An enterprise risk prediction and early warning method and system based on an industry chain security mechanism

Abstract

The present application provides an enterprise risk prediction and early warning method and system based on an industry chain security mechanism, which comprises the following steps: obtaining text data related to a target industry; preprocessing the text data and extracting features; based on expert rules and computable structured features, describing each element of a specific policy and analyzing the influence of the policy on each link of the industry chain to identify possible risk points of the enterprise; based on the current data and the extracted features, machine learning training is performed to identify potential policy risks and their influence on the enterprise; based on the identification result, risk scores are given to each enterprise; and the evaluation result and decision support are output. By structuring and coding the policy tool elements and combining with the training of industry competition and enterprise feature data, the present application realizes the early identification and quantitative evaluation of the external constraints that the enterprise may face in the future, thereby improving the foresight and effectiveness of risk response.

Description

Technical Field

[0001] This application relates to the field of enterprise risk prediction technology, and in particular to an enterprise risk prediction and early warning method and system based on the supply chain security mechanism. Background Technology

[0002] Among current related technologies, 1) Policy text analysis: Existing technologies use Natural Language Processing (NLP) to analyze policies, regulations, and news reports, extracting features such as keywords, sentiment, and threatening semantics to identify policy risk factors; 2) Sentiment analysis and mood detection: Existing technologies use sentiment analysis to quantify negative emotions in texts and assess whether the text contains potentially threatening language, such as "sanctions" or "blockades," which are important bases for judging policy risks; 3) Machine learning and risk prediction models: Existing risk assessment methods use machine learning algorithms (such as random forests and support vector machines) to model historical events and data to help assess the impact of certain risk events on enterprises. However, these methods rely on historical data and are difficult to effectively predict future risks. Summary of the Invention

[0003] The technical solution adopted in this invention is to provide enterprises with a forward-looking risk prediction and early warning system based on policy and industry competition / threat perception information. This system can predict and identify external pressures that may affect enterprise operations in advance, and while providing risk prediction results, it also provides explanations of the corresponding risk links and mechanisms, illustrating the matching combinations of industry competition factors, threat signals, policy tools, and enterprise characteristics from which the risks originate. External risk pressures for enterprises mainly come from policy changes, industry sanctions, technology export restrictions, and investment reviews. In view of this, this invention provides a method and system for enterprise risk prediction and early warning based on supply chain security mechanisms.

[0004] The present invention proposes a method for predicting and warning enterprise risks based on a supply chain security mechanism, comprising: Step 1: Obtain text data related to the target industry; Step 2: Preprocess the text data and extract its features; Step 3: Based on expert rules and computable structured features, characterize the elements of a specific policy and analyze the impact of the policy on each link of the industrial chain, thereby identifying potential risk points for enterprises. Step 4: Based on the current data and extracted features, perform machine learning training to identify potential policy risks and their impact on enterprises; Step 5: Based on the identification results, assign a risk score to each enterprise; Step 6: Output the evaluation results and decision support.

[0005] In one implementation, the text data includes policy documents, news reports, regulations, and think tank reports.

[0006] In one implementation, the text data is preprocessed and features are extracted, including: word segmentation, sentiment analysis and keyword extraction of the text data, and the extracted text features include: policy intervention intensity, threat semantic score and risk expectation density.

[0007] In one implementation, characterizing the elements of a particular policy includes: The elements of specific policy tools, including the applicable targets, the strength of constraints, the triggering conditions, and the implementation methods of export bans and investment reviews, are characterized.

[0008] In one implementation, step 4 involves using an XGBoost model for machine learning training.

[0009] In one implementation, the risk scoring for each enterprise includes: Enterprises are categorized based on the external pressures they face: high risk, medium-high risk, medium risk, and low risk. The risk score is calculated based on supply chain analysis, policy tool characteristics, enterprise characteristics, and historical event data.

[0010] In one implementation, the output evaluation results are used in decision support: Output the enterprise's risk assessment report, which includes macro risk trends, industry distribution, enterprise risk level and its change trajectory, and conducts a chain analysis and mechanism explanation of the enterprise's specific risks, clarifying the risk source elements and their transmission paths.

[0011] Another aspect of the present invention provides an enterprise risk prediction and early warning system based on a supply chain security mechanism, comprising: The data acquisition and integration module is used to acquire text data related to the target industry; The feature extraction and preprocessing module is used to preprocess the text data and extract features. The supply chain risk analysis module is used to characterize the various elements of a specific policy based on expert rules and computable structured features, and to analyze the impact of the policy on each link of the supply chain, thereby identifying potential risk points for enterprises. The risk prediction model and training module are used to train machine learning based on current data and extracted features to identify potential policy risks and their impact on enterprises. The risk assessment and early warning module is used to score the risk of each enterprise based on the identification results; The Output Results and Decision Support module is used to output scoring results and decision support.

[0012] Another aspect of the present invention provides an electronic device, including: a memory and a processor, wherein the memory is configured with a computer program, and the processor is configured to execute the computer program to implement the enterprise risk prediction and early warning method based on the supply chain security mechanism as described in any of the preceding claims.

[0013] Another aspect of the present invention provides a computer storage medium on which a computer program is configured, the computer program being executed to implement the enterprise risk prediction and early warning method based on the supply chain security mechanism as described in any of the preceding claims.

[0014] By adopting the above technical solution, the present invention has at least the following advantages: The method provided by this invention, through the fusion of multi-source heterogeneous data and machine learning modeling, can predict in advance the policy constraints, sanctions risks and other external pressures that enterprises may face. While outputting the enterprise risk prediction results, it also provides explanations of risk links and mechanisms, thereby providing a scientific and forward-looking solution that can help enterprises identify potential risk sources and their transmission paths, thereby improving the efficiency and accuracy of risk management. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of the enterprise risk prediction and early warning method based on the supply chain security mechanism according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the model training logic according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the risk time extraction logic under the time window according to an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the implementation process of the enterprise risk prediction and management module according to an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments.

[0017] While exemplary embodiments of the invention are shown in the accompanying drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey its scope to those skilled in the art. The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0018] The first embodiment of this application provides a method for predicting and warning enterprise risks based on a supply chain security mechanism, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain text data related to the target industry; Step 2: Preprocess the text data and extract its features; Step 3: Based on expert rules and computable structured features, characterize the elements of a specific policy and analyze the impact of the policy on each link of the industrial chain, thereby identifying potential risk points for enterprises. Step 4: Based on the current data and extracted features, perform machine learning training to identify potential policy risks and their impact on enterprises; Step 5: Based on the identification results, assign a risk score to each enterprise; Step 6: Output the evaluation results and decision support.

[0019] The following will be combined with the appendix Figure 2-4 The method provided in this application will be described in detail.

[0020] The implementation process of the technical solution of the enterprise risk prediction and early warning method based on the supply chain security mechanism proposed in this invention is as follows: Figure 2 As shown in the diagram. The core objective of this solution is to predict the level of risk faced by enterprises when encountering external pressures such as policy and industry sanctions by integrating multi-source data and applying machine learning algorithms, and to output explanations of risk chains and mechanisms corresponding to the level of risk. Each process module will be described in detail below.

[0021] 1) Data collection and integration Figure 2 The left side displays multiple data source modules, including policy documents, news reports, and think tank reports. These data sources provide the foundational information for risk prediction. First, the policy data source module obtains policy documents and regulations related to the target industry. In addition, the news and media reporting module captures the latest industry-related developments, ensuring the timeliness and comprehensiveness of the data.

[0022] 2) Feature extraction and preprocessing The collected data undergoes preprocessing, including word segmentation, sentiment analysis, and keyword extraction. These steps identify and extract key features such as policy intervention intensity, threat semantic score, and risk expectation density. This processing ensures the extraction of efficient features relevant to risk assessment from large-scale text data for subsequent model training.

[0023] 3) Supply Chain Risk Analysis Module After feature extraction, the data enters the supply chain risk analysis module. This module assesses the potential risks faced by enterprises by analyzing their position in the supply chain and their upstream and downstream relationships. Based on expert rules and computable structured features, this process characterizes the applicable objects, binding strength, triggering conditions, and implementation methods of specific policy tools (such as export bans and investment reviews), and analyzes the impact of these policies on each link of the supply chain, thereby identifying potential risk points for enterprises.

[0024] 4) Risk prediction model and training After data processing and feature extraction, the data proceeds to the risk prediction module. This module employs ensemble learning models based on gradient boosting trees (such as XGBoost and AutoGluon) for machine learning training to identify potential policy risks and their impact on enterprises, thereby better handling the complex data characteristics of the multi-dimensional graph of the industrial chain. XGBoost is an efficient gradient boosting algorithm capable of handling non-linear relationships and providing accurate risk scoring. At this stage, the AutoGluon framework is used for automated model optimization, enhancing the system's adaptability to different types of data.

[0025] 5) Risk assessment and early warning After model training, the risk assessment module assigns a risk score to each enterprise. The assessment results are categorized based on the external pressures faced by the enterprise: high risk, medium-high risk, medium risk, and low risk. The risk score is calculated based on factors such as supply chain analysis, policy tool characteristics, enterprise characteristics, and historical event data. The enterprise's risk level is updated in real time and promptly notified to relevant personnel through a predictive early warning system. Simultaneously, the system's linked "risk link analysis module" automatically generates a linked explanation of the "policy tools—enterprise-side risk elements—event-side risk characteristics" chain (including key driving factors, transmission paths, and corresponding evidence fragments / feature contributions) based on the enterprise's risk score, supporting traceable and verifiable risk assessment and response decisions.

[0026] 6) Output Results and Decision Support Finally, this technology outputs a risk assessment report for the enterprise, including macroeconomic risk trends, industry distribution, enterprise risk level and its trajectory, and provides a chain-based analysis and mechanism explanation of specific risks, clarifying the risk source elements and their transmission paths. Enterprises can make corresponding decisions based on these reports, such as adjusting the supply chain and optimizing resource allocation, to reduce the potential negative impact of policy changes.

[0027] Figure 3 This demonstrates the technical process for extracting and assessing risk events from multi-source data within a specific time window. The modules and their functions in the process are as follows: 1) Data collection and integration The left side of the workflow displays the multi-source data acquisition module, where the system obtains the latest policy documents, news reports, regulations, and think tank reports from multiple data sources. These data sources are integrated through a data interface to ensure data integrity and timeliness.

[0028] 2) Division and application of time windows This step involves dividing data sources such as policies and news into time windows to generate event data for different time periods. Specifically, the system filters and categorizes data according to set time windows (such as the past month, three months, or six months) to ensure that all risk events within the assessment timeframe are effectively covered.

[0029] 3) Risk event identification and screening Within the time window, the system performs event identification and risk screening on the collected text data. Using natural language processing technology, the system identifies relevant risk events such as policy changes, sanctions, and market shifts. This process is completed through an event classification module, which pre-classifies the identified events (e.g., technology export bans, investment reviews) and labels them with their respective risk types.

[0030] 4) Risk assessment and weight calculation After identifying risk events, the system enters the risk assessment module, which calculates and adjusts the weights of different risk events to evaluate the potential impact of each event. For example, the system assigns different weight scores to each event based on its source, degree of impact, and timeliness, thereby determining its specific impact on the enterprise.

[0031] 5) Intelligent recommendation and decision support Finally, the data after risk assessment enters the decision support module. The system combines machine learning algorithms and industry expert rules to generate a risk assessment report for the enterprise based on the event's scoring results and weights. It also outputs the corresponding risk pathways and causal explanations to clarify the sources and transmission paths of risks, providing intelligent risk warnings. Enterprises can then take corresponding preventative measures against potential risks based on these assessment results, such as adjusting their supply chain structure or optimizing their investment strategies.

[0032] 6) Output Results and Feedback The assessment results will be presented in the form of a report through the output module. The report includes specific risks faced by the enterprise and potential sources of policy pressure, as well as corresponding strategies and risk response recommendations. Furthermore, the system supports a feedback mechanism based on historical events to continuously optimize and adjust the assessment model, improving the accuracy and real-time nature of risk assessment.

[0033] Figure 4 The flowchart in the document illustrates the implementation process of the enterprise risk prediction and management module: 1) Data collection and integration The left side of the flowchart shows the data acquisition module, where the system obtains relevant information from multiple sources (such as corporate financial data, policy documents, news reports, etc.). These data sources are integrated through different interfaces and modules to generate the dataset required for corporate risk assessment. Data integration includes the company's historical financial data, industry development trends, and external policy changes.

[0034] 2) Risk event identification and classification The system identifies and classifies risk events based on the collected data. This step includes: Risk event extraction: Extract potential risk events related to the enterprise from various data sources, such as changes in external policies, market fluctuations, and technological updates.

[0035] Risk classification: Identified risk events are classified according to their nature and impact, such as economic policy risks, market risks, and technological risks.

[0036] 3) Risk assessment and analysis After risk events are categorized, they proceed to the risk assessment and analysis module. In this module, the system employs a comprehensive analysis method based on multi-source data to assess each risk event. By evaluating factors such as the event's timeliness, scope of impact, and the company's exposure level, the system calculates the severity of each risk and assigns a risk score to each event. During the assessment process, the system also incorporates the MACD model (Moving Average Convergence / Divergence Index) to dynamically track risks and predict risk trends.

[0037] 4) Risk Link Identification and Transmission Path Analysis Based on the risk assessment output, the system calls the risk link analysis module to decompose and attribute the enterprise's risk situation, establish a multi-layered heterogeneous graph representation combining enterprise nodes, policy tool nodes, and risk element nodes, and calculate and generate a traceable risk transmission path based on graph structure relationships and feature contributions; at the same time, it outputs a set of key driving elements and their contribution and weight, realizing the tracing and visualization of the causes of risk.

[0038] 5) Decision support and risk warning Once the risk assessment is complete, the system transmits the results to the decision support module. This module generates a detailed warning report based on different risk levels and warning times. This report provides the enterprise with specific countermeasures, including: Short-term and long-term risk response strategies: Propose response strategies based on the types of risks faced by the enterprise and the predicted impact. Strategy Priority: Based on the severity and urgency of the risk, the system will recommend a priority order and solutions for responding to it.

[0039] 6) Enterprise Risk Management and Control Finally, the system transmits the assessment results and early warning information to the enterprise risk management and control module. This module is responsible for taking specific risk management actions based on the early warning report, including: Resource Adjustment: Based on the risk assessment report, adjust the company's resource allocation and strategic direction; Risk control: Take reinforcement measures for high-risk areas and optimize the company's operating strategies.

[0040] 7) Feedback Mechanism and Model Optimization The system also features a feedback mechanism. By collecting data on the effectiveness of enterprises' actual risk management, the system continuously optimizes its risk assessment model and decision support algorithms to improve accuracy and real-time performance. Feedback information is then transmitted to the data acquisition and integration module to ensure that the model can be adjusted and optimized based on the latest data.

[0041] In the technical solution of this invention, the following five key points are the technical contents that need to be emphasized in this application, respectively involving the feature engineering of information sources and variables, the processing method of time windows, the model training strategy, the quantification of enterprise multi-chain participation, and the risk chain representation. Each key point plays a crucial role in the risk prediction and early warning system of this invention, ensuring the effectiveness and foresight of the technical solution.

[0042] 1) Characterization of information sources and variables In the technical solution of this invention, feature engineering of information sources and variables is the foundation of the risk prediction model. To achieve accurate risk early warning, this invention proposes a method for feature extraction and processing based on multi-source data (such as policy documents, news reports, regulations, and think tank research reports). The main steps include: Diverse data sources: The system collects information from multiple data sources, including policy documents, news reports, think tank research reports, and legal provisions, and processes each data source differently according to its content characteristics.

[0043] Feature extraction: Extracting valuable features from various types of text data, mainly including: Sentiment analysis: Identifying the emotional tone in text and calculating the sentiment score of policy documents or news reports using sentiment lexical and sentiment models.

[0044] Threat semantic scoring: By using keyword matching and contextual analysis, threatening semantics in the text, such as "sanctions," "blockade," and "containment," are extracted to assess the potential threat level of external risks.

[0045] Intensity of policy intervention: This measures the density and intensity of language used in policy texts that emphasize strong guidance, strong supervision, and strong constraints, thereby assessing the policy's effectiveness and potential impact.

[0046] Through this multi-layered feature engineering, the present invention can extract core information reflecting potential risks from a large amount of raw text data, ensuring the accuracy and reliability of risk prediction.

[0047] 2) Handling of time windows The handling of time windows is another innovation of this invention. To improve the accuracy of risk assessment, this invention introduces the concept of a time dimension into the model and optimizes risk prediction through reasonable time window division and processing. This mainly includes the following aspects: Time window segmentation: Based on different business needs and data characteristics, the system supports the segmentation of multiple time windows, such as short-term (e.g., 1 month), medium-term (e.g., 3 months), and long-term (e.g., 6 months, 1 year). These different time windows can capture policy changes and market fluctuations within different time periods, thereby better predicting the short-term and long-term impact of policies on enterprises.

[0048] Lag window and cumulative window mechanism: To fully consider the lag and cumulative effects of risk events, this invention employs two methods: lag window and cumulative window. Lagging window: This uses risk scores from a past period as input features to reflect the lagging impact of policy or market changes on enterprises.

[0049] Cumulative window: Accumulates the risk intensity of all risk events within a certain period of time, simulating the continuous accumulation effect of policy pressure, thereby accurately reflecting the long-term risks faced by enterprises.

[0050] The segmented variable construction and time weighting strategy of this invention characterizes the lagged triggering and phased accumulation process of policy pressure, effectively solving the shortcomings of traditional models that simplify the impact of policy.

[0051] This time window approach ensures that the invention can dynamically adjust and adapt to risk assessments in different time periods, thereby improving the model's ability to predict future uncertain events.

[0052] 3) Model training strategy This invention employs an ensemble learning approach in model training, improving overall prediction accuracy by combining the outputs of multiple models. Specific training strategies include: Multi-model ensemble: The system uses machine learning frameworks such as XGBoost and AutoGluon, combined with bagging and stacking techniques, to train multiple sub-models. The results of each sub-model are weighted and averaged through ensemble learning methods, thereby reducing the bias that may be caused by a single model and improving the robustness and accuracy of the model.

[0053] XGBoost: As a gradient boosting tree algorithm, XGBoost has a strong fitting ability when processing structured data (such as policy texts, industrial chain information, etc.) and can capture non-linear relationships in the data.

[0054] AutoGluon: This framework combines various machine learning algorithms (such as random forest, XGBoost, etc.) and further improves the model's generalization ability and prediction performance through bagging (multiple random sampling) and stacking (multi-layer structure).

[0055] Supervised fine-tuning (SFT) and low-rank adaptation (LoRA): To improve the model's adaptability in enterprise risk prediction tasks, this invention fine-tunes the pre-trained model using supervised fine-tuning (SFT) and low-rank adaptation (LoRA) techniques. This reduces training time and computational resource requirements while maintaining the model's prediction accuracy.

[0056] Gradient accumulation and learning rate scheduling: During training, the system employs gradient accumulation and learning rate scheduling techniques to balance computational resources and training efficiency, while avoiding memory overflow issues in large-scale data processing. The learning rate scheduling uses a cosine annealing strategy, which smoothly reduces the learning rate, enhancing the model's exploratory capabilities during training and preventing oscillations and overfitting.

[0057] This comprehensive training strategy, through multi-model integration and intelligent adjustment, greatly improves the accuracy and real-time performance of risk prediction, ensuring that the system can handle complex multi-source data and cope with dynamically changing risk environments.

[0058] 4) Measuring Enterprise Participation under the Soft Label Mechanism To address the problem that existing systems generally rely on static industry labels or single directories to identify enterprise affiliation, neglecting the participation distribution and role flexibility of enterprises across multiple industry chains, this invention proposes a soft labeling mechanism. By constructing an affiliation probability vector, it accurately expresses the participation degree and role flexibility of enterprises across multiple industry chains. The introduction of weighted calculations during risk aggregation significantly improves the coverage and accuracy of risk scoring results, avoiding generalization and omission issues in risk transmission paths.

[0059] 5) Multi-level map construction and risk link representation This invention achieves in-depth explanation of risk links and mechanisms by constructing a multi-level heterogeneous graph. Specifically, the system nests enterprise nodes, policy tool nodes, and risk element nodes into a structured link, and calculates traceable risk transmission paths based on graph structure relationships and feature contributions. This structured representation supports the visualization of causal paths and the tracing of scoring causes, greatly enhancing the transparency and operability of scoring results and providing a solid basis for enterprises to formulate risk countermeasures.

[0060] The second embodiment of this application provides an enterprise risk prediction and early warning system based on the supply chain security mechanism, including: The data acquisition and integration module is used to acquire text data related to the target industry; The feature extraction and preprocessing module is used to preprocess the text data and extract features. The supply chain risk analysis module is used to characterize the various elements of a specific policy based on expert rules and computable structured features, and to analyze the impact of the policy on each link of the supply chain, thereby identifying potential risk points for enterprises. The risk prediction model and training module are used to train machine learning based on current data and extracted features to identify potential policy risks and their impact on enterprises. The risk assessment and early warning module is used to score the risk of each enterprise based on the identification results; The Output Results and Decision Support module is used to output scoring results and decision support.

[0061] The third embodiment of this application provides an electronic device, including: a memory and a processor, wherein a computer program is configured on the memory, and the processor is used to execute the computer program to implement the enterprise risk prediction and early warning method based on the supply chain security mechanism as described in the first embodiment.

[0062] The fourth embodiment of this application provides a computer storage medium on which a computer program is configured. The computer program is executed to implement the enterprise risk prediction and early warning method based on the supply chain security mechanism as described in the first embodiment of this application.

[0063] Compared to existing technologies, this invention, through multi-source heterogeneous data fusion and machine learning modeling, provides forward-looking predictions and early warnings of enterprise risks, thus overcoming the shortcomings of existing technologies. Most existing technologies rely on historical data for risk assessment and lack the ability to predict future policy changes. This invention, by structurally encoding policy tool elements and combining them with industry competition and enterprise characteristic data for fusion training, achieves early identification and quantitative assessment of potential external constraints that enterprises may face in the future, such as inclusion in lists, export restrictions, and investment reviews. It achieves high-precision risk quantification scoring and provides a clear explanation of risk transmission chains and mechanisms. This enables enterprises to intuitively trace the sources of risk and understand the profound impact of policy changes on their multiple industrial chains, thereby supporting enterprises in deploying compliance, supply chain, and operational contingency plans before policy changes occur, thus improving the foresight and effectiveness of risk response.

[0064] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims. All of these forms are within the protection scope of this application.

Claims

1. A method for predicting and warning enterprise risks based on supply chain security mechanisms, characterized in that, include: Step 1: Obtain text data related to the target industry; Step 2: Preprocess the text data and extract its features; Step 3: Based on expert rules and computable structured features, characterize the elements of a specific policy and analyze the impact of the policy on each link of the industrial chain, thereby identifying potential risk points for enterprises. Step 4: Based on the current data and extracted features, perform machine learning training to identify potential policy risks and their impact on enterprises; Step 5: Based on the identification results, assign a risk score to each enterprise; Step 6: Output the evaluation results and decision support.

2. The enterprise risk prediction and early warning method based on the supply chain security mechanism according to claim 1, characterized in that, The text data includes: policy documents, news reports, regulations, and think tank reports.

3. The enterprise risk prediction and early warning method based on the supply chain security mechanism according to claim 2, characterized in that, The text data is preprocessed and features are extracted, including word segmentation, sentiment analysis and keyword extraction. The extracted text features include: policy intervention intensity, threat semantic score and risk expectation density.

4. The enterprise risk prediction and early warning method based on the supply chain security mechanism according to claim 3, characterized in that, The description of the various elements of a specific policy includes: The elements of specific policy tools, including the applicable targets, the strength of constraints, the triggering conditions, and the implementation methods of export bans and investment reviews, are characterized.

5. The enterprise risk prediction and early warning method based on the supply chain security mechanism according to claim 4, characterized in that, In step 4, the XGBoost model is used for machine learning training.

6. The enterprise risk prediction and early warning method based on the supply chain security mechanism according to claim 5, characterized in that, The risk scoring for each enterprise includes: Enterprises are categorized based on the external pressures they face: high risk, medium-high risk, medium risk, and low risk. The risk score is calculated based on supply chain analysis, policy tool characteristics, enterprise characteristics, and historical event data.

7. The enterprise risk prediction and early warning method based on the supply chain security mechanism according to claim 6, characterized in that, The output evaluation results and decision support include: Output the enterprise's risk assessment report, which includes macro risk trends, industry distribution, enterprise risk level and its change trajectory, and conducts a chain analysis and mechanism explanation of the enterprise's specific risks, clarifying the risk source elements and their transmission paths.

8. A corporate risk prediction and early warning system based on a supply chain security mechanism, characterized in that, include: The data acquisition and integration module is used to acquire text data related to the target industry; The feature extraction and preprocessing module is used to preprocess the text data and extract features. The supply chain risk analysis module is used to characterize the various elements of a specific policy based on expert rules and computable structured features, and to analyze the impact of the policy on each link of the supply chain, thereby identifying potential risk points for enterprises. The risk prediction model and training module are used to train machine learning based on current data and extracted features to identify potential policy risks and their impact on enterprises. The risk assessment and early warning module is used to score the risk of each enterprise based on the identification results; The Output Results and Decision Support module is used to output scoring results and decision support.

9. An electronic device, characterized in that, include: A memory and a processor, wherein the memory is configured with a computer program, and the processor is used to execute the computer program to implement the enterprise risk prediction and early warning method based on the supply chain security mechanism as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The medium is equipped with a computer program, which is executed to implement the enterprise risk prediction and early warning method based on the supply chain security mechanism as described in any one of claims 1 to 7.

Images