An intelligent enterprise decision-making execution method, system and electronic device

Abstract

The application relates to the technical field of artificial intelligence and big data analysis, and discloses an intelligent enterprise decision execution method and system and electronic equipment. The method comprises the following steps: continuously collecting and integrating internal and external multi-source heterogeneous data of a target enterprise to construct and dynamically update a market value digital twin; setting a preset multi-dimensional factor based on the multi-source heterogeneous data, extracting data corresponding to each dimension factor in a preset order, performing adaptive weight fusion calculation of the data through an adaptive value quantification model deployed in the digital twin to generate enterprise decision basis data; generating a strategy report draft with a preset directory structure based on the decision basis data, and forming a final version of the report through man-machine collaborative review; analyzing the final version of the report to generate executable tasks and dispatching the executable tasks to an execution end in response to a management confirmation instruction; and collecting execution feedback data and synchronizing the execution feedback data to the digital twin and the model. The application improves the scientific nature, accuracy and efficiency of enterprise decision execution, and effectively reduces compliance risks and operation and maintenance costs.

Description

Technical Field

[0001] This invention relates to artificial intelligence and big data analysis, specifically to an intelligent enterprise decision-making execution method, system, and electronic device. Background Technology

[0002] To enhance international competitiveness and influence, and to facilitate market expansion and quality improvement, market capitalization management, as a core function of listed companies' decision-making and management, no longer needs to be limited to bridging the gap between intrinsic value and market performance; it must also possess a global perspective. For listed companies, focusing solely on value delivery is far from sufficient; how to achieve significant growth within a global context is crucial for long-term development. However, the existing corporate decision-making and execution support system faces the following key bottlenecks, severely hindering the achievement of these goals. Specifically: 1. Insufficient forward-looking insights in analysis dimensions. Existing tools are mostly data silos, with financial, public opinion, and industry dimensions being separated from each other, lacking a dynamic analysis framework that deeply integrates multi-source information.

[0003] 2. Disconnect between decision-making and execution. Analytical outputs often remain at the level of macro-level decision-making recommendations, severely disconnected from specific action plans of business departments. The lack of a technological bridge to translate decisions into actionable and traceable tasks leads to "reports exist, but timeliness is poor." Existing systems generate mostly unstructured reports, lacking supporting natural language understanding and automated task decomposition technologies. These reports cannot be directly recognized and processed by downstream business execution systems such as OA and ERP, resulting in significant manual intervention and secondary transformation of decision-making information during transmission. The system has low automation and poor information flow efficiency.

[0004] 3. Rigid models lack dynamic adaptation. The value logic of different enterprises varies greatly depending on their industry, business scale, and life cycle stage. Existing solutions are mostly static, general, and fixed models that cannot intelligently identify and adapt to the unique endowments of enterprises, resulting in low practicality.

[0005] 4. Management processes lack iterative optimization. A complete management loop has not been formed, from problem diagnosis and decision generation to action implementation and effect evaluation. Each link is isolated, and the effectiveness of strategies cannot be quantified and fed back to the system, hindering the accumulation of experience and continuous iterative optimization of strategies. Summary of the Invention

[0006] To address the technical problems of data silos, rigid models, and the disconnect between decision-making and execution in existing enterprise decision-making and execution systems, this invention provides an intelligent enterprise decision-making and execution method, system, and electronic device. It solves the problem of cognitive fragmentation through digital twins, generates enterprise decision-making basis data through an adaptive value quantification model, and solves the problem of the disconnect between decision-making and execution through intelligent report generation and automatic task distribution, ultimately forming a powerful technological closed loop.

[0007] In a first aspect, the present invention provides an intelligent enterprise decision-making execution method, comprising: Continuously collect and integrate heterogeneous data from multiple sources inside and outside the target company, and process it through a multi-source consistency verification algorithm to build and dynamically update the company's market value digital twin; Based on multi-source heterogeneous data, preset multi-dimensional factors are set, and data corresponding to each dimension factor is extracted in a preset order. Through the adaptive value quantification model deployed in the market value digital twin, adaptive weight fusion calculation of the data is performed to generate data for enterprise decision-making basis. Based on enterprise decision-making data, a preliminary draft of a strategy report with a preset directory structure is automatically generated, and a human-machine collaborative interface is provided to review and revise the preliminary draft of the report, forming a final version to be confirmed; In response to the management confirmation instruction of the final version, the structured content of the final version is parsed, and corresponding executable tasks are generated and assigned to the corresponding execution terminals according to the predefined responsibility mapping relationship; Collect task execution result feedback data and synchronize it to the market value digital twin and the adaptive value quantification model to evaluate the execution effect and optimize the adaptive value quantification model.

[0008] This invention constructs a dynamic market capitalization digital twin by integrating and verifying multi-source heterogeneous data, breaking down data silos and providing comprehensive, high-quality data support for decision-making. Relying on multi-dimensional factors and an adaptive value quantification model, it accurately mines the value transmission logic, solving the rigidity problem of traditional models. Intelligent report generation and human-machine collaborative review are efficiently integrated, balancing the advantages of AI automation with human experience to ensure reports are relevant to reality. Report parsing and automatic task assignment bridge the gap between decision-making and execution, clarifying responsibility boundaries and improving collaboration and implementation efficiency. Execution result feedback and model optimization form a closed loop, continuously adapting to enterprise development and market changes, reducing the cost of repeated trial and error, strengthening compliance risk control, and building a long-term digital closed-loop management barrier, comprehensively improving the accuracy, efficiency, and sustainability of enterprise decision-making and execution.

[0009] Secondly, embodiments of the present invention provide an intelligent enterprise decision-making and execution system, comprising: The multi-source heterogeneous data acquisition module is used to continuously collect and integrate multi-source heterogeneous data from the target enterprise, and after processing by the multi-source consistency verification algorithm, to build and dynamically update the enterprise's market value digital twin. The enterprise decision-making basis data generation module is used to set preset multi-dimensional factors based on multi-source heterogeneous data, extract the data corresponding to each dimension factor in a preset order, and generate enterprise decision-making basis data by performing adaptive weight fusion calculation of the data through the adaptive value quantification model deployed in the market value digital twin. The report generation module is used to automatically generate a draft strategy report with a preset directory structure based on enterprise decision-making data, and provides a human-machine collaborative interface to review and revise the draft report to form a final version to be confirmed; The task generation and dispatch module is used to respond to the management confirmation instruction of the final version, parse the structured content of the final version, and generate corresponding executable tasks according to the predefined responsibility mapping relationship and dispatch them to the corresponding execution end. The feedback and optimization module is used to collect task execution result feedback data and synchronize it to the market value digital twin and the adaptive value quantification model, so as to evaluate the execution effect and optimize the adaptive value quantification model.

[0010] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the intelligent enterprise decision execution method of the first aspect or any corresponding embodiment described above.

[0011] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the intelligent enterprise decision-making execution method of the first aspect or any corresponding embodiment described above.

[0012] Fifthly, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute the intelligent enterprise decision-making execution method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0013] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating an intelligent enterprise decision-making execution method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating another intelligent enterprise decision-making execution method according to an embodiment of the present invention; Figure 3 This is a structural block diagram of an intelligent enterprise decision-making and execution system according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0017] This invention provides an embodiment of an intelligent enterprise decision-making execution method, aiming to construct a digital twin for enterprise market value management. This twin not only dynamically reflects the internal and external state of the enterprise but also integrates non-financial and financial indicators through a built-in value quantification model driven by a large model, ensuring the accuracy and reliability of the twin mapping. This enables the value quantification model to possess adaptive and self-learning capabilities, dynamically adjusting the weights and structure of its core driving factors based on the specific characteristics of the target company. Thus, while maintaining the generality of the system architecture, it achieves intelligent execution of various enterprise decisions. It should be noted that the steps shown in the flowcharts can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0018] Figure 1 This is a flowchart of an intelligent enterprise decision-making execution method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Continuously collect and integrate heterogeneous data from multiple sources inside and outside the target enterprise, and process it through a multi-source consistency verification algorithm to construct and dynamically update the enterprise's market value digital twin.

[0019] Specifically, embodiments of the present invention construct a digital twin of market capitalization as a virtual mirror of a company's market capitalization through comprehensive data collection and dynamic integration. In an optional embodiment, the heterogeneous data from multiple sources both internally and externally includes: 1. Internal business data: Obtain structured data such as enterprise financial system (revenue, profit, cash flow, etc.), operational data (production capacity, supply chain flow, employee activities), and internal management data (governance, ESG) from business systems such as ERP, CRM, and SCM through API interfaces.

[0020] 2. External industry data: Obtain market data (e.g., stock price, trading volume, analyst ratings), industry data (industry-related regulations, industry prosperity, upstream and downstream dynamics of the industrial chain), unstructured data (news sentiment, social media narratives, research report opinions) from financial data terminals (e.g., Wind, Bloomberg), as well as third-party benchmark data (industry ESG ratings, market share statistics).

[0021] 3. Data collected by IoT terminals: Real-time data collected through devices such as smart sensors and vehicle terminals, such as micro-behavioral data like production line energy consumption, carbon emission data, and supply chain logistics trajectories.

[0022] Furthermore, the collected multi-source heterogeneous data undergoes format standardization processing, including data alignment, feature extraction, and contradiction verification, to ensure data consistency and reliability, thereby constructing and dynamically updating a digital twin of the enterprise's market value and realizing a panoramic presentation of the enterprise's status.

[0023] In this embodiment of the invention, a pre-release verification mechanism for internal enterprise information is provided. This mechanism automatically compares the information to be released with internal business data, specifically scanning for "greenwashing" risks in ESG-related statements. This improves information quality from the source and reduces compliance and reputational risks. In one embodiment, identifying ESG greenwashing risks in the target enterprise's internal business data specifically includes: Quantitative Claim Verification: Extract quantitative data from ESG claims and compare it with the original data in the internal management system for consistency. For example, extract quantitative claims (such as "30% emission reduction") from publicly available corporate texts and automatically query the corresponding data in the internal database. Use rule matching and numerical comparison algorithms to calculate the relative deviation |claimed value - actual value| / actual value. If the deviation exceeds the threshold (such as 10%), a "data exaggeration" warning will be triggered.

[0024] Qualitative Claim Analysis: Natural language processing technology is used to perform semantic analysis on vague qualitative statements and evaluate their credibility by comparing them with third-party industry benchmark data. For example, based on knowledge graph semantic reasoning algorithms, when the text contains vague statements such as "industry leader" or "outstanding contribution", the system compares it with the benchmark values ​​provided by relevant third parties. If the company's actual performance indicators are not in the top 20% of the industry (configurable), the system will prompt "claims lack benchmark support".

[0025] Consistency Tracking: Establish a database of corporate ESG commitments, comparing new statements with historical commitments to check for contradictions or failures to meet commitments. For example, when a company releases a new statement, it is automatically compared with historical commitments to check for any actions that fail to meet commitments but still promote future goals, or any contradictions between previous and current statements.

[0026] Step S102: Based on multi-source heterogeneous data, preset multi-dimensional factors are set, and data corresponding to each dimension factor is extracted in a preset order. Through the adaptive value quantification model deployed in the market value digital twin, adaptive weight fusion calculation of the data is performed to generate enterprise decision-making basis data.

[0027] Specifically, in this embodiment of the invention, for multi-source heterogeneous data that determines market capitalization, multi-dimensional factors including financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors are selected, and the data corresponding to the financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors are extracted in sequence.

[0028] In this embodiment of the invention, the financial factor, systemic importance factor, and environmental and social governance factor form a progressive dimensional space, while the short-term disturbance factor is a superimposed dimension, and the data corresponding to this dimension can be positive or negative. Specifically: 1. Financial Factors (CFA_Base related): Focusing on the intrinsic value of a company, covering dimensions such as earnings quality (earnings sustainability, cash content of revenue), growth quality (internal growth momentum, growth sustainability), and asset health (asset turnover efficiency, hidden liabilities). By continuously accessing structured data such as the company's financial statements and operating data, and using financial calculation algorithms (such as the discounted cash flow (DCF) model) based on real-time data, the company's basic financial value is dynamically calculated.

[0029] 2. Systemic Importance Factor (related to SCI_Modifier): This involves adapting and modifying the Systemic Commercial Influence (SCI) framework for assessing the systemic importance of financial institutions. It calculates amplification or reduction factors to evaluate a company's position within the industry chain. Through factors such as scale, relevance, substitutability, complexity, and cross-border activity, it quantifies a company's position within the industry and supply chain.

[0030] 3. Environmental and Social Governance Factors (ESG_Modifier Related): These are sustainable development adjustment coefficients, calculated based on ESG performance to adjust for risks and opportunities. They serve as a dynamic risk and opportunity pricer, quantifying the impact of non-financial factors on long-term value. The Environmental (E) dimension includes compliance risks and opportunities, carbon cost efficiency, etc.; the Social (S) dimension covers employee satisfaction, supply chain labor rights, etc.; and the Governance (G) dimension involves board effectiveness, allocation capabilities, etc. These are transformed into quantifiable financial impacts through risk mapping and opportunity identification models.

[0031] The adaptive value quantification model provided by this invention uses the following formula for value quantification:

[0032] Where MV_t represents the market value of the enterprise at time t, Ω_t represents the dynamic information set at time t, CFA_Base(Ω_t) is a financial factor that represents the basic financial value dynamically calculated based on the information set Ω_t at time t, and SCI_Modifier(Ω_t) is a system importance factor, the standard of which is based on the enterprise's position in the industry chain and industry. ESG_Modifier(Ω_t) is an environmental and social governance factor, which represents the risk and opportunity adjustment coefficient based on the quantitative calculation of a company's environmental, social and governance performance. f_sentiment is a short-term disturbance capture factor, representing a non-linear function that captures short-term disturbances. The input variables of the f_sentiment function include: Investor_Structure_t, representing investor structure data; Narrative_t, representing market narrative quantitative data; Expectation_Gap_t, representing the difference between the consensus expectation and the intrinsic expectation calculated by the model; and Technicals_t, representing a technical indicator feature vector including trend following, momentum, and volume indicators. The final result may be positive or negative.

[0033] In this embodiment of the invention, ESG_Modifier(Ω_t) is calculated in the following way: ESG_Modifier(Ω_t)=(1-Risk_Discount)×(1+Opportunity_Premium), Among them, Risk_Discount quantifies the negative performance of the environmental and social governance dimension into potential financial losses through a preset risk mapping model, and Opportunity_Premium quantifies the positive performance of the environmental and social governance dimension into incremental financial value through a preset opportunity identification model.

[0034] When a company performs well in ESG, Risk_Discount is close to 0 and Opportunity_Premium is positive, resulting in a value > 1, thus generating an ESG premium. When a company performs poorly in ESG, Risk_Discount is close to 1 and Opportunity_Premium is close to 0, resulting in a value < 1, thus generating an ESG discount. This formula realizes the discount of risk and the premium of opportunity, and combines their interaction in a multiplicative form to dynamically adjust the intrinsic value of the company.

[0035] In an optional embodiment, the process of calculating Risk_Discount using the risk mapping model is as follows: 1) Input: Extract negative event data for each dimension of E, S, and G from the dynamic information set.

[0036] In one embodiment, based on a pre-defined ESG negative event feature lexicon and semantic recognition model, negative event data from macro and micro dimensions of E, S, and G are extracted from a dynamic information set, for example: The Environmental (E) dimension includes: macro-level data covering industry pollution emission exceedance rates, keyword matching in environmental protection documents, quantitative index of resource scarcity, data on industries with low carbon emission rankings, and statistics on non-compliance with green energy usage ratios; and micro-level data covering abnormal values ​​in enterprise pollution discharge data, non-compliance with environmental protection equipment investment ratios, data on penalties for violations of waste disposal regulations, green production certification failures, and data on non-compliance in supply chain environmental compliance spot checks.

[0037] The social (S) dimension includes macro-level factors such as the frequency of labor disputes, the industry ranking of employee satisfaction, and data on the failure to meet social responsibility project completion standards; and micro-level factors such as abnormal fluctuations in departmental employee turnover rates, the classification labels of concentrated customer complaints, and records of social responsibility violations by supply chain partners.

[0038] Governance (G) dimension: Macro aspects include compliance scores of legal provisions corresponding to effectiveness and rule of law level, sovereign credit rating data, debt ratio exceeding threshold, etc.; Micro aspects include organizational structure compliance test results corresponding to governance structure defects, related party transaction compliance risk warning data, untimely or incomplete information disclosure records, control loopholes identified by internal audit, risk event logs caused by unreasonable departmental authority settings, etc.

[0039] 2) Quantification: Quantitative indicators of negative event data are mapped to scores using linear or nonlinear interpolation based on benchmark comparison, and qualitative indicators are converted to scores using conditional matching based on a pre-defined rule dictionary. Based on the obtained scores, a loss function is established for each dimension to calculate its expected financial impact.

[0040] Specifically, quantitative indicators are based on linear or nonlinear interpolation methods using benchmark comparisons. Actual values ​​are mapped to a score of 0-100 using corresponding formulas. Examples of quantitative indicators include carbon emissions and water consumption, which are obtained by directly extracting their values.

[0041] Qualitative indicators are based on conditional matching using a rule dictionary. For example, clear if-then rules are developed to transform textual descriptions into specific scores. Scoring: A scoring rule is established: A well-developed environmental management system with third-party certification (such as ISO 14001) and detailed performance disclosure receives 90-100 points; an environmental management system without certification receives 60-80 points for performance disclosure; lack of a systematic management system and inadequate disclosure receive 0-40 points; a major environmental accident or severe administrative penalty results in a -50 point deduction; minor violations or small fines result in a -20 point deduction.

[0042] To make companies in different industries comparable, the absolute values ​​need to be converted to levels relative to the industry average; the indicator score = (company's indicator value - industry worst value) / (industry best value - industry worst value) × 100; after weighted averaging, we get the macro performance score and the micro carbon cost efficiency score.

[0043] Macroeconomic performance score = Σ(Scores of various regulations, compliance, and management indicators × weight) Microscopic carbon cost efficiency score = Σ(scores of various intensity and efficiency indicators × weight) Furthermore, by combining macro and micro quantitative scores, the Escore (negative score in the environmental dimension), Sscore (negative score in the social dimension), and Gscore (negative score in the governance dimension) are calculated respectively. Higher scores indicate more pronounced negative performance in the corresponding dimension. Taking the negative score in the environmental dimension as an example, let's substitute these two dimension scores into the formula to calculate the quantitative score in the E dimension: Escore = α × Macroeconomic Environmental Performance Score + β × Microeconomic Carbon Cost Efficiency Score The weights α and β are allocated through statistical regression (analyzing the correlation between various indicators and financial risk) or meta-learning (dynamically adjusting based on enterprise meta-characteristics (industry, size, etc.) using machine learning).

[0044] 3) Output: Summing up the expected losses of each dimension yields the total expected loss, and then mapping the total loss to the range of 0-1 using a normalization function to obtain Risk_Discount.

[0045] Specifically, based on Escore, Sscore, and Gscore, a loss function is established for each dimension. Combining factors such as industry characteristics and company size, the expected financial losses (e.g., fines, revenue losses, compliance rectification costs, etc.) corresponding to negative performance in each dimension are calculated. The expected losses of the three dimensions are summed to obtain the total expected loss. Then, a normalization function is used to map the total expected loss to the range of 0-1, ultimately yielding the Risk_Discount (ESG risk discount), which is used for subsequent calculation of the ESG_Modifier coefficient.

[0046] In one example, the normalization function is configured with an industry adjustment parameter, so that the same total expected financial loss maps to different Risk_Discount values ​​in industries with different risk characteristics. For example, the following formula calculates: Risk_Discount=min(max( ,0),1) Where Total_Loss is the total expected loss, K is the industry adjustment coefficient (0.001 for high-risk industries and 0.002 for ordinary industries) to control the mapping range, Vasset is the company's latest audited total assets; min / max function: ensures that the result falls within the 0-1 range to avoid interference from extreme values.

[0047] This invention overcomes the limitations of traditional qualitative ESG assessments by comprehensively capturing negative performance across all dimensions through macro and micro-level data coverage. It transforms abstract ESG risks into calculable financial losses, improving the accuracy of risk assessment. Unified data quantification standards and normalization processes ensure that Risk_Discount remains within a fixed range of 0-1, avoiding discrepancies in risk assessments across different companies and dimensions, while also facilitating horizontal comparisons and model application.

[0048] In one embodiment, the process of calculating Opportunity_Premium using the opportunity recognition model is as follows: 1) Input: Extract positive performance data for each dimension (E, S, G) from the dynamic information set; for example: Environment (E): Green revenue percentage, energy conservation and emission reduction benefits, number of green patents; Society (S): Employee satisfaction rating, amount of investment in public welfare, customer satisfaction; Governance (G): Board diversity score, information disclosure rating, and anti-corruption training coverage.

[0049] 2) Quantification: This involves calculating the corresponding financial returns using a payoff function related to the event to assess its incremental value. For example: Green financing returns = (ordinary bond interest rate - green bond interest rate) × bond issuance size, which directly reduces costs and increases corporate value; Talent attraction benefits = (industry average turnover rate - company turnover rate) × per capita recruitment and training cost. The turnover rate × per capita recruitment and training cost is reflected in operating costs and improves human resource efficiency.

[0050] 3) Output: Sum the various financial gains and normalize them relative to the enterprise value (such as total assets or market value) to Opportunity_Premium.

[0051] Specifically, prioritize using the company's latest audited total assets or current market capitalization as the benchmark (denoted as V). You can choose one of the two: choosing total assets is suitable for asset-heavy industries and companies with large market capitalization fluctuations, and can better reflect the stable benchmark of the company's operating scale; choosing market capitalization is suitable for asset-light industries and listed companies, and is more in line with the value assessment logic in the market capitalization management scenario.

[0052] This invention summarizes the financial gains (denoted as Total_Gain) corresponding to all positive ESG events, namely: Total Gain = Green Financing Gain + Talent Attraction Gain + Green Market Access Gain + Subsidy Gain, etc. (Sum of gains from various positive ESG events). The formula is as follows, using a linear normalization method: Opportunity_Premium=Total_Gain / (V×K) Wherein, K is the industry adjustment coefficient, with a value range of 1-5, determined according to industry characteristics: High ESG sensitive industries (such as new energy and environmental protection): K=1-2, amplify the weight of ESG benefits; Medium ESG sensitive industries (such as manufacturing and consumption): K=3, maintain a neutral weight; Low ESG sensitive industries (such as traditional resources): K=4-5, appropriately reduce the weight to avoid overestimating benefits.

[0053] If the calculated result exceeds the 0-1 range, a truncation adjustment is performed: when Opportunity_Premium > 1, it is set to 1 (representing that the positive ESG benefits have reached the upper limit of the enterprise's value carrying capacity); when Opportunity_Premium < 0, it is set to 0 (representing no substantial incremental ESG benefits, avoiding negative interference). This normalization logic considers both differences in enterprise size (calibrated by V) and industry characteristics (adjusted by K), ultimately stabilizing Opportunity_Premium in the 0-1 range. This facilitates the calculation of the ESG_Modifier coefficient in conjunction with Risk_Discount and enables horizontal comparison of the ESG opportunity value between different enterprises.

[0054] 4. Short-term disturbance capture factor (f_sentiment related): Used to capture market dynamics, including investor structure (long-term institutional share, retail investor share), market narrative (intensity of industry concept dissemination), expectation gap (difference between analyst consensus expectation and model intrinsic expectation), and technical indicators (trend following, momentum, trading volume). Trend following indicators include the position relative to its long-term moving average, the value and direction of the MACD (Moving Average Convergence Divergence); the overbought / oversold state reflected by the RSI (Relative Strength Index); and the volume-price relationship and its divergence from price. The model learns the non-linear relationship between these technical indicators and subsequent short-term price fluctuations, thereby dynamically adjusting the output weights of the f_sentiment function. This allows the model to capture early short-term disturbance changes not yet reflected in fundamental information, enhancing the system's sensitivity and timeliness in responding to market dynamics.

[0055] Step S103: Based on the enterprise decision-making data, automatically generate a draft strategy report with a preset directory structure, and provide a human-machine collaborative interface to review and revise the draft report to form the final version to be confirmed.

[0056] Specifically, this step achieves efficient iteration of the report from draft to final version through intelligent generation and human-machine collaboration. In one embodiment, based on enterprise decision-making data, a pre-trained domain-specific large model and a natural language generation module automatically output a structured Word-formatted draft report with an intelligent table of contents. The core content includes: 1. Core Value Analysis Section: Identify the core drivers of the company's current market capitalization (e.g., in financial factors, the continuous increase in ROE contributes 30% to the value increment, and in ESG factors, environmental compliance risks lead to a 5% value discount), value bottlenecks (e.g., in short-term disturbance capture factors, the proportion of institutional investors is less than 15%, which restricts valuation improvement), and causal transmission paths (e.g., increased R&D investment → improved patent conversion rate → extended product life cycle → enhanced future cash flow expectations → market capitalization growth). 2. Decision Recommendations Section: Based on the output of the adaptive value quantification model, propose targeted strategies (such as promoting the implementation of green energy projects, which is expected to reduce the ESG risk discount by 3 percentage points, and optimizing the investor structure, with the goal of introducing 2 QFII institutions before Q3). 3. Implementation Path Section: Initial planning of task breakdown direction, clarification of responsible department matching suggestions (e.g., carbon emission reduction tasks are matched with the Planning Department and Production Department, investor relations optimization tasks are matched with the Investor Relations Department). 4. Data Support Appendix: Includes quantitative data on key factors, charts for capturing and analyzing short-term disturbances, and benchmarking data on ESG performance, ensuring the report is traceable.

[0057] Furthermore, a version-controlled human-computer collaborative interaction interface is provided, with the specific process as follows: 1. Permission allocation: The department responsible for market value management has the permission to edit the initial draft of the report, and the management has the permission to review it. The interface clearly marks the scope of operation for each role. 2. Revision Operations: Editors can make online annotations, add or delete content, and correct data through the interface; all modification traces are automatically recorded. 3. Review and approval process: The revised report is automatically pushed to the management review interface through the system. The management can choose to approve or modify it online. Review comments are fed back to the editors in real time. The one-way review process reduces time costs. 4. Version Management: The system saves the initial draft, revised draft and final draft of the report, supports the review and comparison of historical versions, and forms a complete audit record chain to ensure traceability.

[0058] The above process eliminates the need for manual drafting of framework content. The large model automatically integrates value analysis and data support based on standardized logic, avoiding human error while ensuring a standardized report structure and accurate data. This reduces the initial report draft generation cycle from several days to several hours. Through access control, record keeping, and version management, a standardized review and transfer mechanism is established to avoid information distortion caused by manual transmission, while also meeting compliance requirements and facilitating subsequent auditing and accountability. Furthermore, the structured directory and initial task adaptation lay the foundation for the automatic distribution of subsequent tasks, ensuring that the decision recommendations in the report are accurately implemented and avoiding the problem of a disconnect between reporting and execution.

[0059] Step S104: Respond to the final version management confirmation instruction, parse the structured content of the final version, and generate corresponding executable tasks according to the predefined responsibility mapping relationship and assign them to the corresponding execution end.

[0060] Specifically, this step achieves precise strategy penetration through structured parsing and intelligent assignment, as shown in the following example: 1. Structured content analysis 1) After management confirms the final version report, the parsing engine is automatically triggered. Based on the document structure parsing algorithm of natural language processing, the core task information is extracted: Key information is parsed from the "Core Value Analysis - ESG Optimization" section: The main task is "reducing the risk discount of ESG environmental dimension", the action requirement is "implementing 2 green energy projects", the quantitative indicator is "completing the wind power project filing before Q4 and achieving grid connection and power generation in the first half of next year", and the expected effect is "reducing the ESG risk discount by 3 percentage points". 2) Key information can be extracted from the "Investor Structure Optimization" section: The main task is to "increase the proportion of institutional investors", the action requirement is to "connect with QFII institutions and complete reverse roadshows", the quantitative indicator is "connect with 3 target institutions and successfully complete reverse roadshows for 2 institutions before Q3", and the time node is "to be completed before September 30". 3) During the parsing process, the data support in the report appendix (such as feasibility analysis of green energy projects and list of target QFII institutions) will be automatically associated and distributed as task attachments.

[0061] 2. Responsibility Mapping and Task Generation The system has a built-in "department-responsibility mapping library" (built based on the enterprise's organizational structure and departmental work assignments), which uses an entity linking algorithm to achieve precise matching between tasks and departments. 1) The above-mentioned task of “implementation of green energy projects” is matched with “planning department (project coordination) + production department (technology implementation) + finance department (financial support)”, generating 3 sub-tasks: the planning department is responsible for project filing and progress control, the production department is responsible for determining the technical route and coordinating construction, and the finance department is responsible for project budget preparation and fund allocation. 2) The "QFII Institution Matching" task is matched with "Investor Affairs Department (leading) + Legal Department (compliance review)" and generates 2 sub-tasks: the Investor Affairs Department is responsible for institution matching, roadshow plan formulation and execution, and the Legal Department is responsible for the compliance review of roadshow materials and agreement drafting; 3) Each subtask automatically generates a standardized format, including task name, responsible department, person in charge, quantitative target, time node, data feedback requirements, related report chapters and attachments, to ensure that the task is executable and traceable.

[0062] 3. Example of task assignment and execution terminal reception 1) The system automatically pushes subtasks to the corresponding department's OA system to-do module via API interface, and sends SMS / APP reminders to the person in charge, specifying the task priority (e.g., "QFII docking task" is marked as high priority). 2) After logging into the OA system, the head of the planning department can view task details, download related attachments, update task progress in real time (such as "Project filing in progress" or "Technical briefing completed"), and provide feedback on phased results (such as filing receipts and technical solution documents). 3) For cross-departmental collaborative tasks (such as green energy projects), the market value management department monitors the progress of each sub-task in real time through a task visualization Gantt chart. When a sub-task is found to be lagging behind, a reminder rule is automatically triggered. The task is created through an interface integrated with the OA system, and the sub-task leader is sent a reminder message by calling the SMS gateway API simultaneously, thus avoiding collaboration gaps.

[0063] Step S105: Collect task execution result feedback data and synchronize it to the market value digital twin and the adaptive value quantification model for evaluating the execution effect and optimizing the adaptive value quantification model.

[0064] Specifically, the data collection scope covers task completion data (such as whether quantitative targets were achieved on schedule and completion quality scores), process data (such as resource input during task execution and cross-departmental collaboration records), result-related data (such as carbon emission data after the implementation of ESG projects and changes in institutional shareholding ratios after investor structure optimization), and abnormal feedback data (such as explanations of reasons for task incompleteness and new risk points that emerged during execution). The data collection method supports a combination of automatic collection and manual feedback through multiple channels. Automatic collection is achieved by connecting to OA systems, business systems, IoT terminals, etc., to obtain structured data such as task progress and achievement of quantitative indicators in real time. Manual feedback is achieved through the system's built-in feedback interface, allowing the executing departments to provide feedback on unstructured results (such as reports and supporting documents) and abnormal explanations to ensure data comprehensiveness.

[0065] Furthermore, the collected data will be synchronized to the market capitalization digital twin. The process involves updating the collected execution result data to the twin according to the corresponding factors. For example, the actual emission reduction of carbon emission reduction tasks will be synchronized to the ESG factor, and the institutional docking results will be synchronized to the short-term disturbance capture factor. This will enable the twin to reflect the real status of the enterprise after execution in real time and maintain dynamic consistency with the physical enterprise.

[0066] The process of synchronizing to the adaptive value quantification model involves using execution result data as input samples for model optimization. This includes market capitalization changes after task completion and actual factor impact data (such as the actual changes in the ESG_Modifier coefficient after ESG projects are implemented), providing real feedback data for model parameter adjustments. The adaptive value quantification model uses this feedback data to fine-tune factor weight configurations. For example, if the actual value contribution of a certain type of ESG project is found to be higher than the initial model prediction, the weight of the corresponding ESG factor is increased. The model is then retrained using hierarchical grouped LASSO regression technology to correct the correlation between factors, making the model more closely reflect actual corporate operations and market capitalization changes.

[0067] The method provided in this invention breaks down data silos and ensures the authenticity and reliability of underlying data by collecting multi-source heterogeneous data, verifying consistency, and dynamically constructing digital twins throughout the entire process of decision data-strategy report-task execution. This provides accurate and dynamic market value assessment basis for decision-making and significantly reduces decision bias caused by data distortion. Through an adaptive value quantification model that integrates the weights of multi-dimensional factors and a human-machine collaborative report review and revision mechanism, it can output initial drafts of strategy reports tailored to enterprise needs and optimize report quality through human intervention, improving the scientific rigor and feasibility of strategy recommendations. It transforms qualitative factor characteristics into quantitative operational thresholds, avoiding subjectivity in decision execution. Automatic task assignment based on responsibility mapping achieves seamless transformation of strategy reports into executable tasks, shortening the cycle from decision to implementation, enhancing the accuracy and timeliness of strategy execution, and helping enterprises respond efficiently to market value fluctuations, improving the intelligence and refinement of market value management. Through the above closed-loop optimization mechanism, successful execution experience is converted into model parameters, avoiding repeated trial and error with similar strategies. Simultaneously, it quickly identifies and adjusts ineffective execution behaviors, improving the overall efficiency and return on investment of the market value digital twin.

[0068] In one embodiment, the present invention provides an intelligent enterprise decision-making execution method, such as... Figure 2 As shown, it includes: Step S201: Continuously collect and integrate heterogeneous data from multiple sources inside and outside the target enterprise, and process it through a multi-source consistency verification algorithm to construct and dynamically update the enterprise's market value digital twin.

[0069] In this embodiment of the invention, a cloud-edge-device collaborative data acquisition architecture is used. Step S1 specifically includes the following steps: S2011 uses IoT terminals, edge computing devices, and base stations with integrated communication and sensing capabilities to collect internal business data, external industry data, and enterprise operational behavior data in real time from IoT terminals.

[0070] S2012, cleans, verifies consistency, and extracts features from the raw data collected at the edge computing device node; S2013 uploads the processed standardized data to the cloud to achieve real-time data synchronization and lossless transmission, and dynamically updates the market value digital twin.

[0071] In one optional embodiment, the edge device achieves real-time acquisition and lossless transmission of multi-source data through API interfaces, RPA robots, and 6G communication modules. The edge device leverages the inherent AI capabilities of the 6G network to deploy lightweight data cleaning, verification, and feature extraction models to preprocess the raw data in real time, filtering invalid information and correcting data deviations. The cloud receives standardized data uploaded from edge nodes, performs fusion modeling and dynamic updates, ensuring that the market capitalization digital twin is synchronized with the actual operational status of the enterprise in real time.

[0072] The above process standardizes the format of the collected multi-source data, including data alignment, feature extraction, and contradiction verification, to ensure data consistency and reliability. It establishes an incremental update mechanism that automatically updates the corresponding factor information in the digital twin as new data continues to flow in, without requiring a full recalculation, thus ensuring the timeliness and lightweight operation of the twin. It integrates multi-dimensional factor data correlation analysis to construct a full-link mapping from internal business data to external market performance, thereby achieving a panoramic presentation of the enterprise's operational status.

[0073] Step S202: By deploying an adaptive value quantification model in the market capitalization digital twin, multi-dimensional factors are set based on multi-source heterogeneous data, and adaptive weight fusion calculation is performed to generate data for enterprise decision-making.

[0074] The adaptive value quantification model in this embodiment of the invention is described in step S102 above, and will not be repeated here.

[0075] In an optional embodiment, the adaptive value quantification model provided by this invention employs hierarchical grouped LASSO regression technology for training or updating, including: a1. Obtain the dataset for model training. The dataset includes massive multi-dimensional factor data and corresponding market value data of sample companies. The multi-dimensional factors include at least financial factors, environmental and social governance factors, systemic importance factors, and short-term disturbance capture factors belonging to different business levels. a2, Construct a loss function that includes a hierarchical grouping LASSO penalty term, the loss function being:

[0076] Where N is the number of samples, y i Let β be the sample market capitalization, β0 be the intercept term, and β j Let x be the coefficient of the j-th factor. ij Here, λ is the factor value, λ is the hyperparameter controlling the severity of the penalty, G is the total number of factor groups, and p... g Let ||β(g)||2 be the number of factors in the g-th group, and ||β(g)||2 be the L2 norm of the factor coefficients in the g-th group. a3, with the goal of minimizing the loss function, matches or quickly generates initial factor weight configurations from a preset prior knowledge base, trains the adaptive value quantification model, and during the training process, retains or compresses the factor coefficients to zero as a whole in the factor group. a4, an ensemble learning algorithm combining tree models, suppresses multicollinearity through intra-group and inter-group constraints, and fine-tunes the loss function with group regularization during training.

[0077] This invention employs hierarchical grouping LASSO regression technology for training and updating. Factor groups are divided based on business levels, enabling coefficient selection at the group level. This retains core driving factors and eliminates redundant factors, improving the model's interpretability and simplicity. Initial weights are configured using a prior knowledge base, significantly shortening the model training cycle and reducing data dependence. A tree-based ensemble learning algorithm is introduced, effectively suppressing multicollinearity among multi-dimensional factors through intra-group and inter-group constraints, improving model prediction accuracy and stability. Its hierarchical grouping training logic adapts to the characteristics of multi-dimensional factors in market capitalization management, making model weights more closely aligned with the actual operational status of enterprises, providing reliable support for accurate market capitalization quantification.

[0078] In one embodiment, the ensemble learning algorithm for tree models combines multiple weak learners (such as decision trees) and obtains the final result through voting or averaging, thereby achieving superior performance compared to a single model. Gradient boosting decision trees are a representative algorithm. Taking gradient boosting machines as an example, the process involves iteratively constructing multiple decision trees, where the learning objective of each tree is the residual of all previous tree combinations, expressed by the following formula: F m (x)=F m-1 (x)+ν×γ m ×h m (x) Among them, F m (x) is the ensemble model F after m iterations. m-1 (x) is the model after the first m-1 iterations; h m(x) is the weak learner trained in the m-th round; γm is the optimal weight of the m-th weak learner; ν is the learning rate, an important hyperparameter that controls the contribution of each tree to the final model. By reducing the weights at each step, overfitting is prevented, and more stable convergence is achieved.

[0079] The training method provided in this invention does not penalize each coefficient individually like in ordinary LASSO, but instead penalizes the coefficients of an entire group together. During optimization, the algorithm tends to either retain all coefficients in the group or compress them all to zero. If the profitability quality dimension is not important, then the return on net assets and profit margin coefficients under it should be removed simultaneously.

[0080] In one embodiment, a standardized micro dataset is provided, and the loss function is set to λ=0.5. The corresponding values ​​of each factor for companies A, B, and C are shown in Table 1.

[0081] Table 1

[0082] Step 1: Initial Coefficient Estimation Assume that the initial coefficient estimates given by the unregularized OLS regression are: β1 (ROE) = 0.5, β2 (profit margin) = 0.6, β3 (revenue growth) = 0.8, β4 (R&D growth) = 0.7.

[0083] Step 2: Calculate the L2 norm of each group of coefficients. 1. Formula: For a vector v = [v1, v2, ..., vn], its L2 norm is calculated as: ||v||² = This involves summing the squares of each element in the vector and then taking the square root of the sum.

[0084] 2. Grouping logic of coefficients: The regression coefficients are divided into two groups: Group 1 (G1): β1 (ROE) = 0.5, β2 (profit margin) = 0.6 (profitability factor coefficients); Group 2 (G2): β3 (revenue growth) = 0.8, β4 (R&D growth) = 0.7 (growth factor coefficients).

[0085] 3. Substituting the vector βG1 = (β1, β2) into the definition of the L2 norm, we get: |βG1||2 = Substituting the vector βG2 = (β3, β4) into the definition of the L2 norm, we get: |βG2||2 = .

[0086] The L2 norm of group 1 (G1): |βG1||2= ≈0.78; The L2 norm of group 2 (G2): |βG2||2= ≈1.06.

[0087] Step 3: Apply grouped LASSO penalties Algorithms for solving grouped LASSO (such as coordinate descent) iteratively shrink each group of coefficients. The shrinkage rule is: If the L2 norm of a set is less than a certain threshold (related to λ), the coefficients of the entire set are set to zero. Otherwise, the coefficients of the entire set are scaled down proportionally.

[0088] For group 1 (G1): due to its relatively small L2 norm (0.78), it may be eliminated entirely under the penalty of λ=0.5. The final coefficients are compressed to β1=0, β2=0.

[0089] For group 2 (G2): Due to its larger L2 norm (1.06), it shows a stronger joint explanatory power for market capitalization, so the entire group is retained, but the coefficients are reduced proportionally. Assuming the final estimates are β3=0.65, β4=0.55, Table 2 compares the differences between ordinary least squares (OLS) and grouped LASSO regression in estimating factor coefficients for the value quantification model. OLS estimates of factor coefficients do not consider collinearity interference between factors. Grouped LASSO, based on business logic, divides factors into "profitability quality" and "growth quality" groups, and performs coefficient screening and compression on a group basis, demonstrating the advantages of factor collinearity and optimized model stability. Table 2: Example comparison of grouped LASSO and OLS estimation results.

[0090] Table 2

[0091] Step S203: Visualize the output of the adaptive value quantification model to display the causal relationship path between factors using the constraint-based causal discovery algorithm.

[0092] Specifically, this embodiment of the invention uses a dashboard to visualize the output results. Through a constraint-based causal discovery algorithm, abstract factor relationships are transformed into intuitive causal paths. When analyzing massive amounts of historical data, this algorithm not only relies on statistical relationships but also introduces hard constraint rules extracted from domain knowledge to accurately distinguish the causal relationships and correlations between factors. In the form of an intuitive path diagram, it displays the dynamic causal transmission logic between multiple dimensions of factors such as financial, systemic importance, ESG, and short-term disturbance capture, as well as between factors and corporate market value. At the same time, it marks key quantitative impact coefficients and supports path drill-down and scenario simulation operations.

[0093] In one embodiment, an improved Peter-Clark algorithm is employed, a causal discovery method based on conditional independence testing, incorporating three types of domain-specific constraints: 1. Temporal Priority Constraint: Set timestamp association rules for causal events. For example, the ESG performance in the first quarter can only affect the market capitalization in the second quarter, not the other way around. This is the gold standard for causal inference.

[0094] 2. Value Stream Constraint: Based on financial theory, the causal hierarchy of characteristic variables is defined, with financial performance indicators as upstream causal variables and short-term disturbance indicators as downstream outcome variables. For example, financial performance can influence short-term disturbances, but short-term disturbances usually do not directly affect financial performance in the reverse.

[0095] 3. Business Logic Constraints: Based on management practices, establish hierarchical transmission path rules for the impact of ESG factors on market capitalization. For example, improvements in governance structure should first affect operational risks, then transmit to financial costs, and finally affect market capitalization. Direct, leapfrog effects on market capitalization due to governance improvements are prohibited.

[0096] In one embodiment, the algorithm input is panel data. For example, it includes N companies and T time points (e.g., the past 20 quarters), with each time point containing multiple variables: ESG variables: carbon emissions, employee turnover rate, board diversity score, etc.; Financial variables: ROE, gross profit margin, debt ratio, etc. Market variables: analyst attention, institutional shareholding ratio, financing costs, etc. Outcome variables: market capitalization, Tobin's Q; Algorithm output: A directed acyclic graph where nodes are variables, directed edges represent causal relationships, and edge weights represent the strength of causal effects.

[0097] The dashboard provided in this embodiment of the invention is not a static chart, but an interactive exploration system, with a functional flow diagram, for example: flowchartLRA [“Global Causal Network Graph”] (Overview Page)"]-->B["Single Path Deepening Analysis"] (Click the path to expand)”]B-->C[“Node Influence Measurement” (Click on the node to view)]B-->D["Scenario Simulation Panel" (Right-click node operation)]C-->E["Driver Factor Analysis" (Positive impact)”]C-->F[“Bottleneck factor analysis” (Negative impact)"]D-->G["Policy Simulator"] [Predicting Intervention Effects] In one example, a force-directed graph can be visualized using a global causal network diagram (overview page), where: Nodes: Different shapes represent different variables (e.g., circles represent ESG factors, squares represent financial factors, and diamonds represent short-term disturbance factors).

[0098] Edge: Directed arrow, color indicates the direction of the effect (green for positive, red for negative), thickness indicates the intensity of the causal effect.

[0099] Default view: Only shows the main transmission path from ESG factors to market capitalization, avoiding excessive clutter.

[0100] In practical applications, clicking any node (such as financing costs) will highlight it: 1. In-degree influence: What factors affect it (upstream causes); 2. Impact of output degree: What factors (downstream results) does it affect? 3. Total Effect: The overall impact of this node on market capitalization through all paths. Core Driving Factor: The node with the largest total effect is the focus of decision-making. 4. Key bottleneck factors: Nodes with negative effects and large absolute values ​​are the shortcomings that urgently need to be improved.

[0101] In another example, when clicking on a path (such as employee satisfaction -> productivity -> gross profit margin -> market capitalization), the side panel displays: 1. Path strength: The overall causal effect value (standardized coefficient) of this path; 2. Mediation effect test results: Statistical significance (p-value), proving that the path is indeed valid; 3. Historical trend: The strength of this path has changed over the past 8 quarters.

[0102] In an optional embodiment, the dashboard can perform scenario simulation operations: the user can trigger any ESG factor node (such as carbon emission intensity) to pop up a strategy simulator by clicking the corresponding function button. The simulation process is as follows: 1. User input: Set an intervention strategy, such as reducing carbon intensity by 10% in the coming year; 2. Model Calculation: Based on the estimated causal effects, the system calculates how this change will affect its direct downstream variables (such as environmental compliance costs). The effect then spreads along the causal network, progressively affecting intermediate variables (such as operating costs and net profit). Finally, the expected impact on market capitalization and the confidence interval are calculated.

[0103] 3. Results Presentation: The projected increase in market capitalization is X billion yuan (95% confidence interval: YZ billion yuan). The process of market capitalization transmission is also graphically displayed, highlighting the key links with the greatest impact.

[0104] This step lowers the barrier to understanding complex factor relationships by intuitively presenting the driving logic, helping to quickly grasp the core impact path of market capitalization; based on domain-specific intervention constraints (such as defining the positive and negative effects of variables), it standardizes the variable selection and path analysis process for causal inference, ensuring the reliability of causal inference, avoiding false associations that mislead decision-making, improving the scientific nature of decision-making, supporting scenario simulation and path drilling, providing intuitive reference for decision-making, and helping to accurately predict the effect of strategies.

[0105] Step S204: Based on the enterprise decision-making data, a preliminary draft of a strategy report with a preset directory structure is automatically generated, and a human-machine collaborative interface is provided to review and revise the preliminary draft of the report to form a final version to be confirmed; see step S103 above for details, which will not be repeated here.

[0106] Step S205: Respond to the final version management confirmation instruction, parse the structured content of the final version, and generate corresponding executable tasks according to the predefined responsibility mapping relationship and assign them to the corresponding execution terminals; see step S104 above for details, and will not be repeated here.

[0107] Step S206 involves collecting task execution result feedback data and synchronizing it to the market capitalization digital twin and the adaptive value quantification model to evaluate the execution effect and optimize the adaptive value quantification model. See step S105 above for details, which will not be repeated here.

[0108] The method provided in this embodiment of the invention further includes Sx: a public opinion risk intelligent perception and linkage response step, which is executed independently of or in parallel with steps S1-S5, including: Sx1: Analyze external public opinion data through natural language processing technology to identify negative events and their subjects; This embodiment of the invention relies on natural language processing technology to realize the automated analysis of external public opinion, replacing the manual screening mode, which can quickly identify negative events and related subjects, greatly shorten the risk perception cycle, and avoid the omissions and misjudgments that may occur in manual screening.

[0109] Sx2: Based on the relationship between the subject and the target company, events are automatically classified into different risk levels; this frees the determination of public opinion risk levels from the limitations of subjective experience, ensures the consistency and rationality of risk classification, and provides a scientific basis for subsequent differentiated responses.

[0110] Sx3: Based on the risk level, automatically trigger the corresponding level of internal early warning, and generate and distribute special self-inspection, strategy assessment, or emergency plan activation tasks matching the risk level. This embodiment of the invention automatically triggers the corresponding early warning mechanism based on the risk level, and simultaneously generates and distributes special self-inspection, strategy assessment, or emergency plan activation tasks. This avoids problems such as delayed response or mismatch between countermeasures and risk levels, promptly identifies potential risks, adjusts strategies, or activates emergency plans, effectively reduces the impact of public opinion events on corporate reputation, market value, and business development, and improves risk management efficiency.

[0111] In one optional embodiment, this invention classifies public opinion events into four risk levels based on their relevance to the enterprise, their scope of impact, and their urgency, corresponding to differentiated intelligent response strategies to achieve closed-loop management from risk perception to response execution. Specifically: 1. Level 1: Direct risks to the main entity (highest priority) The risk source is a public opinion event that directly targets the company, including key elements such as the company name, brand logo, and core products / services. Its core characteristic is that the event is 100% related to the company, the facts are clear and unambiguous, and it may directly trigger serious consequences such as inquiries from regulatory authorities, significant stock price fluctuations, and customer churn.

[0112] The system response strategy focuses on "emergency response and comprehensive self-inspection": First, it triggers the highest-level alarm, with the monitoring dashboard turning red across the entire screen and a pop-up notification appearing. Simultaneously, it pushes alarm information to key personnel through multiple channels, including phone calls, SMS, and app push notifications. Second, it automatically activates the emergency plan, directly linking and pushing the corresponding crisis response plan to the emergency response team, skipping the intermediate assessment stage. Third, it generates a comprehensive self-inspection work order with the highest standards, covering all relevant departments and requiring preliminary response results to be provided within 2 hours. Fourth, it synchronizes alarm information, emergency plans, and self-inspection work orders to top management or spokespersons in real time, ensuring that senior management is informed of the situation immediately.

[0113] 2. Level Two: Mirror Investor Risk (High Priority) The risk source is that the subject of the public opinion event is our company's direct competitor, leading company in the same industry, or benchmark company. The core characteristic is that the event is highly replicable and referential, which can easily trigger a crisis of trust in the entire industry, or lead to cross-inspection by regulatory authorities, indirectly affecting our company's operations and reputation.

[0114] The system's response strategy focuses on "prevention and early warning, and targeted investigation": First, it issues high-level warnings, prominently displaying orange indicators on the monitoring dashboard, and simultaneously sending strong in-app reminders to relevant departments such as risk control, compliance, and public relations; second, it automatically generates special self-inspection lists based on event tags (such as financial fraud, data leakage, product security, etc.) and accurately assigns them to the corresponding business departments; third, it sets task completion deadlines for public opinion response ranging from 4 to 24 to 72 hours, automatically tracks progress, and initiates an escalation and reminder mechanism if the task is not completed by the deadline, requiring relevant departments to submit self-inspection reports and targeted rectification measures.

[0115] 3. Level 3: Systemic Risks in the Industry / Market (Medium Priority) Risk sources are systemic negative events that affect the entire industry or market, such as the exposure of industry-wide scandals or significant fluctuations in the upstream of the supply chain. Their core characteristic is that their impact is broad but indirect, not targeting any single company, yet they alter the overall operating environment and market rules of the industry, potentially impacting the company's medium- to long-term development.

[0116] The system's response strategy focuses on "strategic assessment and risk warning": First, it generates a "Risk Analysis Briefing" to extract the core points of the event, conduct in-depth analysis of its medium- and long-term impact on the company's business, and remind management to pay attention to changes in industry trends; second, it triggers a strategic review process, prompting core departments such as strategy, investment, and R&D to conduct stress tests and impact assessments on existing strategic plans and ongoing projects, and generates a special assessment report; third, it provides decision support by generating a decision matrix that includes options such as "accelerate progress, maintain the status quo, strategic transformation, and contraction and adjustment," providing a reference for management to discuss and formulate response strategies.

[0117] 4. Level Four: Risks Related to Parties / Ecosystem (Low to Medium Priority) The sources of risk are the company's key suppliers, major customers, investment targets, and important partners (such as auditing firms and service providers) involved in the public opinion events. A key characteristic is that the risk is transmitted through business cooperation, financial transactions, or reputational connections, requiring a thorough assessment of its specific and substantive impact on the company.

[0118] The system response strategy focuses on "related assessment and risk isolation": First, it triggers a related assessment warning, automatically identifies the related parties of the event, pops up a risk warning, and initiates the related impact assessment process; second, it precisely triggers investigation tasks based on the roles of the related parties. For example, for core suppliers, it requires the supply chain department to assess supply security and develop backup plans, and for investment targets, it requires the investment department to assess asset impairment risks; third, it provides reputation management support, prompting the public relations and brand departments to assess whether a statement needs to be issued for reputation isolation, and providing statement template suggestions.

[0119] This invention also provides an intelligent enterprise decision-making and execution system, such as... Figure 3As shown, it includes: The multi-source heterogeneous data acquisition module 31 is used to continuously collect and integrate multi-source heterogeneous data from the target enterprise, and after processing by the multi-source consistency verification algorithm, to build and dynamically update the enterprise's market value digital twin. The enterprise decision-making basis data generation module 32 is used to generate enterprise decision-making basis data by setting multi-dimensional factors for multi-source heterogeneous data and performing adaptive weight fusion calculation through the adaptive value quantification model deployed in the market value digital twin; The report generation module 33 is used to automatically generate a draft strategy report with a preset directory structure based on enterprise decision-making data, and provides a human-machine collaborative interface to review and revise the draft report to form a final version to be confirmed; The task generation and dispatch module 34 is used to respond to the management confirmation instruction of the final version, parse the structured content of the final version, and generate corresponding executable tasks according to the predefined responsibility mapping relationship and dispatch them to the corresponding execution end. The feedback and optimization module 35 is used to collect task execution result feedback data and synchronize it to the market value digital twin and the adaptive value quantification model, so as to evaluate the execution effect and optimize the adaptive value quantification model.

[0120] In an optional embodiment, the multi-source heterogeneous data acquisition module 31 includes: The raw data acquisition unit is used to collect enterprise internal business data, external industry data, and enterprise operational behavior data in real time from IoT terminals, edge computing devices, and base stations with integrated communication and sensing capabilities. The preprocessing unit is used to clean, verify consistency, and extract features from the raw data collected at the edge computing device nodes. The data transmission unit is used to upload the processed standardized data to the cloud, enabling real-time data synchronization and lossless transmission, and dynamically updating the market value digital twin.

[0121] In one optional embodiment, the enterprise decision-making data generation module 32 includes core multi-dimensional factors such as financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors; the adaptive value quantification model uses the following formula for value quantification: MV_t=[CFA_Base(Ω_t)×SCI_Modifier(Ω_t)]×ESG_Modifier(Ω_t)+f_sentiment(Investor_Structure_t,Narrative_t,Expectation_Gap_t,Technicals_t) Where MV_t represents the market value of the enterprise at time t, Ω_t represents the dynamic information set at time t, CFA_Base(Ω_t) is a financial factor that represents the basic financial value dynamically calculated based on the information set Ω_t at time t, and SCI_Modifier(Ω_t) is a system importance factor, which is an adjustment coefficient determined based on the assessment of the enterprise's system importance in the industry chain and industry. ESG_Modifier(Ω_t) is an environmental and social governance factor, which represents the risk and opportunity adjustment coefficient based on the quantitative calculation of a company's environmental, social and governance performance. f_sentiment is a short-term disturbance capture factor, representing a non-linear function that captures short-term disturbances. The input variables of the f_sentiment function include: Investor_Structure_t, representing investor structure data; Narrative_t, representing market narrative quantitative data; Expectation_Gap_t, representing the difference between consensus expectations and the intrinsic expectations calculated by the model; and Technicals_t, representing a technical indicator feature vector that includes trend following, momentum, and volume indicators.

[0122] In an optional embodiment, the ESG_Modifier(Ω_t) is calculated in the following manner: ESG_Modifier(Ω_t)=(1-Risk_Discount)×(1+Opportunity_Premium), Among them, Risk_Discount quantifies the negative performance of the environmental and social governance dimension into potential financial losses through a preset risk mapping model, and Opportunity_Premium quantifies the positive performance of the environmental and social governance dimension into incremental financial value through a preset opportunity identification model.

[0123] In an optional embodiment, the process of quantifying negative performance in the environmental and social governance dimension into potential financial losses through a preset risk mapping model includes: Negative event data from three dimensions—environment, society, and governance—are extracted from dynamic information sets. Quantitative indicators of negative event data are mapped to scores using linear or nonlinear interpolation based on benchmark comparison, while qualitative indicators are converted to scores using condition matching based on a pre-defined rule dictionary. Based on the obtained scores, a loss function is established for each dimension to calculate its expected financial impact; The expected loss of each dimension is summed to obtain the total expected loss, and the total loss is mapped to the range of 0-1 using a normalization function to obtain Risk_Discount.

[0124] In an optional embodiment, quantifying the positive performance of ESG factors into incremental financial value through a preset opportunity identification model includes: Extracting positive event data related to ESG factors from dynamic information sets; Calculate the corresponding financial returns using an event-related payoff function; The various financial gains are summed up and normalized relative to the enterprise value to obtain Opportunity_Premium.

[0125] The adaptive value quantification model is trained or updated using hierarchical grouped LASSO regression, including: Obtain a dataset for model training. The dataset includes massive multi-dimensional factor data and corresponding market capitalization data of sample companies. The multi-dimensional factors include at least financial factors, environmental and social governance factors, systemic importance factors, and short-term disturbance capture factors belonging to different business logic levels. Based on the business logic hierarchy of the multi-dimensional factors, a hierarchical grouping structure is predefined, wherein at least multiple factors belonging to the same core business dimension are divided into the same factor group. Construct a loss function that includes a hierarchical grouping LASSO penalty term, wherein the loss function is:

[0126] Where N is the number of samples, y i Let β be the sample market capitalization, β0 be the intercept term, and β j Let x be the coefficient of the j-th factor. ij Here, λ is the factor value, λ is the hyperparameter controlling the severity of the penalty, G is the total number of factor groups, and p... g Let ||β(g)||2 be the number of factors in the g-th group, and ||β(g)||2 be the L2 norm of the factor coefficients in the g-th group. With the goal of minimizing the loss function, an initial factor weight configuration is matched or quickly generated from a preset prior knowledge base to train the adaptive value quantification model. During the training process, the factor coefficients are retained or compressed to zero as a whole, with the factor group as the unit. An ensemble learning algorithm combining tree models is used to suppress multicollinearity through intra-group and inter-group constraints, and fine-tuning is performed during training using a loss function with group regularization terms.

[0127] In an optional embodiment, it further includes an ESG greenwashing risk identification module, used to identify ESG greenwashing risks in the target company's internal data, specifically including: The quantitative claim verification unit is used to extract quantitative data from ESG claims and compare them with the original data in the internal management system. For example, it can extract quantitative claims (such as "30% emission reduction") from publicly available corporate texts and automatically query the corresponding data in the internal database. It uses rule matching and numerical comparison algorithms to calculate the relative deviation |claim value - actual value| / actual value. If the deviation exceeds the threshold (such as 10%), it will trigger a "data exaggeration" warning.

[0128] The qualitative claim analysis unit is used to perform semantic analysis on vague qualitative statements using natural language processing technology and evaluate their credibility by comparing them with third-party industry benchmark data. For example, based on a knowledge graph semantic reasoning algorithm, when the text contains vague statements such as "industry leader" or "outstanding contribution", the system compares it with the benchmark values ​​provided by relevant third parties. If the company's actual performance indicators are not in the top 20% of the industry (configurable), the system will prompt "claims lack benchmark support".

[0129] The consistency tracking unit is used to build a database of corporate ESG commitments, comparing new statements with historical commitments to check for contradictions or failures to meet those commitments. For example, when a company releases a new statement, it is automatically compared with historical commitments to check for any behavior that fails to meet commitments but still promotes future goals, or any contradictions between previous and current statements.

[0130] In an optional embodiment, it further includes: a visualization module for converting the output of the adaptive value quantification model into a visual dashboard, and using a constraint-based causal discovery algorithm to display the causal relationship paths between the factors.

[0131] In an optional embodiment, the device further includes a public opinion risk intelligent perception and linkage response module, which includes: The public opinion data analysis unit is used to analyze external public opinion data through natural language processing technology to identify negative events and their perpetrators. The risk level identification unit is used to automatically classify the negative event into different risk levels based on the relationship between the subject and the target enterprise; The risk emergency activation unit is used to automatically trigger the corresponding level of internal early warning based on the risk level, and generate and distribute special self-inspection, strategy evaluation or emergency plan activation tasks that match the risk level.

[0132] The intelligent enterprise decision-making execution system provided in this embodiment of the invention can execute the intelligent enterprise decision-making execution method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0133] Figure 4This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0134] The following is a detailed reference. Figure 4 This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0135] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0136] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the intelligent enterprise decision execution method of the embodiments of the present invention.

[0137] Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0138] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the intelligent enterprise decision-making execution method shown in the above embodiments is implemented.

[0139] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0140] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An intelligent enterprise decision-making and execution method, characterized in that, include: Continuously collect and integrate heterogeneous data from multiple sources inside and outside the target company, and after processing through a multi-source consistency verification algorithm, construct and dynamically update the company's market value digital twin; Based on multi-source heterogeneous data, preset multi-dimensional factors are set, and data corresponding to each dimension factor is extracted in a preset order. The data corresponding to each dimension factor is input into the adaptive value quantification model deployed in the market value digital twin. The adaptive value quantification model is used to fuse and calculate the input data according to the weights corresponding to each dimension factor to generate data for enterprise decision-making. Based on enterprise decision-making data, a preliminary draft of a strategy report with a preset directory structure is automatically generated, and a human-machine collaborative interface is provided to review and revise the preliminary draft of the report, forming a final version to be confirmed; In response to the management confirmation instruction of the final version, the structured content of the final version is parsed, and corresponding executable tasks are generated and assigned to the corresponding execution terminals according to the predefined responsibility mapping relationship; Collect task execution result feedback data and synchronize it to the market value digital twin and the adaptive value quantification model, so as to evaluate the execution effect based on the execution result feedback data and fine-tune the weights of each dimension factor in the adaptive value quantification model; The multi-dimensional factors include financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors, and the data corresponding to the financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors are extracted sequentially; the adaptive value quantification model uses the following formula for value quantification: Where MV_t represents the market value of the enterprise at time t, Ω_t represents the dynamic information set at time t, CFA_Base(Ω_t) is a financial factor, representing the basic financial value dynamically calculated based on the information set Ω_t at time t, and SCI_Modifier(Ω_t) is a system importance factor, representing the adjustment coefficient determined based on the assessment of the enterprise's system importance in the industry chain and industry. ESG_Modifier(Ω_t) is an environmental and social governance factor, which represents the risk and opportunity adjustment coefficient based on the quantitative calculation of a company's environmental, social and governance performance. f_sentiment(Investor_Structure_t, Narrative_t, Expectation_Gap_t, Technicals_t) represents the short-term disturbance quantification value determined by the adaptive value quantification model based on investor structure data, the industry concept propagation strength corresponding to the market narrative, the difference between the analysis consensus expectation and the intrinsic expectation calculated by the model, and the technical indicator feature vector. Here, Investor_Structure_t represents investor structure data; Narrative_t represents the industry concept propagation strength corresponding to the market narrative; Expectation_Gap_t represents the difference between the analysis consensus expectation and the intrinsic expectation calculated by the model; and Technicals_t represents the technical indicator feature vector including trend following, momentum, and volume indicators. The adaptive value quantification model learns the nonlinear relationship between the technical indicator feature vector and subsequent short-term price fluctuations, and dynamically adjusts the output weights of f_sentiment based on the learning results.

2. The method according to claim 1, characterized in that, The continuous collection and integration of heterogeneous data from multiple sources both inside and outside the target enterprise, and the processing of this data through a multi-source consistency verification algorithm to construct and dynamically update the enterprise's market capitalization digital twin, includes: By using IoT terminals, edge computing devices, and base stations with integrated communication and sensing capabilities, internal business data, external industry data, and enterprise operational behavior data are acquired; wherein, the enterprise operational behavior data is collected in real time by the IoT terminals. The raw data collected at the edge computing device nodes is cleaned, its consistency is verified, and features are extracted. The processed standardized data is uploaded to the cloud to achieve real-time data synchronization and lossless transmission, and to dynamically update the market value digital twin.

3. The method according to claim 1, characterized in that, The ESG_Modifier(Ω_t) is calculated in the following way: ESG_Modifier(Ω_t)=(1-Risk_Discount)×(1+Opportunity_Premium), Among them, Risk_Discount quantifies the negative performance of the environmental and social governance dimension into potential financial losses through a preset risk mapping model, and Opportunity_Premium quantifies the positive performance of the environmental and social governance dimension into incremental financial value through a preset opportunity identification model.

4. The method according to claim 3, characterized in that, The process of quantifying negative performance in the environmental and social governance dimension into potential financial losses through a pre-set risk mapping model includes: Extract negative event data from three dimensions: environment, society, and governance, from dynamic information sets; Quantitative indicators of negative event data are mapped to scores using linear or nonlinear interpolation based on benchmark comparison. Qualitative indicators are converted to scores using conditional matching based on a pre-defined rule dictionary. Based on the obtained scores, a loss function is established for each dimension to calculate its expected financial impact. The expected loss of each dimension is summed to obtain the total expected loss, and the total loss is mapped to the range of 0-1 using a normalization function to obtain Risk_Discount.

5. The method according to claim 3, characterized in that, The process of quantifying positive performance in the environmental and social governance dimension into incremental financial value through a pre-defined opportunity identification model includes: Extract positive event data from three dimensions: environment, society, and governance, from dynamic information sets; Calculate the corresponding financial returns using an event-related payoff function; The various financial gains are summed up and normalized relative to the enterprise value to obtain Opportunity_Premium.

6. The method according to claim 1, characterized in that, The adaptive value quantification model is trained or updated using hierarchical grouped LASSO regression technology, including: Obtain a dataset for model training. The dataset includes massive multi-dimensional factor data and corresponding market capitalization data of sample companies. The multi-dimensional factors include at least financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors belonging to different business logic levels. Based on the business logic hierarchy of the multi-dimensional factors, a hierarchical grouping structure is predefined, wherein at least multiple factors belonging to the same core business dimension are divided into the same factor group. Construct a loss function that includes a hierarchical grouping LASSO penalty term, wherein the loss function is: Where N is the number of samples, y i Let β be the sample market capitalization, β0 be the intercept term, and β j Let x be the coefficient of the j-th factor. ij Here, λ is the factor value, λ is the hyperparameter controlling the severity of the penalty, G is the total number of factor groups, and p... g Let ||β(g)||2 be the number of factors in the g-th group, and ||β(g)||2 be the L2 norm of the factor coefficients in the g-th group. With the goal of minimizing the loss function, an initial factor weight configuration is matched or quickly generated from a preset prior knowledge base to train the adaptive value quantification model. During the training process, the factor coefficients are retained or compressed to zero as a whole, with the factor group as the unit. An ensemble learning algorithm combining tree models is used to suppress multicollinearity through intra-group and inter-group constraints, and fine-tuning is performed during training using a loss function with group regularization terms.

7. The method according to claim 1, characterized in that, Also includes: ESG greenwashing risk identification of the target company's internal data, specifically including: Quantitative claim verification: Extract quantitative data from ESG claims and compare them with the original data from the internal environmental system for consistency. Qualitative claim analysis: Natural language processing technology is used to perform semantic analysis on vague qualitative statements and their credibility is evaluated by comparing them with third-party industry benchmark data; Tracking consistency between words and deeds: Establish a corporate ESG commitment database, compare new statements with historical commitments, and check for any contradictions or failures to meet commitments.

8. The method according to claim 1, characterized in that, Also includes: The output of the adaptive value quantification model is visualized to demonstrate the causal relationship paths between factors using a constraint-based causal discovery algorithm.

9. The method according to claim 1, characterized in that, It also includes a public opinion risk intelligent perception and linkage response step, which is executed independently or in parallel with the steps in claim 1, including: analyzing external public opinion data through natural language processing technology to identify negative events and their subjects; Based on the relationship between the entity and the target company, the negative events are automatically classified into different risk levels; Based on the risk level, the corresponding level of internal early warning is automatically triggered, and special self-inspection, decision assessment, or emergency plan activation tasks matching the risk level are generated and distributed.

10. An intelligent enterprise decision-making and execution system, characterized in that, include: The multi-source heterogeneous data acquisition module is used to continuously collect and integrate multi-source heterogeneous data from the target enterprise, and after processing by the multi-source consistency verification algorithm, construct and dynamically update the enterprise's market value digital twin. The enterprise decision-making basis data generation module is used to set preset multi-dimensional factors based on multi-source heterogeneous data, extract the data corresponding to each dimension factor in a preset order, input the data corresponding to each dimension factor into the adaptive value quantification model deployed in the market value digital twin, and perform fusion calculation on the input data according to the weights corresponding to each dimension factor through the adaptive value quantification model to generate enterprise decision-making basis data. The report generation module is used to automatically generate a draft strategy report with a preset directory structure based on enterprise decision-making data, and provides a human-machine collaborative interface to review and revise the draft report to form a final version to be confirmed; The task generation and dispatch module is used to respond to the management confirmation instruction of the final version, parse the structured content of the final version, and generate corresponding executable tasks according to the predefined responsibility mapping relationship and dispatch them to the corresponding execution end. The feedback and optimization module is used to collect task execution result feedback data and synchronize it to the market value digital twin and the adaptive value quantification model, so as to evaluate the execution effect based on the execution result feedback data and fine-tune the weights of each dimension factor in the adaptive value quantification model. The multi-dimensional factors include financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors, and the data corresponding to the financial factors, systemic importance factors, environmental and social governance factors, and short-term disturbance capture factors are extracted sequentially; the adaptive value quantification model uses the following formula for value quantification: Where MV_t represents the market value of the enterprise at time t, Ω_t represents the dynamic information set at time t, CFA_Base(Ω_t) is a financial factor, representing the basic financial value dynamically calculated based on the information set Ω_t at time t, and SCI_Modifier(Ω_t) is a system importance factor, representing the adjustment coefficient determined based on the assessment of the enterprise's system importance in the industry chain and industry. ESG_Modifier(Ω_t) is an environmental and social governance factor, which represents the risk and opportunity adjustment coefficient based on the quantitative calculation of a company's environmental, social and governance performance. f_sentiment(Investor_Structure_t, Narrative_t, Expectation_Gap_t, Technicals_t) represents the short-term disturbance quantification value determined by the adaptive value quantification model based on investor structure data, the industry concept propagation strength corresponding to the market narrative, the difference between the analysis consensus expectation and the intrinsic expectation calculated by the model, and the technical indicator feature vector. Here, Investor_Structure_t represents investor structure data; Narrative_t represents the industry concept propagation strength corresponding to the market narrative; Expectation_Gap_t represents the difference between the analysis consensus expectation and the intrinsic expectation calculated by the model; and Technicals_t represents the technical indicator feature vector including trend following, momentum, and volume indicators. The adaptive value quantification model learns the nonlinear relationship between the technical indicator feature vector and subsequent short-term price fluctuations, and dynamically adjusts the output weights of f_sentiment based on the learning results.

11. An electronic device, characterized in that, include: A memory and a processor are interconnected, the memory stores computer instructions, and the processor executes the computer instructions to perform the intelligent enterprise decision-making execution method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the intelligent enterprise decision-making execution method according to any one of claims 1 to 9.

13. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the intelligent enterprise decision-making execution method as described in any one of claims 1 to 9.

Images