ESG-based enterprise comprehensive financial performance assessment system

Abstract

The present application relates to the technical field of enterprise financial performance assessment, and in particular to an ESG-based enterprise comprehensive financial performance assessment system. The system involves: receiving various financial and non-financial data of an enterprise, performing ESG factor identification and KPI identification by means of an intelligent agent, and storing identification results in an information pool; parsing ESG-related indicators and KPIs to generate an evaluation indicator set; calling the intelligent agent to perform comprehensive evaluation, and acquiring an evaluation result and a risk level; segmenting the data into data units of different granularities, and generating contextual prompt information corresponding to the granularities; performing performance evaluation on the basis of the contextual prompt information, and acquiring and storing evaluation information; and fusing the evaluation information of different granularities to generate a final comprehensive financial performance assessment result. The present application can improve the comprehensiveness and accuracy of performance evaluation, and exhibits an outstanding performance, especially in terms of performing historical data evaluation.

Description

An ESG-based comprehensive financial performance evaluation system for enterprises Technical Field

[0001] The present invention belongs to the technical field of enterprise performance management, and specifically is an ESG-based enterprise comprehensive financial performance evaluation system. Background Art

[0002] With the growing popularity of corporate social responsibility (ESG) concepts, incorporating ESG factors into corporate performance appraisals has become a crucial tool for improving a company's overall financial performance. However, existing corporate performance appraisal systems still have some shortcomings in incorporating and evaluating ESG factors, hindering the comprehensiveness and accuracy of these assessments.

[0003] After searching, a performance appraisal method, server, and performance appraisal system with publication number CN108133306B were disclosed on March 23, 2021. This patent provides a method for performance appraisal of review behavior based on user identification. The server receives information related to review behavior sent by the user terminal, collects and analyzes the review behavior data, and generates performance appraisal results. However, this technical solution mainly focuses on the data statistics and analysis of user review behavior, and does not involve the inclusion and evaluation of ESG factors, making it difficult to fully reflect the comprehensive financial performance of the enterprise. In addition, the system lacks strict control over the source and quality of data, which may lead to deviations in the appraisal results.

[0004] A search revealed a performance appraisal method and system based on a performance appraisal indicator system, with publication number CN117273549B, dated January 26, 2024. This patent provides a performance appraisal method that generates advanced performance appraisal plans from performance appraisal indicator system parameters. By acquiring appraisal indicator data, constructing a basic performance appraisal plan, analyzing primary performance indicators, and conducting primary performance appraisals, it improves the accuracy and effectiveness of performance appraisals. However, this technical solution primarily focuses on the data processing and analysis of performance indicators and does not involve the systematic incorporation and comprehensive evaluation of ESG factors, making it difficult to fully reflect a company's environmental, social, and governance performance. Furthermore, the system lacks flexible adaptability to the specific ESG requirements of different industries and companies, potentially leading to one-sided and limited appraisal results. Technical issues

[0005] The above issues demonstrate that existing corporate performance appraisal systems still have shortcomings in terms of ESG factor incorporation, data source control, comprehensive assessment, and industry adaptability. Therefore, the present invention provides an ESG-based comprehensive corporate financial performance appraisal system designed to fully incorporate ESG factors, strictly control data sources and quality, systematically assess a company's environmental, social, and governance performance, enhance the comprehensiveness and accuracy of performance appraisals, and meet the needs of companies for efficient and comprehensive performance appraisals. Technical Solutions

[0006] In order to solve the above technical problems, this application provides an ESG-based enterprise comprehensive financial performance assessment method, system and device.

[0007] The technical solution provided in this application is described below:

[0008] The first aspect of this application provides an ESG-based comprehensive financial performance assessment method for an enterprise, including:

[0009] Receive various financial and non-financial data of the enterprise, use the first agent to call at least one second agent to identify ESG factors in the data, obtain ESG-related indicators and their corresponding categories in the data, and store the identification results in an information pool;

[0010] Invoking at least one third agent to identify key performance indicators (KPIs) in the data, obtain corresponding weight information, and store the identification results of the KPIs and weight information in an information pool;

[0011] Analyze the ESG-related indicators and KPIs, and generate a set of evaluation indicators based on a predetermined ESG template library;

[0012] Invoking at least one fourth agent to perform a comprehensive evaluation on the set of evaluation indicators, obtaining an evaluation result and its corresponding risk level, and storing the evaluation result and risk level in an information pool;

[0013] Divide the data into data units of different granularities, including quarterly, annual, and historical data;

[0014] For each data unit, the system queries the information pool for ESG-related indicators, KPIs, assessment results, and risk levels, and integrates these to generate contextual prompts of the corresponding granularity.

[0015] Based on the context prompt information Prompt, call at least one agent to perform performance evaluation on the quarterly, annual and historical data units respectively, and obtain and store corresponding evaluation information;

[0016] The evaluation information of different granularities is integrated to generate the final comprehensive financial performance evaluation results.

[0017] Optionally, parsing the ESG-related indicators and KPIs and generating a set of evaluation indicators based on a predetermined ESG template library includes:

[0018] The first agent traverses all ESG-related indicators and KPIs and parses them to obtain all evaluation indicators and form an evaluation indicator set;

[0019] The first agent calls the fourth agent to perform a comprehensive evaluation on each evaluation indicator in the evaluation indicator set to obtain all evaluation results and their corresponding risk levels.

[0020] Optionally, the data is divided into data units of different granularities, including quarterly, annual, and historical data, including:

[0021] Dividing the data into quarterly data by the first agent;

[0022] For each data unit, the relevant ESG indicators, KPIs, assessment results, and risk levels are queried from the information pool, and integrated to generate contextual prompts of corresponding granularity, including:

[0023] For each quarterly data, the first agent queries the information pool for the ESG-related indicators, KPIs, assessment results, and risk levels involved;

[0024] Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate the first prompt information corresponding to the quarter.

[0025] Optionally, the data is divided into data units of different granularities, including quarterly, annual, and historical data, including:

[0026] Dividing the data into annual data by the first agent;

[0027] For each data unit, the relevant ESG indicators, KPIs, assessment results, and risk levels are queried from the information pool, and integrated to generate contextual prompts of corresponding granularity, including:

[0028] For each year's data, the first agent queries the information pool for the ESG-related indicators, KPIs, assessment results, and risk levels involved;

[0029] Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate the second prompt information corresponding to the year.

[0030] Optionally, the data is divided into data units of different granularities, including quarterly, annual, and historical data, including:

[0031] dividing the data into historical data by the first agent;

[0032] For each data unit, the relevant ESG indicators, KPIs, assessment results, and risk levels are queried from the information pool, and integrated to generate contextual prompts of corresponding granularity, including:

[0033] For each historical data, query the ESG-related indicators, KPIs, assessment results and risk levels involved from the information pool through the first agent;

[0034] Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate a third Prompt message corresponding to the historical data.

[0035] Optionally, the calling of at least one agent to perform performance evaluation on quarterly, annual, and historical data units based on the context prompt information Prompt, and obtaining and storing corresponding evaluation information includes:

[0036] Inputting the first prompt information integrated with the context prompt information into the fifth agent as additional input content through the first agent;

[0037] Through the fifth agent, a performance evaluation of the quarterly data is performed based on the Prompt information, and the first evaluation information obtained by the evaluation is returned to the information pool of the first agent.

[0038] Optionally, the calling of at least one agent to perform performance evaluation on quarterly, annual, and historical data units based on the context prompt information Prompt, and obtaining and storing corresponding evaluation information includes:

[0039] Inputting the second Prompt prompt information integrated with the context prompt information as additional input content into the sixth agent;

[0040] Through the sixth agent, a performance evaluation of the annual data is performed based on the Prompt prompt information, and the second evaluation information obtained by the evaluation is returned to the information pool of the first agent.

[0041] Optionally, the calling of at least one agent to perform performance evaluation on quarterly, annual, and historical data units based on the context prompt information Prompt, and obtaining and storing corresponding evaluation information includes:

[0042] Inputting the third Prompt prompt information integrated with the context prompt information as additional input content into the seventh agent;

[0043] The seventh agent performs a performance evaluation of the historical data based on the Prompt information, and returns the third evaluation information obtained from the evaluation to the information pool of the first agent.

[0044] Optionally, the fusion of evaluation information of different granularities to generate a final comprehensive financial performance evaluation result includes:

[0045] Performing similarity calculations on the quarterly, annual, and historical data-level assessment information by a first agent, wherein the similarity calculations are based on similarity of ESG-related indicators, similarity of KPIs, similarity of assessment results, and similarity of risk levels;

[0046] When the similarity exceeds a preset threshold, similar evaluation information is merged through the first agent;

[0047] When there is a conflict in assessment results or risk levels, historical data-level assessment information is given priority, and the conflict is resolved by combining the assessment results with contextual information;

[0048] Supplement the quarterly and annual assessment information with details, and integrate the supplemented annual assessment information into the historical data assessment information to generate a fused assessment containing detailed information.

[0049] sorting the fused evaluation information by the first agent to construct an evaluation chain, wherein the evaluation chain represents the temporal sequence and causal relationship between the evaluation information;

[0050] Generate integrated comprehensive financial performance evaluation results.

[0051] Optionally, the similarity calculation performed on the quarterly, annual, and historical data level evaluation information by the first agent, wherein the similarity calculation is based on the similarity of ESG-related indicators, KPI similarity, evaluation result similarity, and risk level similarity, including:

[0052] Calculate the similarity of ESG-related indicators, KPI similarity, assessment results similarity, and risk level similarity respectively;

[0053] And the total similarity is calculated by the following formula:

[0054] Total similarity = w1 × ESG-related indicator similarity + w2 × KPI similarity + w3 × assessment result similarity + w4 × risk level similarity;

[0055] Among them, w1, w2, w3, and w4 represent the corresponding weights respectively.

[0056] Optionally, the ESG-related indicator similarity, KPI similarity, assessment result similarity, and risk level similarity are calculated using the following formula:

[0057] Element similarity (e1, e2) = (V(e1)·V(e2)) / (||V(e1)||*||V(e2)||);

[0058] Where: V(e1) and V(e2) are the vector representations of element e1 and element e2 respectively;

[0059] "·" represents the dot product of vectors;

[0060] "||V(e1)||" and "||V(e2)||" represent the norms of vectors V(e1) and V(e2), respectively.

[0061] Optionally, the evaluation result similarity and risk level similarity are calculated using the following formula:

[0062] The evaluation result similarity (R1, R2) = 1 if R1 = R2;

[0063] Evaluation result similarity (R1, R2) = 0, if R1≠R2;

[0064] Risk level similarity (L1, L2) = 1 if L1 = L2;

[0065] Risk level similarity (L1, L2) = 0 if L1≠L2.

[0066] The second aspect of the present application provides an ESG-based enterprise comprehensive financial performance evaluation system, the system comprising:

[0067] An ESG factor identification unit is configured to receive various financial and non-financial data of an enterprise, perform ESG factor identification on the data through a first agent calling at least one second agent, obtain ESG-related indicators and their corresponding classifications in the data, and store the identification results in an information pool;

[0068] a KPI identification unit, configured to call at least one third agent to identify key performance indicators (KPIs) in the data, obtain corresponding weight information, and store the identification results of the KPIs and weight information in an information pool;

[0069] A parsing unit, configured to parse the ESG-related indicators and KPIs and generate a set of evaluation indicators based on a predetermined ESG template library;

[0070] Invoking at least one fourth agent to perform a comprehensive evaluation on the set of evaluation indicators, obtaining an evaluation result and its corresponding risk level, and storing the evaluation result and risk level in an information pool;

[0071] Divide the data into data units of different granularities, including quarterly, annual, and historical data;

[0072] For each data unit, the system queries the information pool for ESG-related indicators, KPIs, assessment results, and risk levels, and integrates these to generate contextual prompts of the corresponding granularity.

[0073] Based on the context prompt information Prompt, call at least one agent to perform performance evaluation on the quarterly, annual and historical data units respectively, and obtain and store corresponding evaluation information;

[0074] The evaluation information of different granularities is integrated to generate the final comprehensive financial performance evaluation results.

[0075] Optionally, the parsing unit is specifically used to:

[0076] The first agent traverses all ESG-related indicators and KPIs and parses them to obtain all evaluation indicators and form an evaluation indicator set;

[0077] The first agent calls the fourth agent to perform a comprehensive evaluation on each evaluation indicator in the evaluation indicator set to obtain all evaluation results and their corresponding risk levels.

[0078] Optionally, the segmentation unit is specifically used to:

[0079] Dividing the data into quarterly data by the first agent;

[0080] The query unit is specifically used to:

[0081] For each quarterly data, the first agent queries the information pool for the ESG-related indicators, KPIs, assessment results, and risk levels involved;

[0082] Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate the first prompt information corresponding to the quarter.

[0083] Optionally, the segmentation unit is specifically used to:

[0084] Dividing the data into annual data by the first agent;

[0085] The query unit is specifically used to:

[0086] For each year's data, the first agent queries the information pool for the ESG-related indicators, KPIs, assessment results, and risk levels involved;

[0087] Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate the second prompt information corresponding to the year.

[0088] Optionally, the segmentation unit is specifically used to:

[0089] dividing the data into historical data by the first agent;

[0090] The query unit is specifically used to:

[0091] For each historical data, query the ESG-related indicators, KPIs, assessment results and risk levels involved from the information pool through the first agent;

[0092] Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate a third Prompt message corresponding to the historical data.

[0093] Optionally, the performance evaluation unit is specifically used to:

[0094] Inputting the first prompt information integrated with the context prompt information into the fifth agent as additional input content through the first agent;

[0095] Through the fifth agent, a performance evaluation of the quarterly data is performed based on the Prompt information, and the first evaluation information obtained by the evaluation is returned to the information pool of the first agent.

[0096] Through the sixth agent, a performance evaluation of the annual data is performed based on the Prompt prompt information, and the second evaluation information obtained by the evaluation is returned to the information pool of the first agent.

[0097] The seventh agent performs a performance evaluation of the historical data based on the Prompt information, and returns the third evaluation information obtained from the evaluation to the information pool of the first agent.

[0098] Optionally, the evaluation fusion unit is specifically configured to: perform similarity calculation on the quarterly, annual, and historical data level evaluation information through the first agent, wherein the similarity calculation is based on ESG-related indicator similarity, KPI similarity, evaluation result similarity, and risk level similarity;

[0099] When the similarity exceeds a preset threshold, similar evaluation information is merged through the first agent;

[0100] When there is a conflict in assessment results or risk levels, historical data-level assessment information is given priority, and the conflict is resolved by combining the assessment results with contextual information;

[0101] Supplement the quarterly and annual assessment information with details, and integrate the supplemented annual assessment information into the historical data assessment information to generate a fused assessment containing detailed information.

[0102] sorting the fused evaluation information by the first agent to construct an evaluation chain, wherein the evaluation chain represents the temporal sequence and causal relationship between the evaluation information;

[0103] Generate integrated comprehensive financial performance evaluation results.

[0104] Optionally, the evaluation fusion unit is specifically used to:

[0105] Calculate the similarity of ESG-related indicators, KPI similarity, assessment results similarity, and risk level similarity respectively;

[0106] And the total similarity is calculated by the following formula:

[0107] Total similarity = w1 × ESG-related indicator similarity + w2 × KPI similarity + w3 × assessment result similarity + w4 × risk level similarity;

[0108] Among them, w1, w2, w3, and w4 represent the corresponding weights respectively.

[0109] In a third aspect, the present application provides an ESG-based enterprise comprehensive financial performance assessment device, comprising:

[0110] processor, memory, input and output units, and buses;

[0111] The processor is connected to the memory, the input and output unit, and the bus;

[0112] The memory stores a program, and the processor calls the program to execute the first aspect and any optional method in the first aspect.

[0113] In a fourth aspect, the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed on a computer, the program executes the first aspect and any optional method in the first aspect. Beneficial effects

[0114] By categorizing corporate data into quarterly, annual, and historical data granularity units, we can more accurately capture ESG-related indicators and KPIs across different timescales. This tiered approach improves the comprehensiveness and accuracy of performance evaluations, particularly when evaluating historical data.

[0115] This approach leverages multiple agents working together to perform different tasks, such as ESG factor identification, KPI identification, and comprehensive assessment. This modularizes the entire performance review process. Each agent focuses on a specific task, improving efficiency and accuracy.

[0116] By integrating ESG-related indicators, KPIs, assessment results and risk levels to generate contextual prompt information, and conducting performance evaluation at different granularities, contextual information is effectively utilized, thereby improving the accuracy of performance evaluation.

[0117] The evaluation results of all agents are stored in an information pool, which supports sharing and querying of results across data units of different granularities. This information pool mechanism makes information exchange and result integration between agents more efficient, ensuring the consistency and integrity of multi-granularity performance evaluation results.

[0118] By calculating and integrating similarity between quarterly, annual, and historical data-level evaluation information, we resolve the differences and conflicts between evaluation information of different granularities, ensuring that the final performance evaluation results are more accurate and have a global perspective.

[0119] This method is specially optimized for performance evaluation of historical data, enabling it to handle complex and multi-dimensional performance evaluation tasks, overcoming the limitations of traditional methods in historical data evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0120] In order to more clearly illustrate the technical solutions in this application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0121] FIG1 is a flow chart of an embodiment of an ESG-based enterprise comprehensive financial performance assessment method provided in this application;

[0122] FIG2 is a flow chart of another embodiment of the ESG-based enterprise comprehensive financial performance assessment method provided in this application;

[0123] FIG3 is a flow chart of another embodiment of the ESG-based enterprise comprehensive financial performance assessment method provided in this application;

[0124] FIG4 is a schematic diagram of the structure of an embodiment of an ESG-based enterprise comprehensive financial performance evaluation system provided in this application;

[0125] FIG5 is a schematic structural diagram of an embodiment of an ESG-based enterprise comprehensive financial performance assessment device provided in this application;

[0126] FIG6 is a schematic diagram of a multi-agent collaborative structure provided in this application. Modes for Carrying Out the Invention

[0127] This invention provides a method, system, and device for evaluating the comprehensive financial performance of enterprises based on ESG (environmental, social, and governance) considerations. These methods aim to overcome the limitations of existing technologies for evaluating enterprise performance and provide a more comprehensive and accurate assessment solution. To illustrate the technical solutions and advantages of this invention, the following detailed description is provided with reference to specific embodiments and accompanying figures.

[0128] Figure 1 shows a flowchart of an embodiment of the ESG-based enterprise comprehensive financial performance assessment method of the present invention. Figures 2 and 3 show flowcharts of two other embodiments of the method, further describing in detail the specific operations of the different steps. Figure 4 shows a structural diagram of an embodiment of an ESG-based enterprise comprehensive financial performance assessment system, Figure 5 shows a structural diagram of an embodiment of an ESG-based enterprise comprehensive financial performance assessment device, and Figure 6 shows a multi-agent collaborative structure diagram. Through these figures, the system architecture and workflow of the present invention can be more intuitively understood.

[0129] The system of the present invention first receives various financial and non-financial data from a company. This data includes, but is not limited to, financial statements, environmental reports, social responsibility reports, and corporate governance reports. After receiving the data, a first agent (Agent 1 in Figure 6) invokes at least one second agent (Agent 2 in Figure 6) to identify ESG factors within the data. Using pre-trained natural language processing models and data mining techniques, the second agent identifies ESG-related indicators and their corresponding categories within the data and stores the identified results in an information pool (Information Pool 1 in Figure 4).

[0130] While identifying ESG factors, the first agent invokes at least one third agent (Agent 3 in Figure 6) to identify key performance indicators (KPIs) in the data. Using machine learning algorithms and expert systems, the third agent extracts KPIs from the financial data and calculates their corresponding weights. The identified KPIs and weights are also stored in Information Pool 1 for subsequent use.

[0131] After identifying ESG-related indicators and KPIs, the first agent invokes a parsing unit (parsing unit 4 in Figure 4) to analyze these indicators. Parsing unit 4 traverses all ESG-related indicators and KPIs in information pool 1 and performs detailed analysis to obtain all evaluation indicators and form an evaluation indicator set. During the parsing process, parsing unit 4 categorizes different indicators based on a predefined ESG template library (ESG template library 5 in Figure 4) and generates a standardized evaluation indicator set.

[0132] After generating a set of evaluation indicators, the first agent invokes at least one fourth agent (agent 4 in Figure 6) to conduct a comprehensive assessment of each indicator in the set. Using a multi-factor analysis model, the fourth agent scores each indicator and determines its corresponding risk level. The assessment results and risk level information are stored in information pool 1, forming a complete evaluation results database.

[0133] Next, the system segments the received data into data units of varying granularity, including quarterly, annual, and historical data. This data segmentation is performed by the segmentation unit in the first agent (Segmentation Unit 6 in Figure 4). For each data unit, the first agent queries Information Pool 1 for the relevant ESG indicators, KPIs, assessment results, and risk levels, and integrates these information to generate contextual prompts of the corresponding granularity.

[0134] Quarterly data segmentation: The data is segmented into quarterly data by segmentation unit 6. For each quarterly data, the first agent queries the information pool 1 for the relevant ESG indicators, KPIs, assessment results, and risk levels, and integrates this information to generate the first prompt information corresponding to the quarter.

[0135] Annual data segmentation: The data is segmented into annual data by segmentation unit 6. For each annual data, the first agent queries the information pool 1 for the relevant ESG indicators, KPIs, assessment results, and risk levels, and integrates this information to generate the second prompt information corresponding to the year.

[0136] Historical data segmentation: The data is segmented into historical data by segmentation unit 6. For each historical data, the first agent queries the information pool 1 for the ESG-related indicators, KPIs, assessment results, and risk levels involved, and integrates this information to generate a third prompt message corresponding to the historical data.

[0137] Based on the generated context prompt information Prompt, the first agent calls different agents to perform performance evaluation on quarterly, annual and historical data units respectively, and stores the evaluation results in information pool 1.

[0138] Quarterly Data Performance Evaluation: The first prompt information is fed as additional input to the fifth agent (Agent 5 in Figure 6). Based on the first prompt information, the fifth agent uses time series analysis and forecasting models to perform a performance evaluation on the quarterly data. The evaluation results include ESG scores, KPI scores, and risk levels. The evaluation results are returned to the first agent and stored in Information Pool 1.

[0139] Annual Data Performance Evaluation: The second prompt is fed as additional input to the sixth agent (Agent 6 in Figure 6). Based on the second prompt, the sixth agent uses regression analysis and an expert system to perform a performance evaluation of the annual data. The evaluation results include ESG scores, KPI scores, and risk levels. The evaluation results are returned to the first agent and stored in Information Pool 1.

[0140] Historical Data Performance Evaluation: The third prompt is fed as additional input to the seventh agent (Agent 7 in Figure 6). Based on the third prompt, the seventh agent uses historical data analysis and trend prediction models to perform a performance evaluation of the historical data. The evaluation results include ESG scores, KPI scores, and risk levels. The evaluation results are returned to the first agent and stored in Information Pool 1.

[0141] To generate the final comprehensive financial performance evaluation results, the first agent calls the evaluation fusion unit (evaluation fusion unit 8 in Figure 4) to fuse the evaluation information of different granularities. The evaluation fusion process includes the following steps:

[0142] Similarity Calculation: The evaluation fusion unit 8 uses the first agent to perform similarity calculations on the quarterly, annual, and historical data-level evaluation information. The similarity calculation is based on the similarity of ESG-related indicators, KPI similarity, evaluation result similarity, and risk level similarity. The specific calculation formula is as follows:

[0143] Similarity of ESG-related indicators: Similarity of ESG-related indicators (e1, e2) = V(e1)⋅V(e2)∣∣V(e1)∣∣×∣∣V(e2)∣∣ Similarity of ESG-related indicators (e1, e2) = ∣∣V(e1)∣∣×∣∣V(e2)∣∣V(e1)⋅V(e2)​

[0144] KPI similarity: KPI similarity (e1, e2) = V(e1)⋅V(e2)∣∣V(e1)∣∣×∣∣V(e2)∣∣KPI similarity (e1, e2) = ∣∣V(e1)∣∣×∣∣V(e2)∣∣V(e1)⋅V(e2)​

[0145] Evaluation result similarity: Evaluation result similarity (R1, R2) = {1, if R1 = R20, if R1 ≠ R2 Evaluation result similarity (R1, R2) = {1, 0, if R1 = R2 if R1 ≠ R2

[0146] Risk level similarity: Risk level similarity (L1, L2) = {1, if L1 = L2 0, if L1 ≠ L2 Risk level similarity (L1, L2) = {1, 0, if L1 = L2 if L1 ≠ L2

[0147] The calculation formula for the total similarity is:

[0148] Total similarity = w1 × ESG-related indicator similarity + w2 × KPI similarity + w3 × assessment result similarity + w4 × risk level similarity Total similarity = w1 × ESG-related indicator similarity + w2 × KPI similarity + w3 × assessment result similarity + w4 × risk level similarity

[0149] Among them, w1,w2,w3,w4w1,w2,w3,w4 represent the weights of each similarity respectively.

[0150] Similar evaluation information merging: When the total similarity exceeds the preset threshold, similar evaluation information is merged through the first agent to ensure the consistency of evaluation information of different granularities.

[0151] Conflict Resolution: When conflicting assessment results or risk levels occur, historical data-level assessment information is prioritized and contextual information is used to resolve the conflict. For example, if a quarterly assessment result is inconsistent with a historical assessment result, the system will prioritize the historical data-level assessment result and make adjustments based on contextual information.

[0152] Detail Supplementation and Integration: Detail is supplemented for the quarterly and annual assessments. The resulting annual assessment information is then integrated into the historical data assessment to create a fused assessment containing detailed information. For example, the annual assessment might include more detailed financial analysis and market trend forecasts, which are then incorporated into the historical data assessment to make it more comprehensive.

[0153] Evaluation Chain Construction: The first agent sorts the fused evaluation information to construct an evaluation chain. The evaluation chain represents the temporal sequence and causal relationships between evaluation information, helping users better understand the changing trends and influencing factors of enterprise performance.

[0154] Finally, the evaluation and fusion unit 8 generates integrated financial performance assessment results. These results include the company's overall ESG score, KPI score, risk level, and a detailed performance analysis report. These results can be output to users via the input and output unit (input and output unit 10 in Figure 5) for reference when the company makes strategic decisions.

[0155] Taking a large manufacturing enterprise as an example, the specific steps for conducting comprehensive financial performance assessment using the system of the present invention are as follows:

[0156] Data Collection: The company collected three consecutive years of financial statements, annual environmental reports, social responsibility reports, and corporate governance reports. This data was transmitted to the system via the data receiving unit (data receiving unit 9 in Figure 4).

[0157] ESG factor identification: The first agent calls the second agent to identify the company's ESG-related indicators such as environmental protection measures, resource utilization efficiency, employee benefits, community contributions, board structure, etc., and classify and store them in information pool 1.

[0158] KPI identification: The first agent calls the third agent to identify the company's key performance indicators such as net profit, operating income growth rate, cost control ability, etc., and calculates their weight information and stores it in information pool 1.

[0159] Generation of an evaluation indicator set: Parsing unit 4 traverses all ESG-related indicators and KPIs in information pool 1 and generates an evaluation indicator set. For example, environmental indicators may include wastewater discharge, exhaust emissions, and energy consumption; social indicators may include employee satisfaction, community relations, and charitable donations; and governance indicators may include board independence, shareholder protection, and regulatory compliance.

[0160] Comprehensive Assessment: The fourth agent conducts a comprehensive assessment of the set of evaluation indicators, generating a score and risk level for each indicator. For example, a company's wastewater discharge volume may be rated as low risk, while its cost control capabilities may be rated as medium risk.

[0161] Data segmentation: Segmentation unit 6 divides the data into quarterly data, annual data and historical data.

[0162] Granular performance evaluation:

[0163] Quarterly data: The fifth agent conducts performance evaluation based on quarterly data. The evaluation results include ESG scores, KPI scores and risk levels for each quarter.

[0164] Annual data: The sixth intelligent entity conducts performance evaluation based on annual data. The evaluation results include the ESG score, KPI score and risk level for each year.

[0165] Historical data: The Seventh Agent conducts performance evaluation based on historical data. The evaluation results include three years of historical ESG scores, KPI scores and risk levels.

[0166] Evaluation Information Fusion: Evaluation Fusion Unit 8 calculates similarities between quarterly, annual, and historical data-level evaluation information, merges similar evaluation information, resolves conflicts, and supplements details. The resulting comprehensive financial performance assessment includes the company's overall ESG score, KPI score, risk level, and detailed analysis of performance trends and influencing factors.

[0167] Technical details and implementation of the present invention

[0168] ESG factor identification: The second agent can use deep learning models such as BERT (Bidirectional Encoder Representations from Transformers) for natural language processing to extract ESG-related keywords and sentences. During the identification process, a pre-trained ESG classification model can be used to classify keywords and sentences into three major categories: environment, society, and governance. For example, the BERT model can calculate the vector representation of a sentence using the following formula:

[0169] V(e)=BERT(e)V(e)=BERT(e)

[0170] Among them, ee represents the input text or data.

[0171] KPI Identification: The third agent can use regression analysis and decision tree algorithms to identify and weight key performance indicators in financial data. Specifically, regression analysis can be used to identify important variables in the data, and decision tree algorithms can be used to calculate the weights of these variables. For example, regression analysis can be performed using the following formula:

[0172] KPI=β0+β1x1+β2x2+⋯+βnxnKPI=β0​+β1​x1​+β2​x2​+⋯+βn​xn​

[0173] Among them, x1, x2,…, xn represent the variables in the financial data, and β1, β2,…, βn represent the weights of each variable.

[0174] Comprehensive evaluation: The fourth agent can use a multi-factor analysis model, such as AHP (Analytic Hierarchy Process), for comprehensive evaluation. The AHP model establishes a hierarchical structure model, hierarchically processes different evaluation indicators, and calculates the weight of each indicator through a pairwise comparison matrix. For example, the pairwise comparison matrix can be expressed as:

[0175] A=[1w12w13⋯w1n1w121w23⋯w2n1w131w231⋯w3n⋮⋮⋮⋱⋮1w1n1w2n1w3n⋯1]A

[0176] =1w12​1​w13​1​⋮w1n​1​​w12​1w23​1​⋮w2n​1​​w13​w23​1⋮w3n​1​​⋯⋯⋯⋱⋯​w1n​w2n​w3n​⋮1​​

[0177] Among them, wijwij represents the relative importance of indicator ii and indicator jj.

[0178] Similarity Calculation: The evaluation fusion unit 8 calculates the similarity of the evaluation information at each granularity to ensure the consistency and integrity of the results. The similarity calculation formula is as described above, and the similarity is quantified through vector representation and dot product operation.

[0179] Conflict Resolution: When conflicting assessment results or risk levels arise, the Assessment Fusion Unit 8 prioritizes historical data-level assessment information and makes adjustments based on contextual information. For example, if a metric in a quarterly assessment is rated high risk, while the same metric in a historical assessment is rated low risk, the system will prioritize the historical assessment and make adjustments based on the specific circumstances of the quarterly data.

[0180] Detail Supplementation and Integration: Assessment Integration Unit 8 provides additional details for assessment information at different granularities, such as short-term performance trends within quarterly data, mid-term performance trends within annual data, and long-term performance trends within historical data. The resulting assessment results include detailed performance analysis and recommendations, helping companies better understand and improve their financial performance.

[0181] Using the system and method of this invention, a large manufacturing company successfully conducted a comprehensive assessment of its three-year integrated financial performance. The system not only identified the company's key ESG-related indicators, but also calculated the weights of key performance indicators and generated a detailed performance evaluation report. Through these evaluation results, the company identified deficiencies in environmental protection and social responsibility and developed corresponding improvement measures. Furthermore, the system integrated assessment information at different granularities to ensure the comprehensiveness and accuracy of the results. The company now uses these performance evaluation results more scientifically in its decision-making process, improving the scientific nature and effectiveness of its strategic decisions.

[0182] Figure 5 shows a block diagram of an embodiment of an ESG-based enterprise comprehensive financial performance evaluation device provided by the present invention. The device includes a processor 11, a memory 12, an input / output unit 10, and a bus 13. The processor 11 is connected to the memory 12, the input / output unit 10, and the bus 13. The memory 12 stores a program, which the processor 11 invokes to execute the steps of the above method.

[0183] Processor 11: The core component of the device, responsible for data processing, agent invocation, and evaluation information fusion. Processor 11 can be a high-performance CPU or GPU, capable of efficiently handling large amounts of data and complex computing tasks.

[0184] Memory 12: Memory 12 stores system programs and data. For example, memory 12 contains the ESG template library, data from information pool 1, evaluation indicator sets, evaluation results, and their risk levels. Memory 12 can be various types of memory, such as RAM or ROM, to ensure data security and fast access.

[0185] Input / Output Unit 10: The input / output unit 10 is used to receive enterprise data and output the final evaluation results. For example, the input / output unit 10 can receive the enterprise's financial statements and non-financial reports through a network interface and output the final performance evaluation report through a display screen or a file.

[0186] Bus 13: Bus 13 is a communication line connecting the processor 11, memory 12 and input / output unit 10, ensuring efficient data transmission and coordinated work among various components.

[0187] The present invention also provides a computer-readable storage medium storing a program that, when executed on a computer, performs the steps of the above method. The storage medium can be an optical disc, a magnetic disk, a USB flash drive, or other type of storage device, ensuring the reliability and portability of the program.

[0188] The present invention improves the comprehensiveness and accuracy of performance evaluation by dividing enterprise data into data units of different granularities at the quarterly, annual and historical data levels, and utilizing multiple intelligent agents to work together, responsible for tasks such as ESG factor identification, KPI identification, and comprehensive evaluation. The generation of contextual prompt information and the fusion of evaluation information of different granularities further ensure the consistency and integrity of the evaluation results. The present invention is particularly optimized for the performance evaluation of historical data, enabling it to handle complex and multi-dimensional performance evaluation tasks, overcoming the limitations of traditional methods in historical data evaluation. Through the system and method of the present invention, enterprises can make strategic decisions more scientifically and improve their overall financial performance.

Claims

1. A comprehensive financial performance assessment method for enterprises based on ESG, characterized by: The method comprises: Receive various financial and non-financial data of the enterprise, use the first agent to call at least one second agent to identify ESG factors in the data, obtain ESG-related indicators and their corresponding categories in the data, and store the identification results in an information pool; Invoking at least one third agent to identify key performance indicators (KPIs) in the data, obtain corresponding weight information, and store the identification results of the KPIs and weight information in an information pool; Analyze the ESG-related indicators and KPIs, and generate a set of evaluation indicators based on a predetermined ESG template library; Invoking at least one fourth agent to perform a comprehensive evaluation on the set of evaluation indicators, obtaining an evaluation result and its corresponding risk level, and storing the evaluation result and risk level in an information pool; Divide the data into data units of different granularities, including quarterly, annual, and historical data; For each data unit, the system queries the information pool for ESG-related indicators, KPIs, assessment results, and risk levels, and integrates these to generate contextual prompts of the corresponding granularity. Based on the context prompt information Prompt, call at least one agent to perform performance evaluation on the quarterly, annual and historical data units respectively, and obtain and store corresponding evaluation information; The evaluation information of different granularities is integrated to generate the final comprehensive financial performance evaluation results.

2. The ESG-based comprehensive financial performance assessment method for an enterprise according to claim 1, characterized in that: The step of analyzing the ESG-related indicators and KPIs and generating a set of evaluation indicators based on a predetermined ESG template library includes: The first agent traverses all ESG-related indicators and KPIs and parses them to obtain all evaluation indicators and form an evaluation indicator set; The first agent calls the fourth agent to perform a comprehensive evaluation on each evaluation indicator in the evaluation indicator set to obtain all evaluation results and their corresponding risk levels.

3. The ESG-based enterprise comprehensive financial performance assessment method according to claim 1, characterized in that: Divide the data into data units of different granularities, including quarterly, annual, and historical data, including: Dividing the data into quarterly data by the first agent; For each data unit, the relevant ESG indicators, KPIs, assessment results, and risk levels are queried from the information pool, and integrated to generate contextual prompts of corresponding granularity, including: For each quarterly data, the first agent queries the information pool for the ESG-related indicators, KPIs, assessment results, and risk levels involved; Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate the first prompt information corresponding to the quarter.

4. The ESG-based enterprise comprehensive financial performance assessment method according to claim 1, characterized in that: The data is divided into data units of different granularities, including annual data: Dividing the data into annual data by the first agent; For each data unit, the relevant ESG indicators, KPIs, assessment results, and risk levels are queried from the information pool, and integrated to generate contextual prompts of corresponding granularity, including: For each year's data, the first agent queries the information pool for the ESG-related indicators, KPIs, assessment results, and risk levels involved; Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate the second prompt information corresponding to the year.

5. The ESG-based enterprise comprehensive financial performance assessment method according to claim 1, characterized in that: The data is divided into data units of different granularities, including historical data including: dividing the data into historical data by the first agent; For each data unit, the relevant ESG indicators, KPIs, assessment results, and risk levels are queried from the information pool, and integrated to generate contextual prompts of corresponding granularity, including: For each historical data, query the ESG-related indicators, KPIs, assessment results and risk levels involved from the information pool through the first agent; Through the first intelligent agent, the ESG-related indicators, KPIs, assessment results and risk levels obtained from the query are integrated to generate a third Prompt message corresponding to the historical data.

6. The ESG-based enterprise comprehensive financial performance assessment method according to claim 3, characterized in that: The calling of at least one agent to perform performance evaluation on quarterly, annual, and historical data units based on the context prompt information Prompt, and obtaining and storing corresponding evaluation information includes: Inputting the first prompt information integrated with the context prompt information into the fifth agent as additional input content through the first agent; Through the fifth agent, a performance evaluation of the quarterly data is performed based on the Prompt information, and the first evaluation information obtained by the evaluation is returned to the information pool of the first agent.

7. The ESG-based enterprise comprehensive financial performance assessment method according to claim 4, characterized in that: The calling of at least one agent to perform performance evaluation on quarterly, annual, and historical data units based on the context prompt information Prompt, and obtaining and storing corresponding evaluation information includes: Inputting the second Prompt prompt information integrated with the context prompt information as additional input content into the sixth agent; Through the sixth agent, a performance evaluation of the annual data is performed based on the Prompt prompt information, and the second evaluation information obtained by the evaluation is returned to the information pool of the first agent.

8. The ESG-based enterprise comprehensive financial performance assessment method according to claim 5, characterized in that: The calling of at least one agent to perform performance evaluation on quarterly, annual, and historical data units based on the context prompt information Prompt, and obtaining and storing corresponding evaluation information includes: Inputting the third Prompt prompt information integrated with the context prompt information as additional input content into the seventh agent; The seventh agent performs a performance evaluation of the historical data based on the Prompt information, and returns the third evaluation information obtained from the evaluation to the information pool of the first agent.

9. The ESG-based enterprise comprehensive financial performance assessment method according to claim 1, characterized in that: The above-mentioned fusion of evaluation information of different granularities to generate the final comprehensive financial performance evaluation results includes: Performing similarity calculations on the quarterly, annual, and historical data-level assessment information by a first agent, wherein the similarity calculations are based on similarity of ESG-related indicators, similarity of KPIs, similarity of assessment results, and similarity of risk levels; When the similarity exceeds a preset threshold, similar evaluation information is merged through the first agent; When there is a conflict in assessment results or risk levels, historical data-level assessment information is given priority, and the conflict is resolved by combining the assessment results with contextual information; Supplement the quarterly and annual assessment information with details, and integrate the supplemented annual assessment information into the historical data assessment information to generate a fused assessment containing detailed information. sorting the fused evaluation information by the first agent to construct an evaluation chain, wherein the evaluation chain represents the temporal sequence and causal relationship between the evaluation information; Generate integrated comprehensive financial performance evaluation results.

10. The ESG-based enterprise comprehensive financial performance assessment method according to claim 9, characterized in that: The similarity calculation of the quarterly, annual, and historical data level evaluation information by the first agent is based on the similarity of ESG-related indicators, KPI similarity, evaluation result similarity, and risk level similarity, including: Calculate the similarity of ESG-related indicators, KPI similarity, assessment results similarity, and risk level similarity respectively; And the total similarity is calculated by the following formula: Total similarity = w1 × ESG-related indicator similarity + w2 × KPI similarity + w3 × assessment result similarity + w4 × risk level similarity; Among them, w1, w2, w3, and w4 represent the corresponding weights respectively.

Images