Q is less than the second rating threshold R, then the target enterprise is classified as a risk-free level (i.e., level N) to indicate that the target enterprise is in the lowest risk level overall, that is, the target enterprise is a high-quality enterprise; if the risk coefficient Q is equal to the second rating threshold R, then the target enterprise is classified as a low-risk level (i.e., level L) to indicate that the target enterprise is in a relatively low risk level overall, that is, the target enterprise is a good-quality enterprise; if the risk coefficient Q is greater than the second rating threshold R, then the target enterprise is classified as a medium-risk level (i.e., level M) to indicate that the target enterprise is in a moderate risk level overall, that is, the target enterprise is an ordinary-quality enterprise.

[0038] For example, as shown in Table (1), if the risk coefficient Q of the target enterprise is greater than or equal to the first rating threshold Y, then the target enterprise is a high-risk enterprise in the high-risk range. Therefore, the second rating threshold S corresponding to the high-risk range is selected, and the second rating threshold S is greater than the first rating threshold Y. Specifically: if the risk coefficient Q is less than the second rating threshold S, then the target enterprise is classified as a high-risk level (i.e., level H) to indicate that the target enterprise is in a high-risk level overall, that is, the target enterprise is regarded as a low-quality enterprise; if the risk coefficient Q is greater than or equal to the second rating threshold S, then the target enterprise is classified as the highest-risk level (i.e., level S) to indicate that the target enterprise is in the highest-risk level overall, that is, the target enterprise is regarded as a low-quality enterprise.

[0039] Table (1) is as follows: Based on this, in this process, the second rating threshold plays a role in further stratifying the risk range, so that risk assessment is no longer limited to macro-level division, but achieves more precise level positioning. Therefore, by continuously obtaining the second rating threshold, calculating the difference and determining the risk level, the step-by-step mapping process of the target company's risk performance from numerical value to range and then to level is completely closed, and finally a risk level that can be directly used for the company rating conclusion is obtained.

[0040] Furthermore, both the first and second rating thresholds are derived from benchmark data provided by authoritative professional institutions. This ensures that the enterprise rating process does not rely on a single company's self-definition or subjective judgment, but rather uses industry-recognized standards as a reference basis. Specifically, on the one hand, the first rating threshold is used to define whether an enterprise is in a high-risk or low-risk range. Its value is generated by authoritative professional institutions based on a comprehensive analysis of extensive historical data, general industry trends, and policy guidance, thus reflecting the market's overall risk tolerance. On the other hand, the second rating threshold is used to further subdivide risk levels within the high-risk or low-risk range. Its value is also set based on the data distribution of authoritative professional institutions, ensuring consistency and comparability in risk level classification. Therefore, this rating method based on unified thresholds also ensures the reasonableness of horizontal comparisons of rating results between different companies, making the final enterprise rating report more authoritative, transparent, and of practical reference value.

[0041] The aforementioned enterprise rating method based on ESG indicators firstly classifies the enterprise information of the target enterprise and constructs a reference dataset and a real dataset, thus providing a basis for comparison in subsequent difference analysis. Secondly, based on the joint analysis of the reference dataset and the real dataset in each ESG indicator dimension, a risk coefficient that can quantify the overall risk level of the enterprise is obtained. Next, the risk range corresponding to the target enterprise is calculated based on the difference between the risk coefficient and the first rating threshold, thus transforming the risk performance of the target enterprise from a numerical form to a range form. Finally, the risk level corresponding to the target enterprise is calculated based on the difference between the risk coefficient and the second rating threshold, thus further refining the risk performance of the target enterprise from a range form to a level form. Based on this, a systematic transformation from raw enterprise information to the final risk level is achieved through layer-by-layer mapping, ensuring the hierarchical and refined nature of the enterprise rating results.

[0042] In an exemplary embodiment, the reference dataset and the real dataset are jointly analyzed according to each ESG indicator dimension to obtain the risk coefficient corresponding to the target enterprise, including steps S201 to S203.

[0043] Step S201: Obtain the target difference between the reference dataset and the real dataset in each ESG indicator dimension.

[0044] Step S202: Obtain the first dynamic weight, nonlinear exponent, and standardized coefficient corresponding to each ESG indicator dimension.

[0045] Step S203: Based on the first dynamic weight, nonlinear index, and standardized coefficient corresponding to each ESG indicator dimension, and combined with the preset correction coefficient, perform weighted nonlinear transformation and standardization on the target difference of the corresponding ESG indicator dimension to obtain the risk coefficient corresponding to the target enterprise.

[0046] The target difference represents the numerical difference calculated between the reference dataset and the real dataset for the same ESG indicator dimension, and is used to reflect the degree of deviation between the target company's actual performance and the benchmark level in that ESG indicator dimension.

[0047] The first dynamic weight represents the variable weight coefficient assigned to each ESG indicator dimension, which is used to adjust the relative influence of different ESG indicator dimensions in the risk coefficient calculation process. For example, when the target difference of the environmental dimension is more volatile in the overall difference or has a stronger correlation with other dimensions, its corresponding first dynamic weight will be increased accordingly.

[0048] The nonlinear exponent refers to the exponential factor used when performing a nonlinear transformation on the target difference. It is used to amplify or compress the numerical response of the target difference. For example, when the target difference is small but its difference needs to be highlighted, the influence can be amplified by a higher exponent.

[0049] The standardization coefficient represents an adjustment factor that maps the calculation results of different dimensions to a uniform scale. It is used to eliminate the differences in the magnitude of numerical values ​​of each dimension. For example, it can unify the emission data of the environmental dimension and the employee indicators of the social dimension into the same range.

[0050] The correction factor represents an additional adjustment parameter introduced on the basis of weighted and standardized calculations. It is used to fine-tune the overall result in combination with specific external requirements or rating preferences. For example, when the industry as a whole tightens regulation, the final risk coefficient is adjusted upward.

[0051] For example, the risk coefficient Q corresponding to the target company can be calculated by referring to formula (1): (1) Wherein, FDR represents the target difference corresponding to the enterprise financial dataset, EDR represents the target difference corresponding to the enterprise environmental dataset, SDR represents the target difference corresponding to the enterprise social dataset, and GDR represents the target difference corresponding to the enterprise governance dataset; This represents the first dynamic weight corresponding to FDR. This represents the first dynamic weight corresponding to EDR. This represents the first dynamic weight corresponding to SDR. This represents the first dynamic weight corresponding to GDR.

[0052] in, This represents the nonlinear exponent corresponding to FDR. This represents the nonlinear exponent corresponding to EDR. This represents the nonlinear exponent corresponding to SDR. 1 represents the nonlinear exponent corresponding to GDR; NF represents the normalized coefficient corresponding to FDR; NE represents the normalized coefficient corresponding to EDR; NS represents the normalized coefficient corresponding to SDR; NG represents the normalized coefficient corresponding to GDR; A represents the correction coefficient.

[0053] For example, in equation (1), in the environmental dimension, the enterprise environmental dataset is used as a reference dataset to compare the relevant information between the reference dataset and the real dataset in the environmental dimension. This yields the difference results of the reference values ​​of the reference dataset and the real values ​​of the real dataset in the environmental dimension for each indicator. Then, the difference results of each indicator are integrated to obtain the target difference value corresponding to the environmental dimension, i.e., EDR. Next, based on... , NE performs a weighted nonlinear transformation and standardization on EDR to obtain the standardized difference result corresponding to the environmental dimension, i.e. .

[0054] In the social dimension, the enterprise social dataset is used as a reference dataset, and its content is compared with the real dataset in the social dimension to obtain relevant information. This yields the difference results of the reference values ​​of the reference dataset and the real values ​​of the real dataset in various indicators of the social dimension. These difference results are then integrated to obtain the target difference value (SDR) corresponding to the social dimension. Next, based on... , NS performs a weighted nonlinear transformation and standardization on EDR to obtain the standardized difference result corresponding to the environmental dimension, i.e. .

[0055] In the governance dimension, the enterprise governance dataset is used as a reference dataset, and relevant information is compared with the real dataset in the governance dimension. This yields the difference results of the reference values ​​of the reference dataset and the real values ​​of the real dataset in various indicators of the governance dimension. These difference results are then integrated to obtain the target difference value corresponding to the governance dimension, i.e., GDR. Next, based on... , NG performs a weighted nonlinear transformation and standardization on GDR to obtain the standardized difference result corresponding to the environmental dimension, i.e. .

[0056] In the combined dimension of environmental, social, and governance dimensions, the enterprise financial dataset is used as a reference dataset, and its content is compared with the real dataset within this combined dimension. This yields the differences between the reference values ​​of the reference dataset and the real values ​​of the real dataset across various indicators within the combined dimension. These differences are then integrated to obtain the target difference (FDR) for the combined dimension. Next, based on... , NF performs a weighted nonlinear transformation and standardization on FDR to obtain the standardized difference result corresponding to the combined dimension, i.e. .

[0057] Among them, the combined dimension refers to the composite dimension formed by introducing corporate financial datasets for comprehensive comparison on the basis of environmental, social and governance dimensions. It is used to reflect the comprehensive relationship between corporate financial datasets and the three aspects of environmental, social and governance factors. For example, analyzing emission differences alone in the environmental dimension may not be sufficient to reflect the company's capital investment capacity. However, by introducing financial data into the combined dimension, emission differences can be compared with capital expenditures in a unified manner, thereby obtaining FDR with financial constraints.

[0058] In this embodiment, firstly, by comparing the reference dataset and the real dataset across each ESG indicator dimension, target differences are extracted and aggregated to form quantifiable deviation inputs. Secondly, the first dynamic weight, nonlinear exponent, and standardization coefficient corresponding to each ESG indicator dimension are determined to allocate the degree of influence of the dimension, characterize the numerical response, and suppress dimensional differences. Thirdly, the target differences are subjected to weighted nonlinear transformation and standardization based on the above parameters, and the standardized difference results of each ESG indicator dimension are integrated with the correction coefficient to obtain a single risk coefficient. Based on this, a unified quantitative indicator that can be directly used for threshold comparison is formed.

[0059] In an exemplary embodiment, the first dynamic weight, nonlinear index, and standardized coefficient corresponding to each ESG indicator dimension are obtained, including steps S301 to S303.

[0060] Step S301: Based on the correlation between the target differences corresponding to each ESG indicator dimension at the statistical feature level, obtain the first dynamic weight corresponding to each ESG indicator dimension.

[0061] For example, statistical feature analysis is introduced for the target differences corresponding to each ESG indicator dimension. This means that during the processing, the relative magnitudes of the target differences for each dimension need to be examined, and their proportional distribution structure further identified. For instance, if the target difference for one dimension is significantly larger than those for other dimensions, it indicates that this dimension exhibits a more significant deviation in the current enterprise rating process, and its corresponding weight should be increased accordingly. Conversely, if the target difference for one dimension is small, it indicates that its contribution to the current enterprise rating process is limited, and its role in overall risk assessment is relatively weak; therefore, its corresponding weight should be decreased accordingly. Based on this, through this relative comparison method, a proportional distribution structure among the target differences can be established, and this proportional distribution structure can be used as the basis for weight allocation.

[0062] Ultimately, each ESG indicator dimension will generate a first dynamic weight corresponding to the corresponding target difference. These first dynamic weights constitute a set of constrained parameters, enabling each ESG indicator dimension to exert different influences according to the degree of difference in subsequent risk coefficient calculations.

[0063] Optionally, for the statistical characteristic analysis of the target differences corresponding to each ESG indicator dimension, the proportion of the target difference to the total of all target differences can be calculated to reflect the relative influence of the corresponding dimension in the whole. Furthermore, the relative deviation of the target difference from all target differences can be measured by comparing the target difference with the normalized value of the maximum or minimum target difference, thus reflecting the relative influence of the corresponding dimension in the whole. Additionally, the relative deviation of the target difference from the overall median level can be measured by comparing the target difference with the overall central value of all target differences, thus reflecting the relative influence of the corresponding dimension in the whole.

[0064] Optionally, the first dynamic weight in equation (1) For example, let's explain one way to calculate the first dynamic weight: (2) In equation (2), This represents the relative proportion of the target difference (EDR) of the enterprise environment dataset among the target differences of all indicator dimensions, thus characterizing the relative weight of the environment dimension in the overall target difference.

[0065] Furthermore, This represents the ratio of the target difference (EDR) corresponding to the enterprise environmental dataset to the maximum target difference in other indicator dimensions, characterizing the correlation between the environmental dimension and other indicator dimensions. For example, if the difference between EDR and the maximum target difference is small, the ratio is close to 1, indicating a strong correlation, and its relative weight is not weakened; if EDR is much smaller than the maximum target difference, the ratio is much smaller than 1, indicating a weak correlation, and its relative weight decreases accordingly. Step S302: Obtain the difference sequence in the historical enterprise rating environment, and based on the changing trend reflected by each ESG indicator dimension in the corresponding difference sequence, obtain the nonlinear index corresponding to each ESG indicator dimension.

[0066] Among them, the historical corporate rating environment refers to the overall background conditions on which the corporate rating system formed in the past period is based, such as the specific performance of policy standards, industry benchmarks, and market acceptance in different historical time periods.

[0067] Among them, the difference sequence in the historical enterprise rating environment represents a continuous record in chronological order of the differences between the reference dataset and the real dataset for multiple enterprises in various ESG indicator dimensions during the historical rating process, which is used to reflect the changing trend of the differences in various ESG indicator dimensions over time.

[0068] For example, firstly, the difference sequence in the historical enterprise rating environment is obtained, that is, in the past enterprise rating process, the difference results between the reference dataset and the real dataset for similar enterprises or the same enterprise in different time periods are arranged in chronological order to form a continuous original difference sequence; then, these original difference sequences are organized according to ESG indicator dimensions, so that each ESG indicator dimension corresponds to a set of difference sequences.

[0069] Furthermore, by observing the growth, decline, or curvilinear changes of these difference sequences over time, we can analyze the trends in these difference sequences. For example, if the difference in a certain dimension shows a slow increase followed by a sharp rise over time, it indicates that the degree of difference in that dimension exhibits a non-linear enhancement characteristic under dynamic conditions; if the change in the difference in a certain dimension remains close to linear, it indicates that the degree of difference in that dimension remains stable over time. Based on this, through the above trend analysis process, a non-linear index can be assigned to each ESG indicator dimension. This non-linear index is used to adjust the response of the target difference in the overall calculation of the corresponding dimension during subsequent processing. That is, the larger the non-linear index, the more significant the amplification effect of the target difference in the final risk calculation; the smaller the non-linear index, the closer the target difference remains to its original proportion.

[0070] Ultimately, each ESG indicator dimension obtains a nonlinear index corresponding to the changing trend of its difference sequence, so as to transform the single target difference into an input that is more in line with the actual evolution characteristics, thereby making the risk coefficient not only reflect static comparison, but also include the influence of dynamic trends.

[0071] Optionally, trend analysis of the difference sequence needs to be based on continuous recording of multiple indicators over time: First, the differences of all indicators under a certain dimension are arranged in chronological order to form a complete difference sequence corresponding to that dimension; then, by observing the overall trend, it is determined whether the difference sequence is gradually rising, gradually falling, or remaining stable, thus obtaining the basic direction of change; in this basic direction of change, it is observed whether there are stage-specific turning points in the difference sequence, such as small changes in the early stages followed by rapid increases in the later stages, to indicate that the evolution of the difference sequence over time has non-linear characteristics; in addition, the consistency between indicators needs to be examined. If most indicators show changes in the same direction, the trend can be confirmed as clear; if the directions of the indicators differ significantly, the differences corresponding to these indicators need to be weighted and integrated to generate a unified trend expression; finally, by comparing the rate and magnitude of change in different time periods, it is determined whether the dimension has accelerating or stable change characteristics. After this series of processing, the difference sequence is transformed into trend information that can be used to assign non-linear exponents, so that each dimension can obtain parameter inputs that match its evolution characteristics.

[0072] Specifically, if the difference in a certain dimension exhibits a near-linear change over time, meaning the numerical variation is relatively uniform across different time periods, then a nonlinear exponent close to linear is assigned to that dimension to ensure that subsequent calculations maintain a response close to the original proportion. If the difference in a certain dimension shows a significant phased acceleration over time, such as a slow change in the early stages followed by a rapid increase in the difference later, then a larger nonlinear exponent is assigned to that dimension to enhance the mapping effect of the target difference in the high range during subsequent processing. If the rate of change of the difference in a certain dimension gradually slows down throughout the overall process, then a smaller nonlinear exponent is assigned to that dimension to suppress the expansion of the target difference in the high range during subsequent processing.

[0073] Optionally, the nonlinear exponent in equation (1) For example, the difference series in the corresponding historical corporate rating environment is as follows: This explains one way to calculate the nonlinear exponent: (3) In equation (3), This represents the cumulative value corresponding to the change in the difference between two adjacent time points in the difference sequence corresponding to EDR, used to reflect the degree of data fluctuation. For example, if the difference sequence changes steadily, the cumulative value is close to 0; if the difference sequence changes drastically, the cumulative value is relatively large.

[0074] Furthermore, This represents the cumulative sum of the difference sequences corresponding to EDR. Based on this, This is equivalent to the proportion of volatility to the total volume. A smaller proportion indicates relatively stable volatility, and it represents a non-linear exponent. A value close to 1; if the proportion is large, it indicates that fluctuations account for a significant portion of the overall value, and the non-linear exponent is strong. The corresponding increase.

[0075] Optionally, for the preprocessed reference dataset and the real dataset, dimensionality reduction techniques can be used to compress the high-dimensional data in the data to a lower-dimensional representation, removing redundant information while maintaining the overall differences. Subsequently, core indicators that can reflect the characteristics of different dimensions are extracted from the dimensionality reduction results, and nonlinear indices are generated accordingly. These nonlinear indices express the characteristics of the dimensionality-reduced data in specific numerical form.

[0076] Step S303: Obtain enterprise attribute information from the historical enterprise rating environment, and cluster the enterprise attribute information according to industry, size, and region to obtain clustering results, so as to determine the standardization coefficients corresponding to each ESG indicator dimension.

[0077] Among them, the enterprise attribute information in the historical enterprise rating environment refers to the set of data related to the characteristics of the enterprise itself collected in the historical rating process. It is used to distinguish enterprise differences in the clustering process, such as the industry category, enterprise size level and the region where it is located.

[0078] For example, the process begins by acquiring enterprise attribute information from historical enterprise rating environments. This involves collecting basic attribute data on enterprises during previous rating processes, such as their industry category, size, and geographical location. This information reflects the differences in enterprise performance under various external and internal conditions. Subsequently, based on this enterprise attribute information, enterprises with similar basic attributes are grouped into the same group, thus forming relatively consistent attribute characteristics within each group, while maintaining differences between groups. Based on this, the common patterns of different enterprise groups across various ESG indicator dimensions can be extracted using this grouping method, and a set of standardized coefficients can be determined for the group to which the current target enterprise belongs.

[0079] Based on this, the relationship between the original target difference and enterprise attribute information is reasonably introduced into the calculation logic, so that the final standardized difference result no longer depends on the absolute performance of a single enterprise, but is transformed into a numerical result that can be compared horizontally between different enterprises after standardization correction based on historical grouping characteristics.

[0080] Optionally, within the same group, companies maintain a high degree of consistency in terms of industry type, size, or regional environment, while different groups exhibit significant differences in these aspects. In this way, all companies are divided into several groups, each with unified basic attributes. Next, within each group, the differences in various ESG indicator dimensions of the corresponding company group are analyzed to extract common patterns within that group. These patterns can be reflected in the numerical range or distribution characteristics of the differences, serving as a unified reference at the group level. Finally, once a target company is identified as belonging to a certain group, corresponding standardized coefficients are matched to its various ESG dimensions based on the patterns extracted for that group, ensuring that the target company's target differences are normalized within the corresponding company group.

[0081] Alternatively, taking the standardized coefficient NE in equation (1) as an example, we can explain one way to calculate the standardized coefficient: (4) In equation (4), This represents the average difference level of the environmental dimension in cluster C where the target company is located. Based on this, NE represents the ratio between the current cluster mean and the target difference of the target enterprise under the environmental dimension. If the target difference of the target enterprise is close to the current cluster mean, then NE is close to 1; if the target difference of the target enterprise deviates from the current cluster mean, then NE is too large or too small. Optionally, after generating each target difference, a cross-evaluation model can be used to validate it. Specifically, the target difference is repeatedly evaluated under different data splitting conditions, and the results are tested for consistency and stability. Then, by comprehensively summarizing these results, the standardization coefficient corresponding to the target difference under standardization processing is finally determined, thus obtaining a set of validated comparable values.

[0082] Furthermore, the aforementioned parameters, such as the first dynamic weight, nonlinear index, and standardized coefficient, are all dynamically determined based on the corresponding sample data. This avoids the rigidity and limitations of static parameter settings. In other words, statically set parameters often lack adaptability to different companies or different periods, easily leading to deviations between results and reality. Dynamically set parameters, on the other hand, can be adjusted according to the real-time distribution characteristics of the sample data, making the parameters more consistent with the actual situation in the current rating environment. Therefore, by dynamically setting parameters, they can be updated as the data sample changes, thereby improving the flexibility and accuracy of company rating results and ensuring reasonable applicability across different industries or rating environments.

[0083] Furthermore, the first and second rating thresholds are provided by authoritative professional institutions as unified benchmark data to ensure the objectivity and authority of the enterprise rating results. The risk coefficient, on the other hand, is dynamically adjusted based on relevant sample data using parameters such as the first dynamic weight, nonlinear index, and standardized coefficients to reflect the specific performance of the enterprise in the current rating environment. The two are complementary in function; the former provides a stable external standard, while the latter reflects flexible internal differences. When the risk coefficient is compared with the rating thresholds, it can showcase the individual circumstances of the enterprise under a unified standard. Based on this, this approach avoids the problem of rigid parameter settings while ensuring that the enterprise rating results have industry recognition and scientific rationality. Therefore, the combination of the two does not conflict but rather enhances the accuracy and authority of the enterprise rating process.

[0084] In this embodiment, firstly, based on the statistical correlation between the target differences corresponding to each ESG indicator dimension, a first dynamic weight is assigned to each dimension to reflect the importance of the target difference. Secondly, based on the trend analysis of the difference sequence in the historical enterprise rating environment, a nonlinear index is set for each dimension to adjust the response to changes in the target difference. Thirdly, based on the clustering of enterprise attribute information in the historical enterprise rating environment to extract the enterprise group pattern, a standardized coefficient is determined for each dimension to ensure the comparability of the target difference. Based on this, through the combined effect of the first dynamic weight, the nonlinear index, and the standardized coefficient, multi-level correction and transformation of the target difference are achieved, establishing a reasonable and unified parameter system for the subsequent calculation of risk coefficients.

[0085] In an exemplary embodiment, the target difference between the reference dataset and the real dataset in each ESG indicator dimension is obtained, including steps S401 to S403.

[0086] Step S401: Obtain the sub-target differences between the reference dataset and the real dataset in the sub-index dimensions of each ESG index dimension.

[0087] Step S402: Based on the correlation between the sub-objective differences corresponding to each sub-indicator dimension in the same ESG indicator dimension at the statistical feature level, the second dynamic weights corresponding to the sub-indicator dimensions of each ESG indicator dimension are obtained respectively.

[0088] Step S403: Based on the second dynamic weights corresponding to the sub-indicator dimensions of each ESG indicator dimension, perform weighted averaging on the sub-target differences of each sub-indicator dimension of the corresponding ESG indicator dimension to obtain the target difference of the corresponding ESG indicator dimension.

[0089] The sub-target difference represents the numerical difference calculated between the reference dataset and the real dataset for the same sub-index dimension within the same ESG indicator dimension. It is used to reflect the degree of deviation between the target company's actual performance in that sub-index dimension and the benchmark level.

[0090] The second dynamic weight represents the variable weight coefficient assigned to each sub-indicator dimension, which is used to adjust the relative influence of different sub-indicator dimensions in the risk coefficient calculation process. For example, when the sub-target difference of a certain sub-indicator dimension in the environmental dimension has greater volatility in the overall difference or stronger correlation with other sub-indicator dimensions, its corresponding second dynamic weight will be increased accordingly.

[0091] For example, in the combined dimensions of environmental, social, and governance dimensions, the calculation method for the target difference FDR corresponding to the enterprise financial dataset can refer to equation (5): (5) In equation (5), the sub-target differences of the sub-indicator dimensions corresponding to the combined dimensions include: the difference in corporate debt ratio (FZL), the difference in corporate cash flow ratio (LDL), the difference in corporate net profit (JLR), the difference in corporate dividend payout ratio (GXL), and the difference in corporate profitability (SYL). This represents the second dynamic weight corresponding to FZL. This represents the second dynamic weight corresponding to LDL. This represents the second dynamic weight corresponding to JLR. This represents the second dynamic weight corresponding to GXL. This represents the second dynamic weight corresponding to SYL. In equation (5), the sub-target differences of the sub-indicator dimensions corresponding to the combined dimension are weighted and averaged according to the corresponding second dynamic weights to obtain the target difference FDR on the combined dimension.

[0092] For example, in the environmental dimension, the target difference EDR corresponding to the enterprise environmental dataset can be calculated by referring to equation (6): (6) In equation (6), the sub-target differences of the sub-indicator dimensions corresponding to the environmental dimension include: the difference in enterprise emissions (PFL) and the difference in the enterprise renewable energy use ratio (SBL). This represents the second dynamic weight corresponding to PFL. This represents the second dynamic weight corresponding to SBL. In equation (6), the sub-target differences of the sub-indicator dimensions corresponding to the environmental dimension are weighted and averaged according to the corresponding second dynamic weights to obtain the target difference EDR on the environmental dimension.

[0093] For example, in the social dimension, the target difference (SDR) of the enterprise social dataset can be calculated by referring to equation (7): (7) In equation (7), the sub-objective differences of the sub-indicator dimensions corresponding to the social dimension include: the difference in corporate community project investment (SJE), the difference in corporate community project participation (SCY), the difference in corporate medical insurance coverage (BXL), and the difference in corporate employee satisfaction survey results (MYD). This represents the second dynamic weight corresponding to SJE. This represents the second dynamic weight corresponding to SCY. This represents the second dynamic weight corresponding to BXL. This represents the second dynamic weight corresponding to MYD. In equation (7), the sub-target differences of the sub-indicator dimensions corresponding to the social dimension are weighted and averaged according to the corresponding second dynamic weights to obtain the target difference SDR on the social dimension.

[0094] For example, in the governance dimension, the target difference GDR corresponding to the corporate governance dataset can be calculated by referring to equation (8): (8) In equation (8), the sub-objective differences of the sub-indicator dimensions corresponding to the governance dimension include: the difference in enterprise labor dispute cases ZYL, the difference in average board tenure WSD, and the difference in enterprise procedural soundness JQX. Among them, This represents the second dynamic weight corresponding to ZYL. This represents the second dynamic weight corresponding to WSD. This represents the second dynamic weight corresponding to JQX. In Equation (8), the sub-target differences of the sub-indicator dimensions corresponding to the governance dimension are weighted and averaged according to the corresponding second dynamic weights to obtain the target difference GDR on the governance dimension.

[0095] Optionally, the sub-objective differences of multiple sub-indicator dimensions under a certain ESG indicator dimension are first collected as a set of data. These sub-objective differences reflect the deviations at different emphase levels within that ESG indicator dimension. Then, the numerical relationships between these sub-objective differences are analyzed to determine their relative roles in the overall difference composition of that ESG indicator dimension. For example, if the sub-objective difference of a certain sub-indicator dimension has a larger value in the overall difference composition, it indicates that its contribution to that ESG indicator dimension is more significant, and the corresponding second dynamic weight should be set higher; conversely, if the sub-objective difference is small, the second dynamic weight should be set lower to maintain the rationality of the overall proportion.

[0096] Simultaneously, it is necessary to examine the interrelationships between the differences of various sub-objectives. For example, when multiple sub-objective differences change in the same direction, relatively balanced weights can be assigned. However, when a particular sub-objective difference exhibits a significant difference, its weight needs to be increased to reflect its uniqueness. Based on this, through this statistically characteristic-based analysis process, each sub-indicator dimension ultimately obtains a second dynamic weight corresponding to its sub-objective difference. These second dynamic weights will be combined with the corresponding sub-objective differences in subsequent weighted processing to derive the overall target difference of the corresponding ESG indicator dimension.

[0097] Optionally, the second dynamic weight in equation (8) For example, let's explain one way to calculate the second dynamic weight: (9) In equation (9), This represents the relative proportion of the difference in enterprise labor dispute cases ZYL among the sub-target differences of all sub-indicator dimensions, thus characterizing the relative weight of the corresponding sub-indicator dimension in the overall sub-target difference.

[0098] Furthermore, The ratio of the difference in enterprise labor dispute cases ZYL to the largest sub-target difference among other sub-indicator dimensions is used to characterize the correlation between the corresponding sub-indicator dimension and other sub-indicator dimensions. For example, if the difference between ZYL and the largest sub-target difference is not large, the ratio between the two is close to 1, the correlation is strong, and its relative weight is not weakened. If ZYL is much smaller than the largest sub-target difference, the ratio between the two is much smaller than 1, the correlation is weak, and its relative weight decreases accordingly.

[0099] In this embodiment, firstly, the sub-target differences of the sub-indicator dimensions are extracted and organized to obtain the difference input that can be refined to the internal level of the dimension. Secondly, based on the statistical characteristic relationship between the sub-target differences within the same ESG indicator dimension, corresponding second dynamic weights are assigned to achieve a reasonable proportion of sub-target differences within the dimension. Thirdly, based on the combined calculation of the sub-target differences and the second dynamic weights, the target difference of the corresponding dimension is derived, thereby transforming the multi-dimensional differences within the dimension into a single result. Based on this, through hierarchical processing and weighted integration of sub-indicator differences, a step-by-step transformation from internal dimension details to the overall dimension result is achieved, laying a complete data structure for subsequent comprehensive risk calculation.

[0100] In an exemplary embodiment, the sub-target differences between the reference dataset and the real dataset on the sub-index dimensions of each ESG index dimension are obtained, including steps S501 to S502.

[0101] Step S501: Obtain the target data corresponding to the same sub-indicator dimension of the same ESG indicator dimension for the reference dataset and the real dataset.

[0102] For example, the reference dataset and the real dataset are compared at the same dimensional level, that is, the environmental, social, or governance dimensions are further subdivided into various sub-indicator dimensions to achieve a fine-grained data comparison process. Specifically, after determining the sub-indicator dimension to be compared, on the one hand, the target data corresponding to the sub-indicator dimension, i.e., the reference value, is located in the reference dataset to ensure that it can represent the benchmark level; on the other hand, the target data corresponding to the sub-indicator dimension, i.e., the real value, is found in the real dataset to ensure that it can reflect the true situation of the target enterprise. In addition, since the data sources are different, the formats, units of measurement, or recording methods may differ. Therefore, normalization processing is also required during acquisition to ensure that the two types of data maintain consistency, which is convenient for subsequent difference calculation. Based on this, the data that were originally scattered in different data sources are aligned at the dimensional and sub-dimensional levels, resulting in a set of target data pairs that can be directly used for difference calculation, forming the prerequisite for subsequent calculation of sub-target differences.

[0103] Step S502: Based on the percentage difference between the target data corresponding to the reference dataset and the target data corresponding to the real dataset, obtain the sub-target difference in the corresponding sub-index dimension of the corresponding ESG index dimension.

[0104] For example, firstly, the difference between target data pairs within the same indicator dimension is calculated to obtain the corresponding degree of deviation. Secondly, to ensure comparability, the difference is expressed as a percentage. That is, the target data in the reference dataset is used as a benchmark, and the target data in the real dataset is compared with this benchmark to calculate the relative deviation ratio. Thus, this method not only reflects the absolute difference of the target company in this sub-indicator dimension but also eliminates the influence of different indicator units, allowing the result to be uniformly expressed as a standardized percentage difference. Finally, the percentage value of the difference corresponding to this sub-indicator dimension is taken as the sub-target difference for that sub-indicator dimension. Based on this, the target data pairs are transformed into highly comparable difference results through percentage calculation, so that the differences in the sub-indicator dimension are systematically quantified, forming directly usable sub-target differences.

[0105] In this embodiment, the target data corresponding to the same sub-indicator dimension of the same ESG indicator dimension are obtained from the reference dataset and the real dataset respectively. Then, the sub-target difference is calculated based on the percentage difference between the target data pairs of the same sub-indicator dimension, so that the difference results are unified into standardized values ​​with strong comparability. Based on this, a quantitative expression of the target enterprise at the subdivided dimension level is realized.

[0106] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0107] The enterprise rating method based on ESG indicators in any of the above embodiments can be implemented using a computer system based on a pre-set AI model. Figure 2The system architecture diagram of the computer system is shown. The system architecture of the computer system consists of multiple layers from bottom to top. The bottom layer is the AI ​​large model layer, which includes various types of AI large models such as vector models, language generation models, and visual recognition models, to provide intelligent analysis and generation capabilities.

[0108] A RAG service layer is further set up, which includes functional modules such as data parsing and data retrieval, to enable parsing and semantic retrieval of multi-source data.

[0109] A data acquisition layer is further included, which supports multiple data access methods, including file upload, online form reporting, link parsing, automatic network collection, API interface, and APP-based collection. It also involves issue identification and automatic classification, which is used to automatically identify the topic of the issue in the text or document during the data input stage and classify and archive it, thereby providing structured input for subsequent indicator calculation and analysis.

[0110] A functional service layer is further set up, which includes user and permission function modules such as user management, permission management, department management and approval, as well as indicator and report function modules such as indicator management, indicator calculation and analysis, report generation and analysis and report template management, to meet the system's management and data processing needs.

[0111] A further interface layer is set up to connect with external systems through SDK or API; the top layer is the application layer, which includes a web browser, application services and mobile APP, to provide interactive interfaces and service access points.

[0112] Overall, the system architecture of this computer system forms a complete closed loop driven by a large underlying AI model, through data collection and analysis, then to functional service processing and interface calls, and finally displayed at the application layer, realizing full-process support for the collection, analysis, rating and display of ESG indicators.

[0113] Specifically, Figure 3 The functional flowchart of the aforementioned computer system is shown. First, the system collects raw data related to the enterprise from various channels using a data collector. This raw data then undergoes sequential processing operations, including data cleaning, filtering, identification, and archiving, to ensure the integrity, consistency, and standardization of the data entering the system. The processed data is then stored in a database, forming a structured data resource that can be accessed later.

[0114] Subsequently, the system retrieves the required data from the database using a data retrieval tool and maps it to multiple preset ESG indicators. Then, the processed ESG indicators are input into the AI ​​big model, which performs comprehensive analysis and reasoning on the ESG indicators based on preset algorithms and logic, and finally outputs a qualified enterprise rating result. Based on the enterprise rating result, the system automatically generates an ESG report to present the enterprise's comprehensive performance in environmental, social and governance dimensions through text, tables or visualization.

[0115] On the one hand, at the data acquisition and integration level of the computer system, the system supports multiple data input methods, including manual entry by users through Excel templates, uploading original enterprise documents (such as Word, Excel, and PPT), and uploading multimodal content such as images and videos. Furthermore, the system uses AI large-scale models to transform unstructured information into structured data and automatically categorizes it to the corresponding indicator positions under semantic mapping, requiring only minimal confirmation and correction from users. Simultaneously, the system can connect in real-time with internal enterprise business systems such as production, sales, and human resources through API interfaces, directly collecting multi-dimensional enterprise data to ensure the comprehensiveness and timeliness of internal information. In addition, the system utilizes web crawling technology to extract industry standards, policies, regulations, and market environment information from external websites as supplementary external data sources. Based on this, user input data, internal enterprise data, and external industry data are organically integrated in this way, forming a unified structured database that provides complete and diversified input support for subsequent enterprise rating based on ESG indicators.

[0116] On the other hand, regarding data processing and analysis within the computer system, firstly, in terms of efficient data integration capabilities, the system can quickly integrate heterogeneous data from multiple sources within an enterprise, including financial, operational, and environmental data, breaking down information silos and improving data utilization, thereby providing comprehensive support for enterprise strategy formulation and risk management. Secondly, in terms of in-depth data analysis and insight, by applying AI-powered large-scale models to conduct multi-dimensional analysis of the integrated data, it can uncover potential patterns and trends. For example, it can analyze the impact of employees' work environment on productivity, and, combined with salary and benefits data, provide a reference for human resource optimization and management decisions. Thirdly, in terms of dynamic data monitoring and early warning, when key indicators show abnormalities or trend changes, it can monitor and generate early warning prompts in real time, helping enterprises respond quickly to potential risks. For example, when an enterprise's carbon emission indicators approach the policy-mandated threshold, the system will automatically prompt adjustment measures to assist the enterprise in taking effective measures. Based on this, the combination of these three aspects can provide enterprises with deeper and more timely intelligent support in the ESG rating process.

[0117] On the other hand, regarding the report generation and display level of the computer system, the system first has an automated report generation function, which can automatically generate reports of different standards and formats according to the needs of enterprises, reducing manual intervention, lowering compilation costs, and improving overall efficiency. As a result, enterprises can quickly generate concise or detailed reports to meet diverse scenarios of internal management or external disclosure. Simultaneously, the system also has report version management and traceability functions, recording and archiving the report generation process and all historical versions to ensure that the modification process is traceable and the data source is verifiable. Therefore, when external disclosure or auditing is required, enterprises can use this function to provide complete evidence of report generation to prove the authenticity and compliance of the data. Based on this, a unified approach to report management efficiency, flexibility, and transparency is achieved, providing high-transparency and high-credibility output support for enterprise rating.

[0118] In this computer system, each calculation formula and each rating threshold is integrated into an independent module according to a preset format to achieve clear, structured processing. Specifically, the calculation formula involving risk coefficients is encapsulated as a parameter calculation module, used to receive and process data inputs from different dimensions and output the corresponding risk coefficients; the definition method involving the first and second rating thresholds is encapsulated as a threshold comparison module, used to compare the risk coefficients with fixed benchmarks one by one, thereby generating the judgment results of risk range and risk level. All of the above modules are uniformly scheduled and invoked by the AI ​​big data model. When enterprise data is input into the computer system, the AI ​​big data model can automatically match the required parameter calculation module and threshold comparison module according to task requirements, and complete the parameter calculation and threshold comparison in a predetermined order. Based on this, through this modular approach, the invocation of calculation formulas and rating thresholds no longer relies on manual intervention, but is automatically completed by the computer system, thereby ensuring that different enterprises follow a unified processing logic in the rating process, while possessing flexible expansion and rapid adaptation capabilities, making the final enterprise rating results both objective and standardized.

[0119] Furthermore, in the aforementioned computer system, blockchain technology is introduced to... Figure 3Managing data at any stage as shown ensures data security and traceability. Specifically, firstly, after data is acquired by the data collector and processed through cleaning, filtering, identification, and archiving, the system generates a corresponding data digest based on the processed data and synchronously writes this data digest to the blockchain, thereby achieving tamper-proof management of data during storage and retrieval. Because the data digest is unique, when verifying whether a file has been tampered with, it is only necessary to compare the newly generated data digest of the current file with the data digest stored in the blockchain. If they match, it indicates that the data remains unchanged; if they do not match, it means the file has been modified. Based on this, when entering the data retrieval and data access stages, the data digest recorded on the blockchain can serve as the basis for data integrity verification, avoiding data distortion caused by human modification or system anomalies.

[0120] Furthermore, during the indicator generation and AI large-scale model analysis phases, the blockchain can simultaneously record the source and processing of various indicator data, ensuring transparent data support for the final ESG report. Based on this, when external regulators or third-party auditors require verification, companies can trace the entire data flow process through the blockchain, ensuring the reliability and compliance of the report results.

[0121] This demonstrates that the introduction of blockchain technology not only strengthens the trustworthy management of data in all aspects of the system, but also enhances the credibility and usability of the rating process.

[0122] Based on the same inventive concept, this application also provides an ESG-based enterprise rating apparatus for implementing the aforementioned ESG-based enterprise rating method. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations of one or more ESG-based enterprise rating apparatus embodiments provided below can be found in the limitations of the ESG-based enterprise rating method described above, and will not be repeated here.

[0123] In one exemplary embodiment, such as Figure 4 As shown, an enterprise rating device based on ESG indicators is provided, including: an acquisition module 101, a joint analysis module 102, a first evaluation module 103, and a second evaluation module 104, wherein: The acquisition module 101 is used to acquire enterprise information of the target enterprise and divide the enterprise information into a reference dataset and a real dataset. The joint analysis module 102 is used to perform joint analysis on the reference dataset and the real dataset according to various ESG indicator dimensions to obtain the risk coefficient corresponding to the target enterprise. The first assessment module 103 is used to obtain the first rating threshold under the current enterprise rating environment, and determine the risk range corresponding to the target enterprise based on the numerical difference between the risk coefficient and the first rating threshold. The second assessment module 104 is used to obtain the second rating threshold that matches the risk range corresponding to the target enterprise, and to determine the risk level corresponding to the target enterprise based on the numerical difference between the risk coefficient and the second rating threshold.

[0124] In an exemplary embodiment, the joint analysis module 102 is further configured to: obtain the target difference between the reference dataset and the real dataset in each ESG indicator dimension; obtain the first dynamic weight, nonlinear exponent and standardization coefficient corresponding to each ESG indicator dimension; and, based on the first dynamic weight, nonlinear exponent and standardization coefficient corresponding to each ESG indicator dimension, and in combination with a preset correction coefficient, perform weighted nonlinear transformation and standardization processing on the target difference of the corresponding ESG indicator dimension to obtain the risk coefficient corresponding to the target enterprise.

[0125] In an exemplary embodiment, the joint analysis module 102 is further configured to: obtain the first dynamic weights corresponding to each ESG indicator dimension based on the correlation between the target differences corresponding to each ESG indicator dimension at the statistical feature level; obtain the difference sequence in the historical enterprise rating environment, and obtain the nonlinear index corresponding to each ESG indicator dimension based on the changing trend reflected by each ESG indicator dimension in the corresponding difference sequence; obtain enterprise attribute information in the historical enterprise rating environment, and cluster the enterprise attribute information according to industry, size, and region to obtain clustering results, so as to determine the standardization coefficients corresponding to each ESG indicator dimension.

[0126] In an exemplary embodiment, the joint analysis module 102 is further configured to: obtain the sub-target differences between the reference dataset and the real dataset at each sub-index dimension of the ESG index dimension; obtain the second dynamic weights corresponding to each sub-index dimension of the ESG index dimension based on the correlation between the sub-target differences corresponding to each sub-index dimension in the same ESG index dimension at the statistical feature level; and perform weighted averaging on the sub-target differences at each sub-index dimension of the corresponding ESG index dimension based on the second dynamic weights corresponding to each sub-index dimension of the ESG index dimension to obtain the target difference at the corresponding ESG index dimension.

[0127] In an exemplary embodiment, the joint analysis module 102 is further configured to: obtain the target data corresponding to the same sub-index dimension of the same ESG index dimension for the reference dataset and the real dataset respectively; and obtain the sub-target difference in the corresponding sub-index dimension of the corresponding ESG index dimension based on the percentage difference between the target data corresponding to the reference dataset and the target data corresponding to the real dataset.

[0128] The modules in the aforementioned ESG-based enterprise rating device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0129] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above embodiments.

[0130] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above embodiments.

[0131] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0132] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0133] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A corporate rating method based on ESG indicators, characterized in that, The method includes: Obtain the target company's enterprise information and divide the enterprise information into a reference dataset and a real dataset; Based on each ESG indicator dimension, the reference dataset and the real dataset are jointly analyzed to obtain the risk coefficient corresponding to the target enterprise; Obtain the first rating threshold under the current enterprise rating environment, and determine the risk range corresponding to the target enterprise based on the numerical difference between the risk coefficient and the first rating threshold; Obtain the second rating threshold that matches the risk range corresponding to the target company, and determine the risk level corresponding to the target company based on the numerical difference between the risk coefficient and the second rating threshold.

2. The method according to claim 1, characterized in that, The step of jointly analyzing the reference dataset and the real dataset according to various ESG indicator dimensions to obtain the risk coefficient corresponding to the target enterprise includes: Obtain the target difference between the reference dataset and the real dataset in each ESG indicator dimension; Obtain the first dynamic weight, non-linear exponent, and standardized coefficient for each ESG indicator dimension; Based on the first dynamic weight, nonlinear index, and standardized coefficient corresponding to each ESG indicator dimension, and combined with the preset correction coefficient, the target difference of the corresponding ESG indicator dimension is subjected to weighted nonlinear transformation and standardization to obtain the risk coefficient corresponding to the target enterprise.

3. The method according to claim 2, characterized in that, The process of obtaining the first dynamic weight, nonlinear exponent, and standardized coefficient corresponding to each ESG indicator dimension includes: Based on the correlation between the target differences corresponding to each ESG indicator dimension at the statistical feature level, the first dynamic weight corresponding to each ESG indicator dimension is obtained. Obtain the difference series in the historical corporate rating environment, and based on the changing trends reflected in the corresponding difference series of each ESG indicator dimension, obtain the nonlinear index corresponding to each ESG indicator dimension. Obtain enterprise attribute information from historical enterprise rating environments, and cluster the enterprise attribute information according to industry, size, and region to obtain clustering results, so as to determine the standardization coefficients corresponding to each ESG indicator dimension.

4. The method according to claim 2, characterized in that, The step of obtaining the target difference between the reference dataset and the real dataset in each ESG indicator dimension includes: Obtain the sub-target difference between the reference dataset and the real dataset in each sub-index dimension of the ESG index; Based on the correlation at the statistical feature level between the sub-objective differences corresponding to each sub-indicator dimension in the same ESG indicator dimension, the second dynamic weights corresponding to the sub-indicator dimensions of each ESG indicator dimension are obtained respectively. Based on the second dynamic weights corresponding to the sub-indicator dimensions of each ESG indicator dimension, the sub-target differences of each sub-indicator dimension of the corresponding ESG indicator dimension are weighted and averaged to obtain the target difference of the corresponding ESG indicator dimension.

5. The method according to claim 4, characterized in that, The step of obtaining the sub-target differences between the reference dataset and the real dataset at each sub-index dimension of the ESG index dimension includes: Obtain the target data corresponding to the same sub-indicator dimension of the same ESG indicator dimension of the reference dataset and the real dataset respectively; Based on the percentage difference between the target data corresponding to the reference dataset and the target data corresponding to the real dataset, the sub-target difference in the corresponding sub-index dimension of the corresponding ESG index dimension is obtained.

6. A corporate rating device based on ESG indicators, characterized in that, The device includes: The acquisition module is used to acquire enterprise information of the target enterprise and divide the enterprise information into a reference dataset and a real dataset. The joint analysis module is used to perform joint analysis on the reference dataset and the real dataset according to various ESG indicator dimensions to obtain the risk coefficient corresponding to the target enterprise. The first assessment module is used to obtain the first rating threshold under the current enterprise rating environment, and determine the risk range corresponding to the target enterprise based on the numerical difference between the risk coefficient and the first rating threshold. The second assessment module is used to obtain the second rating threshold that matches the risk range corresponding to the target enterprise, and to determine the risk level corresponding to the target enterprise based on the numerical difference between the risk coefficient and the second rating threshold.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images