Intelligent decision-making system and method based on enterprise life index large model

Abstract

The invention relates to the technical field of artificial intelligence and enterprise decision making, and discloses an intelligent decision making system and method based on an enterprise life index large model, and the system comprises a multi-source heterogeneous data collection module which is used for obtaining and associating various data in real time, and generating target enterprise multi-source heterogeneous data; the dynamic index system generation module is used for sequencing feature importance through a graph neural network and generating a dynamic evaluation index set containing survival, risk and growth indexes; the enterprise life index large model building module is used for fusing index sets, capturing enterprise state evolution, embedding risk factors and outputting multi-dimensional enterprise life indexes; the intelligent decision dynamic generation module is used for inputting the indexes into the decision strategy tree model and matching a strategy library to generate a decision instruction set; and the closed-loop feedback optimization module is used for realizing closed-loop optimization by acquiring execution data and combining Bayesian optimization dynamic correction parameters. According to the invention, the accuracy and timeliness of enterprise intelligent decision making can be improved.

Description

Background Art

[0002] Traditional enterprise intelligent decision-making systems mostly rely on static financial indicators or single-dimensional data analysis, which makes it difficult to dynamically capture the multi-dimensional correlation of the enterprise's operating status and lacks the ability to predict nonlinear risk factors. In existing technologies, enterprise evaluation models usually use a fixed indicator system that cannot be adaptively adjusted according to industry characteristics or market environment, resulting in delayed evaluation results and poor generalization. In addition, conventional systems often separate data analysis from decision generation and lack an end-to-end closed-loop optimization mechanism, resulting in a disconnect between decision recommendations and real-time business scenarios, making it difficult to cope with dynamic decision-making needs in a complex economic environment.

[0003] Therefore, it is urgent to provide a technical solution to solve the above problems. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an intelligent decision-making system and method based on a large model of enterprise life index.

[0005] In the first aspect, the present invention provides an intelligent decision-making system based on a large model of enterprise life index, and the technical solution of the system is as follows: The multi-source heterogeneous data acquisition module is used to obtain the target enterprise's structured financial data, unstructured public opinion data, supply chain time series data and industry benchmark data in real time, and to perform cross-modal correlation mapping of data through data lineage tracking technology to generate multi-source heterogeneous data of the target enterprise; A dynamic indicator system generation module, which is used to sort the feature importance of the multi-source heterogeneous data using a graph neural network, and generate an extensible dynamic evaluation indicator set according to preset industry classification rules; wherein the dynamic evaluation indicator set includes a viability indicator, a risk transmission indicator, and a growth potential indicator; The enterprise life index big model construction module is used to use the enterprise life index big model containing a preset Transformer architecture to fuse the dynamic evaluation indicator set, capture the state evolution law of the target enterprise through the temporal attention mechanism of the enterprise life index big model, and embed the industry risk disturbance factor generated by adversarial training, and output a multidimensional enterprise life index including survival probability, risk threshold level and trend deviation; An intelligent decision-making dynamic generation module is used to input the multi-dimensional enterprise life index into a pre-trained decision strategy tree model, dynamically match the risk response strategy library based on a reinforcement learning framework, and generate a decision instruction set including financing suggestions, supply chain adjustment plans and crisis warning levels; The closed-loop feedback optimization module is used to obtain the actual execution data after the execution of the decision instruction set through the decision effect collection interface deployed on the business terminal, and dynamically correct the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set based on the actual execution data and combined with the Bayesian optimization algorithm.

[0006] The beneficial effects of the intelligent decision-making system based on the enterprise life index large model of the present invention are as follows: The system of the present invention can effectively solve the problems of traditional enterprise intelligent decision-making systems that are difficult to dynamically capture the multi-dimensional correlation of enterprise operating status, lack of nonlinear risk prediction capabilities, delayed evaluation results and poor generalization, and decision recommendations being out of touch with real-time business scenarios, thereby improving the accuracy and timeliness of enterprise intelligent decision-making.

[0007] On the basis of the above scheme, the intelligent decision-making system based on the enterprise life index large model of the present invention can also be improved as follows.

[0008] In an optional manner, the multi-source heterogeneous data acquisition module is specifically used for: Retrieving the structured financial data from the target enterprise's ERP system through an API interface, and using web crawler technology to obtain the unstructured public opinion data, and based on IoT sensors, obtaining the supply chain time series data, and based on third-party database data, obtaining the industry benchmark data of the target enterprise; Perform field mapping and time granularity alignment on the structured financial data to obtain standard financial data, perform text sentiment analysis and entity extraction on the unstructured public opinion data to obtain standard public opinion data, perform anomaly detection and time window aggregation on the supply chain time series data to obtain standard supply chain data, perform unit consistency and time window matching on the industry benchmark data to obtain standard industry benchmark data; Add metadata tags to each type of standardized data, and build a blood relationship map between data entities based on the graph database; each type of standardized data includes: the standard financial data, the standard public opinion data, the standard supply chain data, and the standard industry benchmark data; Real-time data association is triggered through a rule engine to automatically update the blood relationship map, and each type of standardized data is encapsulated in a preset format to obtain the multi-source heterogeneous data of the target enterprise.

[0009] In an optional manner, the dynamic indicator system generation module is specifically used to: A heterogeneous graph including nodes and edges is constructed according to the multi-source heterogeneous data; wherein the nodes of the heterogeneous graph include: financial indicators, supply chain entities, public opinion events and industry benchmarks; the edge weights of the heterogeneous graph include: cost transmission weights of financial indicators and supply chain entities, and semantic similarity weights of public opinion events and industry benchmarks; A multi-layer graph neural network with an attention mechanism is used to train the heterogeneous graph, node features are aggregated through message passing, and a feature importance score of each node is output; The dynamic evaluation index set is generated according to the feature importance score of each node and the preset industry classification rules; wherein the viability index is: among the financial index and the supply chain related index, the index whose feature importance score is greater than the first threshold; the risk transmission index is: the index corresponding to the edge whose edge weight is greater than the second threshold; the growth potential index is: the deviation index relative to the industry mean is calculated in combination with the industry benchmark data; When changes in industry benchmarks or supply chain entities are detected, the feature importance scores of each node after the change are recalculated based on the incremental graph neural network, and the dynamic evaluation indicator set is updated according to the preset expansion rules.

[0010] In an optional manner, the preset Transformer architecture includes: an encoder unit and a decoder unit; the enterprise life index large model construction module is specifically used for: The viability index, the risk transmission index and the growth potential index in the dynamic evaluation index set are arranged in a time series of a preset time window length, the index value of each time window is normalized, and the time dimension information is added through sine function position coding to generate a model input tensor; The model input tensor is input into the preset Transformer architecture to output the multi-dimensional enterprise life index; wherein the encoder unit uses a multi-head self-attention mechanism to calculate the dependency weights of each indicator across time windows to capture the nonlinear state association law; the decoder unit fuses the indicator value of the current time window with the industry benchmark data to generate the survival probability through autoregressive prediction; Based on the preset industry risk scenarios corresponding to the target enterprise, an anti-disturbance factor is generated and embedded in the model to train the enterprise life index model. Based on the historical risk event data of the target enterprise, the risk threshold level of the target enterprise is determined, and the trend deviation between the current index value of the target enterprise and the industry benchmark average is calculated.

[0011] In an optional manner, the decision strategy tree model is: a hierarchical decision model constructed based on a rule engine and a reinforcement learning framework; the intelligent decision dynamic generation module is specifically used for: Constructing the hierarchical decision model based on the fusion of rule engine and reinforcement learning, inputting the survival probability, the risk threshold level and the trend deviation into the hierarchical decision model, extracting strategy instructions from the leaf nodes matched by the hierarchical decision model, and generating the structured decision instruction set; The reward value is calculated according to the actual effect after the execution of the decision instruction set, and the strategy selection weight of the hierarchical decision model is updated based on the reinforcement learning framework to dynamically update the hierarchical decision model.

[0012] In an optional manner, the actual execution data includes: supply chain adjustment effect data, financing strategy execution data and crisis response effect data; the closed-loop feedback optimization module is specifically used to: The supply chain adjustment effect data, the financing strategy execution data and the crisis response effect data are aligned with the predefined decision-making objectives, quantitative evaluation indicators are generated, and the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set are dynamically corrected in combination with the Bayesian optimization algorithm.

[0013] In a second aspect, the present invention provides an intelligent decision-making method based on a large model of enterprise life index, and the technical solution of the method is as follows: Acquire the target enterprise's structured financial data, unstructured public opinion data, supply chain time series data, and industry benchmark data in real time, and use data lineage tracking technology to perform cross-modal correlation mapping of the data to generate multi-source heterogeneous data of the target enterprise; Utilize graph neural network to sort the feature importance of the multi-source heterogeneous data, and generate an extensible dynamic evaluation indicator set according to preset industry classification rules; wherein the dynamic evaluation indicator set includes viability indicators, risk transmission indicators and growth potential indicators; The enterprise life index big model with a preset Transformer architecture is used to integrate the dynamic evaluation indicator set, and the state evolution law of the target enterprise is captured through the temporal attention mechanism of the enterprise life index big model, and the industry risk disturbance factor generated by adversarial training is embedded, and a multi-dimensional enterprise life index including survival probability, risk threshold level and trend deviation is output; Input the multi-dimensional enterprise life index into the pre-trained decision strategy tree model, dynamically match the risk response strategy library based on the reinforcement learning framework, and generate a decision instruction set including financing suggestions, supply chain adjustment plans and crisis warning levels; The actual execution data after the execution of the decision instruction set is obtained through the decision effect collection interface deployed on the business terminal, and based on the actual execution data, the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set are dynamically corrected in combination with the Bayesian optimization algorithm.

[0014] The beneficial effects of the intelligent decision-making method based on the enterprise life index large model of the present invention are as follows: The method of the present invention can effectively solve the problems of traditional enterprise intelligent decision-making systems that are difficult to dynamically capture the multi-dimensional correlation of enterprise operating status, lack of nonlinear risk prediction capabilities, delayed evaluation results and poor generalization, and decision recommendations being out of touch with real-time business scenarios, thereby improving the accuracy and timeliness of enterprise intelligent decision-making.

[0015] On the basis of the above scheme, the intelligent decision-making method based on the enterprise life index large model of the present invention can also be improved as follows.

[0016] In an optional manner, the step of acquiring structured financial data, unstructured public opinion data, supply chain time series data and industry benchmark data of the target enterprise in real time, and performing cross-modal association mapping of the data through data lineage tracking technology to generate multi-source heterogeneous data of the target enterprise includes: Retrieving the structured financial data from the target enterprise's ERP system through an API interface, and using web crawler technology to obtain the unstructured public opinion data, and based on IoT sensors, obtaining the supply chain time series data, and based on third-party database data, obtaining the industry benchmark data of the target enterprise; Perform field mapping and time granularity alignment on the structured financial data to obtain standard financial data, perform text sentiment analysis and entity extraction on the unstructured public opinion data to obtain standard public opinion data, perform anomaly detection and time window aggregation on the supply chain time series data to obtain standard supply chain data, perform unit consistency and time window matching on the industry benchmark data to obtain standard industry benchmark data; Add metadata tags to each type of standardized data, and build a blood relationship map between data entities based on the graph database; each type of standardized data includes: the standard financial data, the standard public opinion data, the standard supply chain data, and the standard industry benchmark data; Real-time data association is triggered through a rule engine to automatically update the blood relationship map, and each type of standardized data is encapsulated in a preset format to obtain the multi-source heterogeneous data of the target enterprise.

[0017] In a third aspect, a technical solution of an electronic device of the present invention is as follows: The invention comprises a memory, a processor and a program stored in the memory and running on the processor. When the processor executes the program, the steps of the intelligent decision-making method based on the enterprise life index large model of the present invention are implemented.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium having the following technical solution: Instructions are stored in the computer-readable storage medium. When the computer-readable storage medium reads the instructions, the computer-readable storage medium executes the steps of the intelligent decision-making method based on the enterprise life index large model of the present invention.

[0019] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present invention. In addition, the same reference symbols are used to represent the same components throughout the accompanying drawings. In the accompanying drawings: Figure 1 It is a structural schematic diagram of an embodiment of an intelligent decision-making system based on a large enterprise life index model of the present invention; Figure 2 It is a flow chart of an embodiment of an intelligent decision-making method based on a large enterprise life index model of the present invention; Figure 3 The figure is a schematic structural diagram of an embodiment of an electronic device of the present invention. DETAILED DESCRIPTION

[0021] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0022] Figure 1 FIG. 1 is a schematic diagram showing a structure of an embodiment of an intelligent decision-making system 100 based on a large enterprise life index model provided by the present invention. Figure 1 As shown, the system 100 includes: The multi-source heterogeneous data acquisition module 110 is used to obtain the target enterprise's structured financial data, unstructured public opinion data, supply chain time series data and industry benchmark data in real time, and to perform cross-modal correlation mapping of the data through data lineage tracking technology to generate multi-source heterogeneous data of the target enterprise.

[0023] Among them, the target enterprise is the enterprise that needs to make intelligent decisions in this embodiment, such as automobile enterprises, manufacturing enterprises, etc., and there is no restriction here. Structured financial data refers to: numerical financial information with fixed format and clear fields, such as: balance sheet (total assets, current liabilities, etc.), income statement (operating income, net profit, etc.) and cash flow statement (cash flow from operating activities, investment expenditure, etc.). Unstructured public opinion data refers to: text, image or audio and video data without a fixed format, reflecting the market's evaluation, emotions or emergencies of the enterprise, such as: social media comments (discussions on "battery spontaneous combustion" on social platforms); news report titles (such as "XX company's net profit plummeted"); user-generated content (short videos, forum posts). Supply chain time series data refers to: supply chain operation data with timestamps, recording dynamic changes in procurement, production, logistics and other links, such as: logistics sensor data (GPS track, temperature and humidity records); purchase order fulfillment rate (on-time delivery ratio); inventory turnover rate (inventory consumption per unit time). Industry benchmark data refers to: standardized reference data at the industry level, used for horizontal comparison of corporate performance, such as: industry average financial indicators (average gross profit margin, debt-to-asset ratio); raw material price index (such as the quarterly average price of lithium carbonate); policy and regulatory database (carbon emission standards, trade agreements).

[0024] Among them, data lineage tracking technology refers to the technology of building a data life cycle association map by recording data sources, flow paths and dependencies, supporting data traceability, impact analysis and dynamic association updates. Multi-source heterogeneous data refers to: data sets from different systems, with diverse formats and differentiated structures, including multiple modalities such as numerical values, texts, and time series. In this embodiment, multi-source heterogeneous data specifically standardizes structured financial data, unstructured public opinion data, supply chain time series data, and industry benchmark data, and establishes cross-modal association mapping through data lineage tracking technology, so as to achieve dynamically updated data.

[0025] It should be noted that structured financial data is used to quantify a company's debt repayment, profitability, and operational capabilities, unstructured public opinion data is used to capture reputation risks and market expectation fluctuations, supply chain time series data is used to monitor supply chain stability and risk transmission, and industry benchmark data is used to provide horizontal comparison references and position the company's competitive position.

[0026] The dynamic indicator system generation module 120 is used to use a graph neural network to sort the feature importance of the multi-source heterogeneous data and generate an extensible dynamic evaluation indicator set according to preset industry classification rules.

[0027] Among them, the dynamic evaluation indicator set includes viability indicators, risk transmission indicators and growth potential indicators. Graph neural network (GNN) is a deep learning model for processing graph structured data. The graph consists of nodes (entities) and edges (relationships between entities). GNN learns the representation of nodes and graphs by aggregating the information of neighboring nodes, thereby capturing complex relationships. The preset industry classification rules are a set of rules pre-established based on the business characteristics, risk types and assessment requirements of different industries, which are used to guide dynamic indicator generation and model optimization. For example, taking the automotive industry as an example, the types corresponding to the preset industry classification rules include: indicator weight constraints, risk transmission priorities, dynamic expansion conditions and data association thresholds.

[0028] The enterprise life index big model construction module 130 is used to adopt the enterprise life index big model containing a preset Transformer architecture to fuse the dynamic evaluation indicator set, capture the state evolution law of the target enterprise through the temporal attention mechanism of the enterprise life index big model, and embed the industry risk disturbance factor generated by adversarial training, and output a multidimensional enterprise life index including survival probability, risk threshold level and trend deviation.

[0029] Among them, the preset Transformer architecture includes: encoder unit and decoder unit; the encoder unit includes: ① multi-head self-attention layer, used to calculate the cross-time window dependency weights between indicators (such as the attention weight of lithium price fluctuations on cash flow gaps); ② feedforward neural network, used for nonlinear mapping features; ③ layer normalization and residual connection: improve training stability. The decoder unit includes: ① masked self-attention layer, used to prevent future information leakage; ② cross-attention layer: used to fuse encoder output with industry benchmark data; ③ prediction head: used to output survival probability and risk transmission path probability distribution.

[0030] The enterprise life index model is constructed based on the preset Transformer architecture, and the enterprise life index model uses the time series modeling capability of the preset Transformer architecture. The enterprise life index model of this embodiment embeds industry risk disturbance factors (such as lithium price mutation simulation) and a set of dynamic evaluation indicators in the preset Transformer architecture.

[0031] Among them, the law of state evolution refers to the nonlinear pattern of changes in corporate indicators (survivability indicators, risk transmission indicators and growth potential indicators) over time, such as the lag effect of deterioration of financial indicators caused by supply chain disruptions. Industry risk disturbance factors refer to noise data that simulate industry events, which are used to enhance the robustness of the enterprise life index model in abnormal scenarios. For example, construct adversarial training samples: inject noise into industry benchmark data (such as a 50% surge in lithium prices and sudden policy changes). The goal of adversarial training is to minimize the prediction deviation before and after the disturbance (such as survival probability error ≤ 5%).

[0032] Among them, the multidimensional enterprise life index is a set of composite indicators that quantify the enterprise's survival ability (survival probability), risk level (risk threshold level) and competitive position (trend deviation).

[0033] The intelligent decision-making dynamic generation module 140 is used to input the multi-dimensional enterprise life index into the pre-trained decision strategy tree model, dynamically match the risk response strategy library based on the reinforcement learning framework, and generate a decision instruction set including financing suggestions, supply chain adjustment plans and crisis warning levels.

[0034] Among them, the decision strategy tree model realizes multi-condition strategy matching through a tree structure. The decision strategy tree model includes: root node, intermediate node and leaf node. The root node triggers different decision branches based on the risk threshold level (high, medium, low); the intermediate node divides the sub-branches according to the risk type (supply chain, financing, public opinion); the leaf node mounts specific strategies (such as "switching suppliers" and "issuing bonds").

[0035] Among them, based on the historical event library and combined with the priority rules defined by industry experts, the initial model is iteratively trained through supervised learning (such as the decision tree algorithm) to obtain a pre-trained decision strategy tree model, which can match more than 80% of historical correct decisions.

[0036] The closed-loop feedback optimization module 150 is used to obtain the actual execution data after the execution of the decision instruction set through the decision effect collection interface deployed on the business terminal, and dynamically correct the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set based on the actual execution data in combination with the Bayesian optimization algorithm.

[0037] Among them, actual execution data refers to the data used to characterize the execution effect produced after the target enterprise executes the decision instruction set.

[0038] The technical solution of this embodiment can effectively solve the problems of traditional enterprise intelligent decision-making systems that are difficult to dynamically capture the multi-dimensional correlation of enterprise operating status, lack of nonlinear risk prediction capabilities, delayed evaluation results and poor generalization, and decision recommendations being out of touch with real-time business scenarios, thereby improving the accuracy and timeliness of enterprise intelligent decision-making.

[0039] In an optional manner, the multi-source heterogeneous data acquisition module 110 is specifically used for: The structured financial data is retrieved from the ERP system of the target enterprise through the API interface, and the unstructured public opinion data is obtained by using web crawler technology, and the supply chain time series data is obtained based on IoT sensors, and the industry benchmark data of the target enterprise is obtained based on third-party database data.

[0040] Among them, the ERP system refers to the Enterprise Resource Planning system, which is a comprehensive enterprise management software that stores a lot of data including corporate financial data. IoT sensors are IoT sensors, smart terminals deployed on physical devices, which collect supply chain time series data in real time and transmit them through the network.

[0041] The structured financial data is subjected to field mapping and time granularity alignment to obtain standard financial data, and the unstructured public opinion data is subjected to text sentiment analysis and entity extraction to obtain standard public opinion data, and the supply chain time series data is subjected to anomaly detection and time window aggregation to obtain standard supply chain data, and the industry benchmark data is subjected to unit consistency and time window matching to obtain standard industry benchmark data.

[0042] Among them, field mapping refers to: mapping financial fields of different sources / formats to standard accounting account codes. For example, the original field name "Current Assets" is mapped to "Current Assets (Account Code: 1001)". Time granularity alignment refers to: converting data of different periods (such as weekly reports, monthly reports) into a unified time unit (such as monthly, quarterly, annual, etc.). Event granularity alignment methods include but are not limited to: accumulation method (such as adding up 3 months of profits into quarterly profits) and interpolation method (missing month data is filled with linear interpolation). Taking a certain automobile company as an example, the balance sheet for the first quarter is extracted from the ERP system of the automobile company, and the quarterly report is generated using the accumulation method. The fields are standardized according to the IFRS code to obtain the standard financial data of the automobile company in the first quarter.

[0043] Among them, sentiment analysis refers to: using the BERT model or dictionary method (VADER) to extract sentiment polarity scores from unstructured public opinion data, with a value range of [-1, +1]; for example, the sentiment score of "frequent battery failures" is -0.8 (strong negative). The process of entity extraction is: using named entity recognition (NER) to identify information such as company names, product names, and event types. Taking a certain automobile company as an example, the comment "XX company's battery spontaneous combustion incident has not been resolved" captured from social media was analyzed and extracted to obtain standard public opinion data: the sentiment score was -0.75 (negative); the entities were: XX company (company name), battery (product name), spontaneous combustion (event type).

[0044] Among them, anomaly detection uses statistical methods (Z-Score) and machine learning (local outlier factor), and the time window aggregation methods include: fixed window and sliding window. Taking a certain automobile company as an example, in the supply chain time series data (delivery on-time rate data) recorded by IoT sensors, it was detected that the on-time rate of a supplier dropped suddenly from 95% to 60% (Z-Score=4.2), which was marked as anomaly; the "daily average on-time rate 78%" indicator was generated by daily aggregation to generate standard supply chain data.

[0045] The unit consistency process includes currency conversion and dimension unification. The time window matching process includes data cycle alignment and timestamp calibration. Taking a certain automobile company as an example, the lithium price data unit "US dollars / ton" in the industry benchmark data is converted to "RMB / ton" and aligned with the quarterly procurement data time window of the automobile company to generate standard industry benchmark data.

[0046] Metadata tags are added to each type of standardized data, and a blood relationship map between data entities is constructed based on the graph database.

[0047] Among them, each type of standardized data includes: standard financial data, standard public opinion data, standard supply chain data and standard industry benchmark data. Metadata tags include: data source, timestamp, version mark. Graph database is a special type of database that represents and stores data in a graph structure and is suitable for processing complex relationship networks. The blood relationship map refers to: the association network between data entities, for example, the edge weight of "lithium price node and battery cost node" is 0.85). Specifically, the blood relationship map includes: ① nodes: financial indicator entities, supply chain node entities, public opinion event entities and industry indicator entities; ② edges: representing the causal influence relationship, risk transmission relationship and semantic association relationship between nodes.

[0048] Real-time data association is triggered through a rule engine to automatically update the blood relationship map, and each type of standardized data is encapsulated in a preset format to obtain the multi-source heterogeneous data of the target enterprise.

[0049] Among them, the rule engine is a software system that automatically triggers the data processing process based on predefined logical conditions (rule sets), which is used to achieve real-time data association, graph update and format encapsulation. The process of preset format encapsulation includes: ① converting the original field according to the preset key name (such as "delivery_rate" to "delivery on-time rate"); ② adding data source, timestamp, and lineage association nodes; verifying data integrity, and triggering an alarm if it fails.

[0050] In the above optional solutions, the financial data of an automobile company is obtained from the ERP system through API, the public opinion data is captured by crawling social media comments (keyword "battery spontaneous combustion"), and the supply chain data is collected by IoT sensors to collect the delivery records of the Yangtze River Delta Logistics Center. After the data is standardized, the metadata is marked, and a blood relationship map is constructed to show that the edge weight of "lithium price fluctuations and battery costs" is 0.85.

[0051] In an optional manner, the dynamic indicator system generating module 120 is specifically used for: A heterogeneous graph including nodes and edges is constructed according to the multi-source heterogeneous data.

[0052] The nodes of the heterogeneous graph include: financial indicators (such as debt-to-asset ratio, etc.), supply chain entities (such as suppliers and logistics centers, etc.), public opinion events (such as XXX event) and industry benchmarks (industry average inventory turnover rate). The edge weights of the heterogeneous graph include: cost transmission weights of financial indicators and supply chain entities (such as the weight of battery cost increase caused by rising lithium prices is 0.8), and the semantic similarity weights of public opinion events and industry benchmarks (such as the semantic association weight of "battery failure" public opinion and "after-sales cost" = 0.6).

[0053] A multi-layer graph neural network with an attention mechanism is used to train the heterogeneous graph, aggregate node features through message passing, and output the feature importance score of each node.

[0054] Among them, the feature importance score refers to the contribution of the node to the risk assessment. For example, the feature importance score corresponding to the supply chain entity is 0.38, the feature importance score corresponding to the industry benchmark is 0.35, the feature importance score corresponding to the financial indicator is 0.28, and the feature importance score corresponding to the public opinion event is 0.18.

[0055] The dynamic evaluation index set is generated according to the feature importance score of each node and the preset industry classification rules.

[0056] Among them, the viability index is: among the financial index and supply chain related index, the index with feature importance score greater than the first threshold; the risk transmission index is: the index corresponding to the edge with edge weight greater than the second threshold; the growth potential index is: combined with the industry benchmark data, the deviation index relative to the industry mean is calculated. It should be noted that the first threshold defaults to 0.3, the second threshold defaults to 0.6, and the deviation defaults to ±1.5.

[0057] When changes in industry benchmarks or supply chain entities are detected, the feature importance scores of each node after the change are recalculated based on the incremental graph neural network, and the dynamic evaluation indicator set is updated according to the preset expansion rules.

[0058] Among them, the trigger conditions for change detection are: ① Changes in industry benchmarks, such as the monthly increase in lithium prices exceeding the historical fluctuation threshold; ② Changes in supply chain entities, such as adding / replacing suppliers.

[0059] The process of recalculating the feature importance scores of each node after the change based on the incremental graph neural network is as follows: identifying the nodes and edges affected by the change; freezing the model parameters of the unchanged part, and only performing gradient updates on the node embeddings and edge weights of the affected subgraphs; and re-running the message passing of the multi-layer graph neural network based on the updated node embeddings to calculate the importance scores of each node.

[0060] Among them, the preset extension rules include: new indicator rules, exclusion indicator rules and weight adjustment rules.

[0061] In an optional manner, the enterprise life index large model building module 130 is specifically used to: The viability index, the risk transmission index and the growth potential index in the dynamic evaluation index set are arranged in a time series of a preset time window length, the index value of each time window is normalized, and the time dimension information is added through sinusoidal function position encoding to generate a model input tensor.

[0062] The model input tensor is input into the preset Transformer architecture, and the multi-dimensional enterprise life index is output.

[0063] Among them, the preset Transformer architecture includes: an encoder unit and a decoder unit; the encoder unit uses a multi-head self-attention mechanism to calculate the dependency weights of each indicator across time windows to capture the nonlinear state association law. The decoder unit fuses the indicator value of the current time window with the industry benchmark data, and generates the survival probability through autoregressive prediction. It should be noted that the decoder unit outputs the predicted original value, which is converted to the [0,100%] interval by the Sigmoid function.

[0064] Based on the preset industry risk scenarios corresponding to the target enterprise, an anti-disturbance factor is generated and embedded in the model to train the enterprise life index model. Based on the historical risk event data of the target enterprise, the risk threshold level of the target enterprise is determined, and the trend deviation between the current index value of the target enterprise and the industry benchmark average is calculated.

[0065] Among them, the historical quantile method is used, combined with the historical risk event data of the target enterprise, to determine the risk threshold level of the target enterprise, which includes: high risk, medium risk and low risk. The Z-Score is used to calculate the trend deviation between the current indicator value of the target enterprise and the industry benchmark average.

[0066] In an optional manner, the intelligent decision dynamic generation module 140 is specifically used to: Construct the hierarchical decision model based on the fusion of rule engine and reinforcement learning, input the survival probability, the risk threshold level and the trend deviation into the hierarchical decision model, extract the strategy instructions from the leaf nodes matched by the hierarchical decision model, and generate the structured decision instruction set.

[0067] Among them, the decision strategy tree model is: a hierarchical decision model built based on a rule engine and a reinforcement learning framework. The root node of the hierarchical decision model uses the risk threshold level as the initial decision trigger condition to divide the decision branches into three categories: high risk, medium risk, and low risk; the intermediate node constructs a secondary decision branch according to the risk transmission path type (supply chain risk, financing liquidity risk, public opinion crisis risk); the leaf node mounts predefined risk response strategies, including a financing strategy library, a supply chain adjustment strategy library, and a crisis response strategy library.

[0068] The reward value is calculated according to the actual effect after the execution of the decision instruction set, and the strategy selection weight of the hierarchical decision model is updated based on the reinforcement learning framework to dynamically update the hierarchical decision model.

[0069] Among them, the reinforcement learning framework includes: state space, action space and reward function. The state space refers to the survival probability, risk threshold level, and trend deviation. The action space predefines the executable actions in the strategy library (such as issuing bonds, switching suppliers, and launching public relations responses). The reward function calculates the reward value based on the actual effect after the strategy is executed (such as the increase in inventory turnover rate and the reduction rate of financing costs), updates the strategy priority, and thus dynamically updates the hierarchical decision model.

[0070] In an optional manner, the closed-loop feedback optimization module 150 is specifically used to: The supply chain adjustment effect data, the financing strategy execution data and the crisis response effect data are aligned with the predefined decision-making objectives, quantitative evaluation indicators are generated, and the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set are dynamically corrected in combination with the Bayesian optimization algorithm.

[0071] Among them, actual execution data includes: supply chain adjustment effect data, financing strategy execution data and crisis response effect data. For example, supply chain adjustment effect data includes: supplier delivery on-time rate change value, inventory turnover rate optimization range, logistics cost volatility. Financing strategy execution data includes: financing cost actual interest rate, fund arrival cycle, market credit rating changes. Crisis response effect data includes: public opinion negative sentiment index decline rate, crisis event resolution time, and regulatory penalties.

[0072] Among them, quantitative evaluation indicators include: positive effect indicators (such as supply chain resilience improvement coefficient), negative risk indicators (such as financing cost overrun rate) and timeliness indicators (such as financing cost overrun rate).

[0073] Among them, the dynamic correction process includes: weight parameter correction of the enterprise life index large model and dynamic evaluation indicator set correction. The weight parameter correction of the enterprise life index large model specifically includes: adjusting the Transformer attention head weights to strengthen the attention allocation of high-contribution indicators; updating the strength of adversarial training perturbation factors to adapt to the current industry risk level. The dynamic evaluation indicator set correction specifically includes: eliminating long-term low-weight indicators; adding derivative indicators with significant sudden increases in actual execution data. The corrected enterprise life index large model and the dynamic evaluation indicator set are verified for consistency. If the verification error rate is ≤5%, the correction is completed; if the error rate is >5%, roll back to the model and set before the dynamic correction and trigger the manual intervention process.

[0074] Figure 2 FIG. 1 is a flow chart showing an embodiment of an intelligent decision-making method based on a large enterprise life index model provided by the present invention. Figure 2 As shown, the following steps are included: S1. Acquire the target enterprise's structured financial data, unstructured public opinion data, supply chain time series data and industry benchmark data in real time, and perform cross-modal correlation mapping of the data through data lineage tracking technology to generate multi-source heterogeneous data of the target enterprise; S2. Rank the feature importance of the multi-source heterogeneous data using a graph neural network, and generate an extensible dynamic evaluation indicator set according to preset industry classification rules; wherein the dynamic evaluation indicator set includes a viability indicator, a risk transmission indicator, and a growth potential indicator; S3. Adopting a large enterprise life index model with a preset Transformer architecture to fuse the dynamic evaluation indicator set, capturing the state evolution law of the target enterprise through the temporal attention mechanism of the large enterprise life index model, and embedding the industry risk disturbance factor generated by adversarial training, outputting a multidimensional enterprise life index including survival probability, risk threshold level and trend deviation; S4, inputting the multi-dimensional enterprise life index into the pre-trained decision strategy tree model, dynamically matching the risk response strategy library based on the reinforcement learning framework, and generating a decision instruction set including financing suggestions, supply chain adjustment plans and crisis warning levels; S5. Obtain the actual execution data after the execution of the decision instruction set through the decision effect collection interface deployed on the business terminal, and dynamically correct the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set based on the actual execution data and the Bayesian optimization algorithm.

[0075] In an optional manner, S1 includes: Retrieving the structured financial data from the target enterprise's ERP system through an API interface, and using web crawler technology to obtain the unstructured public opinion data, and based on IoT sensors, obtaining the supply chain time series data, and based on third-party database data, obtaining the industry benchmark data of the target enterprise; Perform field mapping and time granularity alignment on the structured financial data to obtain standard financial data, perform text sentiment analysis and entity extraction on the unstructured public opinion data to obtain standard public opinion data, perform anomaly detection and time window aggregation on the supply chain time series data to obtain standard supply chain data, perform unit consistency and time window matching on the industry benchmark data to obtain standard industry benchmark data; Add metadata tags to each type of standardized data, and build a blood relationship map between data entities based on the graph database; each type of standardized data includes: the standard financial data, the standard public opinion data, the standard supply chain data, and the standard industry benchmark data; Real-time data association is triggered through a rule engine to automatically update the blood relationship map, and each type of standardized data is encapsulated in a preset format to obtain the multi-source heterogeneous data of the target enterprise.

[0076] In an optional manner, S2 includes: A heterogeneous graph including nodes and edges is constructed according to the multi-source heterogeneous data; wherein the nodes of the heterogeneous graph include: financial indicators, supply chain entities, public opinion events and industry benchmarks; the edge weights of the heterogeneous graph include: cost transmission weights of financial indicators and supply chain entities, and semantic similarity weights of public opinion events and industry benchmarks; A multi-layer graph neural network with an attention mechanism is used to train the heterogeneous graph, node features are aggregated through message passing, and a feature importance score of each node is output; The dynamic evaluation index set is generated according to the feature importance score of each node and the preset industry classification rules; wherein the viability index is: among the financial index and the supply chain related index, the index whose feature importance score is greater than the first threshold; the risk transmission index is: the index corresponding to the edge whose edge weight is greater than the second threshold; the growth potential index is: the deviation index relative to the industry mean is calculated in combination with the industry benchmark data; When changes in industry benchmarks or supply chain entities are detected, the feature importance scores of each node after the change are recalculated based on the incremental graph neural network, and the dynamic evaluation indicator set is updated according to the preset expansion rules.

[0077] In an optional manner, the preset Transformer architecture includes: an encoder unit and a decoder unit; S3 includes: The viability index, the risk transmission index and the growth potential index in the dynamic evaluation index set are arranged in a time series of a preset time window length, the index value of each time window is normalized, and the time dimension information is added through sine function position coding to generate a model input tensor; The model input tensor is input into the preset Transformer architecture to output the multi-dimensional enterprise life index; wherein the encoder unit uses a multi-head self-attention mechanism to calculate the dependency weights of each indicator across time windows to capture the nonlinear state association law; the decoder unit fuses the indicator value of the current time window with the industry benchmark data to generate the survival probability through autoregressive prediction; Based on the preset industry risk scenarios corresponding to the target enterprise, an anti-disturbance factor is generated and embedded in the model to train the enterprise life index model. Based on the historical risk event data of the target enterprise, the risk threshold level of the target enterprise is determined, and the trend deviation between the current index value of the target enterprise and the industry benchmark average is calculated.

[0078] In an optional manner, the decision strategy tree model is: a hierarchical decision model constructed based on a rule engine and a reinforcement learning framework; S4 includes: Constructing the hierarchical decision model based on the fusion of rule engine and reinforcement learning, inputting the survival probability, the risk threshold level and the trend deviation into the hierarchical decision model, extracting strategy instructions from the leaf nodes matched by the hierarchical decision model, and generating the structured decision instruction set; The reward value is calculated according to the actual effect after the execution of the decision instruction set, and the strategy selection weight of the hierarchical decision model is updated based on the reinforcement learning framework to dynamically update the hierarchical decision model.

[0079] In an optional manner, the actual execution data includes: supply chain adjustment effect data, financing strategy execution data and crisis response effect data; S5 includes: The supply chain adjustment effect data, the financing strategy execution data and the crisis response effect data are aligned with the predefined decision-making objectives, quantitative evaluation indicators are generated, and the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set are dynamically corrected in combination with the Bayesian optimization algorithm.

[0080] It should be noted that the beneficial effects of the intelligent decision-making method based on the enterprise life index big model provided by the above embodiment are the same as the beneficial effects of the above intelligent decision-making system 100 based on the enterprise life index big model, which will not be repeated here.

[0081] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any of the above-mentioned intelligent decision-making methods based on the enterprise life index big model is implemented. That is to say, an electronic device according to an embodiment of the present invention may include but is not limited to: a processor and a memory; a memory for storing a computer program; a processor for executing the intelligent decision-making method based on the enterprise life index big model shown in any embodiment of the present invention by calling the computer program.

[0082] In an alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, such as through a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0083] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of the present invention. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0084] The bus 4002 may include a path to transmit information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 In the figure, only one thick line is used to represent the bus 4002, but this does not mean that there is only one bus or one type of bus.

[0085] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0086] The memory 4003 is used to store application code (computer program) for executing the solution of the present invention, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the application code stored in the memory 4003 to implement the content shown in the above method embodiment.

[0087] Among them, the electronic device can also be a terminal device, and the terminal device can be any terminal device that can install applications and access web pages through applications, including at least one of a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart car device.

[0088] It should be noted that Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0089] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements any of the above-mentioned intelligent decision-making methods based on the enterprise life index large model.

[0090] Optionally, the computer readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0091] In an exemplary embodiment, a computer program product or a computer program is also provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes the above-mentioned intelligent decision-making method based on the enterprise life index large model.

[0092] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0093] It should be understood that the flow charts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the module, the program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0094] The computer-readable storage medium provided by the embodiment of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or component.

[0095] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0096] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) to form a technical solution.

[0097] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects and represent the definition of a specific order or sequence. The order of use of similar objects can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0098] Those skilled in the art know that the present invention can be implemented as a system, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.

[0099] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. An intelligent decision-making system based on the enterprise life index model, characterized in that: include: The multi-source heterogeneous data acquisition module is used to obtain the target enterprise's structured financial data, unstructured public opinion data, supply chain time series data and industry benchmark data in real time, and to perform cross-modal correlation mapping of data through data lineage tracking technology to generate multi-source heterogeneous data of the target enterprise; A dynamic indicator system generation module, which is used to sort the feature importance of the multi-source heterogeneous data using a graph neural network, and generate an extensible dynamic evaluation indicator set according to preset industry classification rules; wherein the dynamic evaluation indicator set includes a viability indicator, a risk transmission indicator, and a growth potential indicator; The enterprise life index big model construction module is used to use the enterprise life index big model containing a preset Transformer architecture to fuse the dynamic evaluation indicator set, capture the state evolution law of the target enterprise through the temporal attention mechanism of the enterprise life index big model, and embed the industry risk disturbance factor generated by adversarial training, and output a multidimensional enterprise life index including survival probability, risk threshold level and trend deviation; An intelligent decision-making dynamic generation module is used to input the multi-dimensional enterprise life index into a pre-trained decision strategy tree model, dynamically match the risk response strategy library based on a reinforcement learning framework, and generate a decision instruction set including financing suggestions, supply chain adjustment plans and crisis warning levels; The closed-loop feedback optimization module is used to obtain the actual execution data after the execution of the decision instruction set through the decision effect collection interface deployed on the business terminal, and dynamically correct the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set based on the actual execution data and combined with the Bayesian optimization algorithm.

2. The system according to claim 1, characterized in that The multi-source heterogeneous data acquisition module is specifically used for: Retrieving the structured financial data from the target enterprise's ERP system through an API interface, and using web crawler technology to obtain the unstructured public opinion data, and based on IoT sensors, obtaining the supply chain time series data, and based on third-party database data, obtaining the industry benchmark data of the target enterprise; Perform field mapping and time granularity alignment on the structured financial data to obtain standard financial data, perform text sentiment analysis and entity extraction on the unstructured public opinion data to obtain standard public opinion data, perform anomaly detection and time window aggregation on the supply chain time series data to obtain standard supply chain data, perform unit consistency and time window matching on the industry benchmark data to obtain standard industry benchmark data; Add metadata tags to each type of standardized data, and build a blood relationship map between data entities based on the graph database; Each type of standardized data includes: the standard financial data, the standard public opinion data, the standard supply chain data and the standard industry benchmark data; Real-time data association is triggered through a rule engine to automatically update the blood relationship map, and each type of standardized data is encapsulated in a preset format to obtain the multi-source heterogeneous data of the target enterprise.

3. The system according to claim 2, characterized in that The dynamic indicator system generation module is specifically used for: A heterogeneous graph including nodes and edges is constructed according to the multi-source heterogeneous data; wherein the nodes of the heterogeneous graph include: financial indicators, supply chain entities, public opinion events and industry benchmarks; the edge weights of the heterogeneous graph include: cost transmission weights of financial indicators and supply chain entities, and semantic similarity weights of public opinion events and industry benchmarks; A multi-layer graph neural network with an attention mechanism is used to train the heterogeneous graph, node features are aggregated through message passing, and a feature importance score of each node is output; The dynamic evaluation index set is generated according to the feature importance score of each node and the preset industry classification rules; wherein the viability index is: among the financial index and the supply chain related index, the index whose feature importance score is greater than the first threshold; the risk transmission index is: the index corresponding to the edge whose edge weight is greater than the second threshold; the growth potential index is: the deviation index relative to the industry mean is calculated in combination with the industry benchmark data; When changes in industry benchmarks or supply chain entities are detected, the feature importance scores of each node after the change are recalculated based on the incremental graph neural network, and the dynamic evaluation indicator set is updated according to the preset expansion rules.

4. The system according to claim 3, characterized in that The preset Transformer architecture includes: an encoder unit and a decoder unit; the enterprise life index large model construction module is specifically used for: The viability index, the risk transmission index and the growth potential index in the dynamic evaluation index set are arranged in a time series of a preset time window length, the index value of each time window is normalized, and the time dimension information is added through sine function position coding to generate a model input tensor; The model input tensor is input into the preset Transformer architecture to output the multi-dimensional enterprise life index; wherein the encoder unit uses a multi-head self-attention mechanism to calculate the dependency weights of each indicator across time windows to capture the nonlinear state association law; the decoder unit fuses the indicator value of the current time window with the industry benchmark data to generate the survival probability through autoregressive prediction; Based on the preset industry risk scenarios corresponding to the target enterprise, an anti-disturbance factor is generated and embedded in the model to train the enterprise life index model. Based on the historical risk event data of the target enterprise, the risk threshold level of the target enterprise is determined, and the trend deviation between the current index value of the target enterprise and the industry benchmark average is calculated.

5. The system according to claim 1, characterized in that The decision strategy tree model is a hierarchical decision model built based on a rule engine and a reinforcement learning framework; the intelligent decision dynamic generation module is specifically used for: Constructing the hierarchical decision model based on the fusion of rule engine and reinforcement learning, inputting the survival probability, the risk threshold level and the trend deviation into the hierarchical decision model, extracting strategy instructions from the leaf nodes matched by the hierarchical decision model, and generating the structured decision instruction set; The reward value is calculated according to the actual effect after the execution of the decision instruction set, and the strategy selection weight of the hierarchical decision model is updated based on the reinforcement learning framework to dynamically update the hierarchical decision model.

6. The system according to claim 1, characterized in that The actual execution data includes: supply chain adjustment effect data, financing strategy execution data and crisis response effect data; the closed-loop feedback optimization module is specifically used for: The supply chain adjustment effect data, the financing strategy execution data and the crisis response effect data are aligned with the predefined decision-making objectives, quantitative evaluation indicators are generated, and the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set are dynamically corrected in combination with the Bayesian optimization algorithm.

7. An intelligent decision-making method based on a large model of enterprise life index, characterized in that: include: Acquire the target enterprise's structured financial data, unstructured public opinion data, supply chain time series data, and industry benchmark data in real time, and use data lineage tracking technology to perform cross-modal correlation mapping of the data to generate multi-source heterogeneous data of the target enterprise; Utilize graph neural network to sort the feature importance of the multi-source heterogeneous data, and generate an extensible dynamic evaluation indicator set according to preset industry classification rules; wherein the dynamic evaluation indicator set includes viability indicators, risk transmission indicators and growth potential indicators; The enterprise life index big model with a preset Transformer architecture is used to integrate the dynamic evaluation indicator set, and the state evolution law of the target enterprise is captured through the temporal attention mechanism of the enterprise life index big model, and the industry risk disturbance factor generated by adversarial training is embedded, and a multi-dimensional enterprise life index including survival probability, risk threshold level and trend deviation is output; Input the multi-dimensional enterprise life index into the pre-trained decision strategy tree model, dynamically match the risk response strategy library based on the reinforcement learning framework, and generate a decision instruction set including financing suggestions, supply chain adjustment plans and crisis warning levels; The actual execution data after the execution of the decision instruction set is obtained through the decision effect collection interface deployed on the business terminal, and based on the actual execution data, the weight parameters of the enterprise life index model and the composition structure of the dynamic evaluation indicator set are dynamically corrected in combination with the Bayesian optimization algorithm.

8. The method according to claim 7, characterized in that The step of acquiring structured financial data, unstructured public opinion data, supply chain time series data and industry benchmark data of the target enterprise in real time, and performing cross-modal association mapping of the data through data lineage tracking technology to generate multi-source heterogeneous data of the target enterprise includes: Retrieving the structured financial data from the target enterprise's ERP system through an API interface, and using web crawler technology to obtain the unstructured public opinion data, and based on IoT sensors, obtaining the supply chain time series data, and based on third-party database data, obtaining the industry benchmark data of the target enterprise; Perform field mapping and time granularity alignment on the structured financial data to obtain standard financial data, perform text sentiment analysis and entity extraction on the unstructured public opinion data to obtain standard public opinion data, perform anomaly detection and time window aggregation on the supply chain time series data to obtain standard supply chain data, perform unit consistency and time window matching on the industry benchmark data to obtain standard industry benchmark data; Add metadata tags to each type of standardized data, and build a blood relationship map between data entities based on the graph database; each type of standardized data includes: the standard financial data, the standard public opinion data, the standard supply chain data, and the standard industry benchmark data; Real-time data association is triggered through a rule engine to automatically update the blood relationship map, and each type of standardized data is encapsulated in a preset format to obtain the multi-source heterogeneous data of the target enterprise.

9. An electronic device, characterized in that: The electronic device includes a processor, which is coupled to a memory. The memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor so that the electronic device implements the intelligent decision-making method based on the enterprise life index model as described in claim 7 or 8.

10. A computer-readable storage medium, characterized in that: At least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by the processor so that the computer-readable storage medium implements the intelligent decision-making method based on the enterprise life index large model as described in claim 7 or 8.

Images