An industry chain risk early warning method and system based on multi-source risk signals

Abstract

This invention relates to the field of risk early warning technology, specifically to a method and system for supply chain risk early warning based on multi-source risk signals. The method includes: detecting anomalies in historical and real-time risk time-series data of the supply chain to obtain historical risk events and real-time abnormal fluctuation points; constructing a causal-oriented directed topology of the supply chain based on the upstream and downstream flow relationships of historical risk events to obtain a causal-oriented graph; evaluating the upstream attribution components of the abnormal fluctuation points based on the causal-oriented graph to obtain causal effect values; tracing the risk transmission of abnormal fluctuation points based on the causal effect values ​​and the causal-oriented graph to obtain risk source points and risk propagation paths; performing counterfactual blocking inference in the causal-oriented graph based on the risk source points and risk propagation paths to obtain risk propagation path blocking data; and compiling the blocking data information to obtain supply chain risk early warning information. This invention can improve the efficiency of supply chain risk early warning based on multi-source risk signals.

Description

Technical Field

[0001] This invention relates to the field of risk warning technology, and in particular to a method and system for supply chain risk warning based on multi-source risk signals. Background Technology

[0002] As an interdependent cluster of enterprises, the supply chain is susceptible to systemic risks due to operational fluctuations at any node, which can be transmitted and amplified through upstream and downstream supply and demand relationships. Therefore, continuous monitoring of the supply chain's operational status and accurate identification and tracing of the source and propagation path of risks are crucial for ensuring the safe and stable operation of the supply chain. Current technologies typically involve collecting operational data from each node of the supply chain and combining this data with statistical methods or pre-defined rules for anomaly detection and risk analysis.

[0003] Existing supply chain risk monitoring technologies lack effective handling of noise interference such as missing values ​​and outliers commonly found in raw time-series data. This limits the accuracy of anomaly detection, easily leading to false alarms caused by misjudging collection errors as operational risks. Furthermore, it struggles to integrate discrete fluctuations into anomalous events with clear operational implications, resulting in a lack of reliable data support for subsequent analysis. Moreover, risk transmission analysis is often based on static upstream-downstream relationships for unidirectional judgments, lacking the mining and quantification of complex causal relationships in historical risk events. This makes it impossible to accurately assess the actual contribution of upstream nodes and the blocking effect of different intervention points, leading to vague risk source identification and distorted propagation path characterization. Consequently, it fails to meet the actual needs of precise source tracing and risk management. Therefore, improving the efficiency of supply chain risk early warning has become an urgent problem to be solved. Summary of the Invention Summary of the Invention

[0004] This invention provides a method and system for early warning of supply chain risks based on multi-source risk signals, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a supply chain risk early warning method based on multi-source risk signals, comprising: S1. Detect anomalies in the historical and real-time risk time series data of the industrial chain to obtain the historical risk events and real-time abnormal fluctuation points of the industrial chain. S2. Based on the upstream and downstream flow relationships in the historical risk events, a directed topology is constructed for the causal orientation of the industrial chain to obtain the causal orientation map of the industrial chain. S3. Based on the causal pointing map, perform upstream attribution component evaluation on the abnormal fluctuation points to obtain the causal effect value of the abnormal fluctuation points. S4. Based on the causal effect value and the causal pointing map, the source of risk transmission is traced at the abnormal fluctuation points to obtain the risk source points and risk propagation paths of the industrial chain; S5. Based on the risk source location and the risk propagation path, perform counterfactual blocking inference in the causal pointing map to obtain the blocking data of the risk propagation path; S6. Compile the blocked data to obtain the industrial chain risk warning information.

[0006] In a preferred embodiment, the step of detecting anomalies in historical and real-time risk time-series data of the industrial chain to obtain historical risk events and real-time abnormal fluctuation points of the industrial chain includes: The operational log data of the nodes in the industry chain are continuously collected to obtain the risk time series data of the industry chain; The risk time series data is smoothed and denoised to obtain the purified risk data of the industrial chain; By detecting change points in the purified risk data, the discrete risk fluctuation points of the industrial chain can be obtained. According to the preset time-series classification rules, the initial abnormal fluctuation points are classified to obtain the historical fluctuation points and real-time fluctuation points of the industrial chain. By extracting events from the aforementioned historical fluctuation points, the historical risk events of the industrial chain can be obtained. The real-time fluctuation points are identified as anomalies to obtain the real-time abnormal fluctuation points of the industrial chain.

[0007] In a preferred embodiment, the step of constructing a directed topology of the causal orientation of the industrial chain based on the upstream and downstream flow relationships in the historical risk events to obtain a causal orientation graph of the industrial chain includes: The flow of the historical risk events is identified to obtain the original flow relationship pairs of the industrial chain; Based on the sequence of occurrence of the historical risk events, the original flow relationship pairs are verified to obtain the refined causal pair of the industrial chain; The refined causal pairs are subjected to a reliable selection process to obtain a list of reliable causal pairs for the industrial chain. Based on the aforementioned confidence causal pointing list, the nodes in the industrial chain are hierarchically arranged to obtain the hierarchical node sequence of the industrial chain; Based on the hierarchical node sequence and the confidence causal pointing list, the upstream and downstream dependencies between the nodes are topologically reconstructed to obtain the causal pointing graph of the industrial chain.

[0008] In a preferred embodiment, the step of evaluating the upstream attribution components of the anomalous fluctuation points based on the causal pointing map to obtain the causal effect value of the anomalous fluctuation points includes: Feature mining is performed on the fluctuation attributes of the abnormal fluctuation points to obtain the fluctuation amplitude characteristics and fluctuation time markers of the abnormal fluctuation points. Based on the topological connection relationship of the causal pointing graph, the upstream transmission path of the abnormal fluctuation point is traced to obtain the upstream node group of the abnormal fluctuation point. Historical load assessment is performed on the upstream node group to obtain the conduction strength parameters of the upstream node group; The path attenuation factor of the upstream transmission path is obtained by analyzing the upstream transmission path between the upstream node group and the abnormal fluctuation point. The causal effect value of the abnormal fluctuation point is obtained by attributing and quantifying the fluctuation amplitude characteristics, the transmission strength parameter, and the path attenuation factor.

[0009] In a preferred embodiment, the formula for calculating the causal effect value is as follows: ; In the formula, where, The causal effect value is... The fluctuation amplitude characteristic, The number of upstream nodes in the upstream node group. For the first The transmission strength parameters from each upstream node to the abnormal fluctuation point. For the causal pointing graph, the first... The upstream transmission path between each upstream node and the abnormal fluctuation point. For the nodes in the upstream transmission path Path decay factor.

[0010] In a preferred embodiment, the step of tracing the source of risk transmission at the abnormal fluctuation points based on the causal effect value and the causal pointing map to obtain the risk source points and risk propagation paths of the industrial chain includes: Based on the causal pointing graph, the upstream adjacency of the abnormal fluctuation point is traced to obtain the upstream adjacency node of the abnormal fluctuation point.

[0011] Based on the causal effect value, the influence contribution of the upstream adjacent nodes is evaluated to obtain the upstream attribution node record of the abnormal fluctuation point. By performing hierarchical backtracking on the nodes in the upstream attribution node record along the causal pointing graph, the complete upstream tracing chain of the abnormal fluctuation point is obtained; By marking the source points of the complete upstream traceability chain, the risk source locations of the industrial chain can be obtained; Based on the node pointing relationship in the complete upstream traceability chain, the path from the risk source point to the abnormal fluctuation point is plotted to obtain the risk propagation path of the industrial chain.

[0012] In a preferred embodiment, the step of performing counterfactual blocking inference on the causal pointing map based on the risk source location and the risk propagation path to obtain blocking data for the risk propagation path includes: The path nodes along the risk propagation path are selected based on their betweenness numbers to obtain a high betweenness number node group for the risk propagation path. The blocking effect of candidate nodes in the high betweenness number node group is simulated to obtain the simulated blocking data of the candidate node group; Based on the simulated blocking data, the blocking coverage rate and blocking implementation cost of the candidate nodes are comprehensively compared to obtain the optimal blocking node for the risk propagation path; The blocking effect of the optimal blocking node is verified to obtain blocking data that blocks the risk propagation path.

[0013] In a preferred embodiment, the step of verifying the blocking effect of the optimal blocking node to obtain blocking data that blocks the risk propagation path includes: The influence domain of the optimal blocking node is delineated to obtain the blocking influence area of ​​the optimal blocking node; Based on the area affected by the blockade, the dependent edges of the optimal blockade node are cut off to obtain the post-blockade dependency relationship of the industrial chain; Based on the post-blocking dependency relationship, a propagation blocking assessment is performed on the area affected by the blocking to obtain the blocking data of the risk propagation path.

[0014] In a preferred embodiment, the step of compiling the blocked data to obtain the supply chain risk warning information includes: The blocked data is semantically parsed to obtain the structured compiled fields of the blocked data; Based on the structured compiled fields, the risk propagation path is condensed into text to obtain a textual risk description of the industrial chain; The structured compiled fields and the textual risk description are encapsulated in a message to obtain the industry chain risk warning information.

[0015] To address the aforementioned problems, this invention also provides a supply chain risk early warning system based on multi-source risk signals, the system comprising: The risk anomaly detection module is used to detect anomalies in the historical and real-time risk time series data of the industrial chain, and to obtain the historical risk events and real-time abnormal fluctuation points of the industrial chain. The causal topology construction module is used to construct a directed topology of the causal orientation of the industrial chain based on the upstream and downstream flow relationships in the historical risk events, so as to obtain the causal orientation map of the industrial chain. The causal effect assessment module is used to assess the upstream attribution components of the abnormal fluctuation points based on the causal pointing map, and obtain the causal effect value of the abnormal fluctuation points. The risk transmission and tracing module is used to trace the source of risk transmission based on the causal effect value and the causal orientation map, so as to obtain the risk source location and risk propagation path of the industrial chain. The risk blocking simulation module is used to perform counterfactual blocking simulation on the causal pointing map based on the risk source location and the risk propagation path to obtain blocking data of the risk propagation path; The early warning information compilation module is used to compile the blocked data to obtain the industrial chain risk early warning information.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention performs smoothing and noise reduction, missing value completion, and outlier replacement on historical risk time-series data. The purified data is then identified as discrete fluctuation points through change point detection and time-series classification. Based on operational attributes, continuous fluctuation points at the same node are integrated into historical risk events with clear operational implications, such as sudden drop in production capacity and inventory backlog. This solves the problems of high noise interference and difficulty in parsing discrete points in the original time-series data, providing real and traceable evidence of abnormal events for subsequent causal mapping, and significantly improving the basic quality of risk data.

[0017] 2. This invention relies on integrated historical risk events to extract upstream and downstream flow relationships and construct a causal orientation map with credibility levels through time-series verification and credibility selection. Based on this, it calculates the causal effect value of real-time abnormal fluctuation points, including fluctuation amplitude, upstream node transmission strength, and path attenuation factor. Then, through upstream attribution and counterfactual blocking inference, it accurately locates the risk source points and the optimal blocking nodes. This breaks through the limitations of traditional static correlation analysis, realizes the quantitative evaluation and intervention effect simulation of complex causal relationships, solves the problems of unclear risk sources and distorted propagation paths, and provides scientific decision support for accurate source tracing and efficient handling. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a supply chain risk early warning method based on multi-source risk signals, provided as an embodiment of the present invention. Figure 2 A functional block diagram of a supply chain risk early warning system based on multi-source risk signals provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides a supply chain risk early warning method based on multi-source risk signals. The execution entity of this supply chain risk early warning method based on multi-source risk signals includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application embodiment: a server, a terminal, etc. In other words, the supply chain risk early warning method based on multi-source risk signals can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating a supply chain risk early warning method based on multi-source risk signals according to an embodiment of the present invention. In this embodiment, the supply chain risk early warning method based on multi-source risk signals includes: S1. Detect anomalies in the historical and real-time risk time series data of the industrial chain to obtain the historical risk events and real-time abnormal fluctuation points of the industrial chain. In this embodiment of the invention, the step of detecting anomalies in historical and real-time risk time-series data of the industrial chain to obtain historical risk events and real-time abnormal fluctuation points of the industrial chain includes: The operational log data of the nodes in the industry chain are continuously collected to obtain the risk time series data of the industry chain; The risk time series data is smoothed and denoised to obtain the purified risk data of the industrial chain; By detecting change points in the purified risk data, the discrete risk fluctuation points of the industrial chain can be obtained. According to the preset time-series classification rules, the initial abnormal fluctuation points are classified to obtain the historical fluctuation points and real-time fluctuation points of the industrial chain. By extracting events from the aforementioned historical fluctuation points, the historical risk events of the industrial chain can be obtained. The real-time fluctuation points are identified as anomalies to obtain the real-time abnormal fluctuation points of the industrial chain.

[0022] Operational log data consists of real-time records generated by each node in the industrial chain throughout the entire process of production, distribution, and operation. It includes node capacity data, supply chain transaction data, inventory data, order completion data, equipment operation data, etc. The operational log data of each level of the industrial chain is automatically collected at fixed time frequencies. Production and manufacturing nodes are collected hourly, and distribution and service nodes are collected minutely. During the collection process, node identifiers and timestamps are added to the log data of each node. The node identifier includes the node's hierarchical attributes and functional attributes in the industrial chain, and the timestamp is accurate to the specific time of collection. The operational log data of each node in the industrial chain with node identifiers and continuous timestamps are integrated to form a data set. This data set presents the continuous change status of the operational data of each node with time as the axis. This data set is the risk time series data of the industrial chain.

[0023] First, missing data in the risk time series data is filled in by selecting the average of normal operating data adjacent to the corresponding node. Then, outlier data in the risk time series data is identified by comparing the data value in a single time series data with the historical normal fluctuation range of the same type of operating data for that node. Values ​​exceeding three times or more of the historical normal fluctuation range are identified as outlier data. Subsequently, the identified outlier data is replaced by selecting the average of the two valid data before and after the original data. Finally, the completed and replaced risk time series data is continuously fitted to eliminate abrupt fluctuations in the data on the time axis, making the data curve show a continuous and smooth trend. After the above processing, the risk time series data of the industrial chain has no missing data and outlier values, and can truly reflect the actual change pattern of the operating data of each node in the industrial chain. This data is the purified risk data of the industrial chain.

[0024] A time-series coordinate system for the purified risk data is constructed with time as the horizontal axis and operational data values ​​as the vertical axis. Continuous trend monitoring of operational data at each node is performed within this coordinate system. The slope of operational data change within adjacent time intervals is calculated. Time points where the absolute value of the slope exceeds a preset stability threshold are marked as candidate change points. The preset stability threshold is determined based on the normal range of slope changes in historical operational data of each node in the industry chain. Candidate change points are then continuously verified by checking whether the operational data change trend of the candidate change point continues to deviate from the original stable trend for the next three consecutive timestamps. Candidate change points that continue to deviate are determined as valid change points. All valid change points are integrated into a set, with each point containing a corresponding node identifier, a specific timestamp, and a data change slope. This set represents the discrete risk fluctuation points of the industry chain.

[0025] The initial abnormal fluctuation points, also known as discrete risk fluctuation points, are determined by first setting a fixed time boundary. This time boundary is 24 hours prior to the current data processing time. This time period is set based on the risk monitoring frequency and risk response needs of the industry chain. Fluctuation points before the time boundary are classified as historical, while fluctuation points after the time boundary until the current time are classified as real-time. This rule is the preset time-series classification rule. Then, the timestamp information of each initial abnormal fluctuation point is extracted and compared with the preset time boundary. Points with timestamps earlier than the time boundary are determined as historical fluctuation points, while points with timestamps later than or equal to the time boundary are determined as real-time fluctuation points. Historical fluctuation points are the set of initial abnormal fluctuation points with timestamps earlier than the preset time boundary, reflecting past operational data trend changes in the industry chain. Real-time fluctuation points are the set of initial abnormal fluctuation points with timestamps later than or equal to the preset time boundary, reflecting recent operational data trend changes in the industry chain up to the present.

[0026] First, historical fluctuation points are grouped by node identifier, with historical fluctuation points of the same industry chain node grouped together. Then, the historical fluctuation points within each group are sorted by timestamp, and the continuous change characteristics of the sorted points are analyzed. Combined with the operational attributes of the node, the operational anomaly type corresponding to the point change is determined. Operational anomaly types include sudden drop in production capacity, inventory backlog, order default, equipment failure, etc. Subsequently, continuous historical fluctuation points under the same operational anomaly type are integrated into a complete anomaly event, and an event identifier is added to this anomaly event. The event identifier includes the node information corresponding to the event, the anomaly type, the time interval of occurrence, and the data change characteristics. Finally, all integrated anomaly events are summarized to form a set. Each event in this set contains complete node information, anomaly type, time interval, and data change characteristics. This set is the historical risk event of the industry chain.

[0027] First, extract the node operation data change characteristics and data change slope corresponding to each real-time fluctuation point. Compare the data change slope with the emergency warning slope threshold of the same type of operation data of the node. The emergency warning slope threshold is determined based on the slope value that triggered actual operation anomalies in the node's historical operation data. Then, combine the upstream and downstream correlation of the node in the industry chain to analyze whether the operation data change of the real-time fluctuation point will have a transmission impact on its direct upstream and downstream nodes. Compare the magnitude of the node's operation data change with the adaptability and tolerance range of upstream and downstream nodes. If it exceeds the range, it is determined that it will have a transmission impact. The adaptability and tolerance range of upstream and downstream nodes is determined based on the supply and demand cooperation agreement between nodes and historical collaborative operation data. Real-time fluctuation points that meet the conditions of data change slope exceeding the emergency warning slope threshold and having a transmission impact on upstream and downstream nodes are determined as valid real-time abnormal fluctuation points. Real-time fluctuation points that only meet a single condition or do not meet any conditions are directly eliminated. Integrate all valid real-time abnormal fluctuation points into a set. Each point includes a node identifier, timestamp, data change characteristics, slope value, and upstream and downstream transmission impact range. This set is the real-time abnormal fluctuation point of the industry chain.

[0028] The beneficial effects include: achieving full and continuous collection of operational data from all nodes in the industrial chain, establishing a complete risk data foundation with time attributes, improving the comprehensiveness and continuity of industrial chain risk data collection, eliminating missing values ​​and outliers in risk time series data, avoiding interference from invalid data, improving the authenticity and effectiveness of industrial chain risk data, providing accurate data support for subsequent change point detection, accurately identifying nodes and time points of significant changes in industrial chain operational data trends, improving the accuracy and effectiveness of identifying industrial chain risk fluctuation points, avoiding misjudgment of normal data fluctuations, achieving accurate classification of fluctuation points, allowing industrial chain risk analysis to be carried out separately according to historical and real-time dimensions, improving the pertinence and execution efficiency of industrial chain risk analysis, integrating discrete historical fluctuation points into complete operational anomaly events, improving the traceability and analyzability of historical risks in the industrial chain, providing specific risk event basis for the construction of causal pointing maps, eliminating invalid fluctuation points without actual transmission impact, improving the accuracy of real-time risk anomaly identification in the industrial chain, avoiding invalid warnings caused by false risk signals, and enhancing the reliability and practicality of industrial chain risk warnings.

[0029] S2. Based on the upstream and downstream flow relationships in the historical risk events, a directed topology is constructed for the causal orientation of the industrial chain to obtain the causal orientation map of the industrial chain. In this embodiment of the invention, the step of constructing a directed topology of the causal orientation of the industrial chain based on the upstream and downstream flow relationships in the historical risk events to obtain the causal orientation graph of the industrial chain includes: The flow of the historical risk events is identified to obtain the original flow relationship pairs of the industrial chain; Based on the sequence of occurrence of the historical risk events, the original flow relationship pairs are verified to obtain the refined causal pair of the industrial chain; The refined causal pairs are subjected to a reliable selection process to obtain a list of reliable causal pairs for the industrial chain. Based on the aforementioned confidence causal pointing list, the nodes in the industrial chain are hierarchically arranged to obtain the hierarchical node sequence of the industrial chain; Based on the hierarchical node sequence and the confidence causal pointing list, the upstream and downstream dependencies between the nodes are topologically reconstructed to obtain the causal pointing graph of the industrial chain.

[0030] Extract the occurrence node information of each historical risk event, and then match the direct upstream and downstream nodes that will be affected by the abnormality after the node occurs through the supply and demand cooperation records of each node in the industrial chain. At the same time, combine the abnormality type and data change characteristics of the historical risk events to confirm the transmission direction of the abnormal impact. Pair the source node of the abnormality with the target node affected by it. Each pair of associations containing the source node identifier, the target node identifier and the transmission direction of the abnormal impact is the original flow relationship pair. The set formed by integrating all such association combinations is the original flow relationship pair of the industrial chain.

[0031] Extract the historical risk event occurrence time intervals for the source and target nodes corresponding to each pair of associations in the original flow relationship pairs. Compare the start time of the abnormal event at the source node with the start time of the abnormal event at the target node. If the start time of the abnormal event at the source node is earlier than the start time of the abnormal event at the target node, and there is a continuous time overlap between the occurrence time intervals of the two events, it is determined that the flow direction of the original flow relationship pair conforms to the actual causal transmission logic. If the start time of the abnormal event at the source node is later than the start time of the abnormal event at the target node, or there is no continuous time overlap between the occurrence time intervals of the two events, it is determined that the flow direction of the original flow relationship pair does not conform to the actual causal transmission logic and is directly eliminated. All original flow relationship pairs that conform to the causal transmission logic are retained. At the same time, a time sequence verification pass mark is added to each retained association combination. The set formed by integrating all association combinations with this mark is the refined causal pointing pair of the industry chain.

[0032] The statistical refinement of causal relationships involves analyzing the frequency of each associated combination in historical risk events within the supply chain. Simultaneously, it analyzes the correlation between the source node anomalies and target node anomalies corresponding to this combination. The correlation analysis method involves counting the number of times the target node experiences a corresponding propagation anomaly when the source node experiences this type of anomaly, and then dividing this number by the total number of times the source node experiences this type of anomaly. The resulting value is the correlation degree. A frequency threshold of 1.5 times the average frequency of this type of anomaly propagation relationship within the statistical period of historical risk events in the supply chain is set, and a correlation degree threshold of 0.7 is set. If the frequency of a certain set of refined causal pairs is higher than the frequency threshold and the correlation is higher than the correlation threshold, the set of pairs is determined to be a reliable causal relationship. If neither of the above two conditions is met, the pair is directly removed. All reliable causal relationships are organized in an orderly manner, and a reliability level label is added to each relationship. The reliability level is divided into three levels according to the frequency and correlation values. All reliable causal relationships with reliability level labels are arranged in descending order of reliability level to form an ordered list. This list is the reliable causal relationship list of the industry chain.

[0033] Extract all initial source nodes from the confidence causal pointing list that are not marked as being influenced by other nodes. These nodes are designated as the first-layer nodes in the industry chain. Then, based on the first-layer nodes, extract all target nodes directly influenced by the first-layer nodes from the confidence causal pointing list. These nodes are designated as the second-layer nodes in the industry chain. Next, based on the second-layer nodes, extract all target nodes directly influenced by the second-layer nodes from the confidence causal pointing list. These nodes are designated as the third-layer nodes in the industry chain. This layer-by-layer derivation method is used to sequentially define the nodes for each subsequent layer of the industry chain until all nodes in the confidence causal pointing list have been hierarchically defined. Simultaneously, a corresponding layer identifier is added to each node. The layer identifier includes the layer number and the node's functional attribute within that layer. All nodes that have been hierarchically defined and have been labeled are arranged from low to high layer number. Nodes within the same layer are arranged from high to low importance of their functional attributes. The resulting ordered set of nodes is the hierarchical node sequence of the industry chain.

[0034] Based on a hierarchical node sequence, nodes at each level are arranged sequentially in the visualized topological space in ascending order of hierarchical number along different horizontal dimensions. Nodes at the same level are arranged in descending order of functional attribute importance at different vertical positions within the same horizontal dimension. Each set of credible causal relationships in the confidence causal pointing list is then extracted. Based on the hierarchical identifiers of the source and target nodes in the relationships, directed lines with arrows are drawn for the corresponding source and target nodes in the topological space, with the arrow direction consistent with the transmission direction of the abnormal influence. A corresponding confidence level identifier is added to each directed line, consistent with the confidence level of the corresponding relationship in the confidence causal pointing list. The identifiers, functional attributes, hierarchical information of all nodes in the topological space, as well as the transmission direction and confidence level information of all directed lines, are then visualized and labeled. This visualized topological structure, with completed node arrangement, directed line drawing, and information labeling, is then solidified. This solidified visualized topological structure is the causal pointing map of the industry chain.

[0035] The beneficial effects include: extracting abnormal node propagation correlation information, covering potential transmission relationships, laying a solid data foundation for causal point verification, eliminating invalid correlation combinations, making causal points more realistic and improving their authenticity and effectiveness, determining credible causal points through double screening, providing precise priority basis for hierarchical node arrangement through hierarchical sorting, clarifying and orderly arranging upstream and downstream nodes, conforming to the industry operation architecture, providing clear node basis for topology reconstruction, intuitively presenting node dependencies and causal transmission relationships, and providing a reference for risk analysis through credibility levels, thereby improving the accuracy and efficiency of risk analysis.

[0036] S3. Based on the causal pointing map, perform upstream attribution component evaluation on the abnormal fluctuation points to obtain the causal effect value of the abnormal fluctuation points. In this embodiment of the invention, the step of evaluating the upstream attribution components of the abnormal fluctuation points based on the causal pointing map to obtain the causal effect value of the abnormal fluctuation points includes: Feature mining is performed on the fluctuation attributes of the abnormal fluctuation points to obtain the fluctuation amplitude characteristics and fluctuation time markers of the abnormal fluctuation points. Based on the topological connection relationship of the causal pointing graph, the upstream transmission path of the abnormal fluctuation point is traced to obtain the upstream node group of the abnormal fluctuation point. Historical load assessment is performed on the upstream node group to obtain the conduction strength parameters of the upstream node group; The path attenuation factor of the upstream transmission path is obtained by analyzing the upstream transmission path between the upstream node group and the abnormal fluctuation point. The causal effect value of the abnormal fluctuation point is obtained by attributing and quantifying the fluctuation amplitude characteristics, the transmission strength parameter, and the path attenuation factor.

[0037] The specific formula for calculating the causal effect value is as follows: ; In the formula, where, The causal effect value is... The fluctuation amplitude characteristic, The number of upstream nodes in the upstream node group. For the first The transmission strength parameters from each upstream node to the abnormal fluctuation point. For the causal pointing graph, the first... The upstream transmission path between each upstream node and the abnormal fluctuation point. For the nodes in the upstream transmission path Path decay factor.

[0038] Fluctuation attributes are various data characteristics exhibited when the operational data of the corresponding industrial chain nodes fluctuates at abnormal fluctuation points. These characteristics include the numerical range of data changes, the rate of change, and the duration of the fluctuation. When conducting feature mining on the fluctuation attributes of abnormal fluctuation points, all operational data fluctuation records corresponding to the abnormal fluctuation points are first extracted. These records include node identifiers, data collection timestamps, and specific values ​​of operational data before and after the fluctuation. Then, the difference between the specific values ​​of operational data before and after the fluctuation is calculated, and the absolute value of the result is the numerical change of the fluctuation. At the same time, the time length from the start of the fluctuation to the peak of the fluctuation is recorded as the fluctuation duration. Combining the numerical change and the fluctuation duration forms a comprehensive feature that can characterize the degree of data fluctuation at the abnormal fluctuation point. This comprehensive feature is the fluctuation amplitude feature. Finally, the first timestamp of the start of the fluctuation in the operational data corresponding to the abnormal fluctuation point is extracted, and this timestamp is associated and bound with the node identifier. The resulting association is the fluctuation time stamp.

[0039] The topological connections in a causal pointing graph are the directional upstream and downstream relationships between nodes in the industry chain. These relationships include direct and indirect connections between nodes, as well as the credibility level of each connection. When tracing the upstream transmission path of anomaly points based on this topological connection, the target node corresponding to the anomaly point in the causal pointing graph is first located. Then, using this target node as the endpoint, the first-level upstream nodes that can directly point to the target node are searched sequentially in the reverse direction of the directed lines in the graph. Based on the first-level upstream nodes, the search continues in the reverse direction of the directed lines to find second-level upstream nodes that can directly point to the first-level upstream nodes, and so on, until the initial source node in the graph with no upstream nodes is reached. During this process, connections with the lowest credibility level and no actual risk transmission record are removed. All first-level upstream nodes, second-level upstream nodes, and so on up to the initial source node found during the path tracing process are integrated to form a node set, which is the upstream node group of the anomaly point.

[0040] Historical load assessment analyzes the impact of each node in an upstream node cluster on the transmission of risks to downstream nodes throughout history. When conducting this assessment on an upstream node cluster, a separate historical risk transmission record is extracted for each node. This record includes the number of times the node transmitted risk to downstream nodes after each anomaly, the degree of fluctuation in the downstream node caused by each transmission, and the duration of the anomaly in the downstream node due to the transmission. The degree of fluctuation in the downstream node caused by a single transmission is then quantified, using the fluctuation range of the node's historical normal operating data as a benchmark. If the fluctuation of the downstream node exceeds the benchmark by more than double, it is recorded as the basic impact value. For each additional double exceeding the benchmark, an equal value is added to the basic impact value. Simultaneously, a weighted average is calculated based on the number of transmissions and the duration of the downstream node anomaly; the more transmissions and the longer the duration, the higher the weighting coefficient. The weighting coefficient is set based on the node's functional importance in the industry chain; the higher the functional importance, the larger the weighting coefficient. The comprehensive value obtained after completing the above quantification and weighting calculations for each upstream node is the transmission strength parameter for that node. Each node in the upstream node cluster corresponds to a unique transmission strength parameter.

[0041] The upstream transmission path is a complete directed connection path from a single node in the upstream node group to the target node corresponding to the abnormal fluctuation point. Each upstream node corresponds to a unique upstream transmission path, which contains several intermediate transmission nodes. When performing attenuation analysis on the upstream transmission path, the risk transmission filtering capability of each intermediate transmission node in each upstream transmission path is first analyzed. This capability is judged based on the operational buffer capability of the intermediate transmission node. The operational buffer capability includes the node's inventory reserves, capacity adjustment capability, and supply chain alternative resource reserves. The higher the inventory reserves, the stronger the capacity adjustment capability, and the more sufficient the alternative resource reserves, the stronger the risk transmission filtering capability of the node, and the greater the attenuation effect on risk transmission. Then, an attenuation value is set for each intermediate transmission node. The setting of the attenuation value is positively correlated with the risk transmission filtering capability. The stronger the filtering capability, the larger the attenuation value. Subsequently, the attenuation values ​​of all intermediate transmission nodes in a single upstream transmission path are accumulated and calculated. The result is the overall attenuation degree of the path. Then, the ratio of the overall attenuation degree of a single path to the maximum overall attenuation degree of all transmission paths in the industry chain is used as the benchmark. The resulting value is the path attenuation factor corresponding to the path.

[0042] When performing attribution quantification on fluctuation amplitude characteristics, transmission strength parameters, and path attenuation factors, firstly, for each node in the upstream node group, the value of its corresponding transmission strength parameter is multiplied by the value of the path attenuation factor of the upstream transmission path from that node to the target node to obtain the attribution quantification value of a single upstream node to the target node. Then, the attribution quantification values ​​of all nodes in the upstream node group are summed to obtain the comprehensive attribution quantification value of the upstream node group to the target node. Finally, the comprehensive characteristic value corresponding to the fluctuation amplitude characteristics is multiplied by the comprehensive attribution quantification value to obtain the final comprehensive value, which is the causal effect value of the abnormal fluctuation point. This value can comprehensively characterize the overall causal influence of the upstream node group on the abnormal fluctuation point through each transmission path.

[0043] In the formula for calculating the causal effect value, The causal effect value of the abnormal fluctuation point is to be determined, which is used to quantify the overall causal influence of the upstream node group on the abnormal fluctuation point. It is the core result of the upstream attribution component assessment and provides a quantitative basis for subsequent risk transmission and source tracing. The fluctuation amplitude characteristics of abnormal fluctuation points are obtained by feature mining of the fluctuation attributes of abnormal fluctuation points. They reflect the actual fluctuation degree of the points themselves and are the core data for formula calculation. The number of upstream nodes in the upstream node group is obtained by tracing the upstream transmission path based on the causal pointing graph topological connection relationship, and is used to determine the range of the formula summation operation. For the first The transmission strength parameter from each upstream node to the abnormal fluctuation point is obtained by evaluating the historical load of the upstream node group, which accurately characterizes the risk transmission influence of a single upstream node on the abnormal fluctuation point. For the causal pointing graph, the first The upstream transmission path from each upstream node to the abnormal fluctuation point is obtained by tracing the upstream transmission path, thus clarifying the selection range of the path attenuation factor; For the nodes in this upstream propagation path The path attenuation factor is obtained by performing attenuation analysis on the upstream transmission path. It reflects the actual attenuation of risk in the transmission path and is used to correct the transmission intensity of a single node. It refers to all nodes within the path. Multiply the path attenuation factors to obtain the overall attenuation coefficient of a single conduction path; arrive It is obtained by summing the products of the transmission intensity parameters of each upstream node and the overall attenuation coefficient of the corresponding path in the order of nodes, and then obtaining the comprehensive attribution quantification value of the upstream node group.

[0044] The overall formula first corrects the path attenuation of the transmission intensity of a single node and sums the results. Then, it multiplies the result with the fluctuation amplitude characteristics to obtain the causal effect value. By comprehensively combining the degree of abnormal fluctuation itself, the transmission intensity of upstream nodes and the path attenuation effect, the overall causal influence of the upstream node group on the abnormal fluctuation point is accurately quantified.

[0045] The beneficial effects include: accurately extracting the core fluctuation characteristics and occurrence time of abnormal fluctuation points, providing accurate basic data for upstream attribution component assessment, improving the accuracy and uniqueness of initial data for attribution assessment; comprehensively tracing upstream nodes based on the directed correlation of causal pointing graphs, while eliminating invalid connections, improving the completeness and effectiveness of the upstream node group; combining historical risk transmission records of nodes with the functional importance of the industrial chain for comprehensive quantitative calculation, allowing the transmission intensity parameter to accurately characterize the risk transmission impact intensity of a single upstream node, improving the scientificity and accuracy of node impact intensity judgment; analyzing the degree of risk transmission attenuation by combining the actual operational buffering capacity of intermediate transmission nodes, allowing the path attenuation factor to accurately reflect the actual attenuation of risk transmission, improving the fit between path impact judgment and actual operation of the industrial chain; and obtaining causal effect values ​​through layer-by-layer calculation of multi-dimensional data, accurately characterizing the overall causal impact of the upstream node group on abnormal fluctuation points, providing accurate quantitative basis for subsequent risk transmission tracing, and improving the accuracy and reliability of risk tracing.

[0046] S4. Based on the causal effect value and the causal pointing map, the source of risk transmission is traced at the abnormal fluctuation points to obtain the risk source points and risk propagation paths of the industrial chain; In this embodiment of the invention, the step of tracing the source of risk transmission at the abnormal fluctuation points based on the causal effect value and the causal pointing map to obtain the risk source points and risk propagation paths of the industrial chain includes: Based on the causal pointing graph, the upstream adjacency of the abnormal fluctuation point is traced to obtain the upstream adjacency node of the abnormal fluctuation point.

[0047] Based on the causal effect value, the influence contribution of the upstream adjacent nodes is evaluated to obtain the upstream attribution node record of the abnormal fluctuation point. By performing hierarchical backtracking on the nodes in the upstream attribution node record along the causal pointing graph, the complete upstream tracing chain of the abnormal fluctuation point is obtained; By marking the source points of the complete upstream traceability chain, the risk source locations of the industrial chain can be obtained; Based on the node pointing relationship in the complete upstream traceability chain, the path from the risk source point to the abnormal fluctuation point is plotted to obtain the risk propagation path of the industrial chain.

[0048] Extract the node identifiers and hierarchical information of the abnormal fluctuation points, retrieve all nodes directly connected to the node and whose arrows point to the node in the causal pointing graph, integrate and record the identifiers, hierarchical attributes, functional attributes, and causal pointing confidence levels of all retrieved nodes with abnormal fluctuation points, and the resulting set of node information is the upstream adjacent node of the abnormal fluctuation point.

[0049] The individual influence components of each upstream neighboring node in the causal effect value are decomposed. The ratio of the individual influence component of each upstream neighboring node to the sum of the individual influence components of all upstream neighboring nodes is calculated to obtain the influence contribution ratio of each upstream neighboring node to the abnormal fluctuation point. The average influence contribution ratio of all upstream neighboring nodes is taken as the judgment threshold. Upstream neighboring nodes with influence contribution ratios higher than the judgment threshold are screened out. At the same time, the identifier, individual influence component, influence contribution ratio, and transmission relationship information of each screened node are recorded. The structured information record formed is the upstream attribution node record of the abnormal fluctuation point.

[0050] Extract the identifier and hierarchical information of each node in the upstream attribution node record. Using each node as the current source node, search for its directly connected upstream neighbor nodes in the causal pointing graph, with arrows pointing to that node. Repeat this search operation until the retrieved node is the initial source node with no upstream neighbor nodes in the causal pointing graph. Organize the node transmission sequence from all initial source nodes to the abnormal fluctuation point and arrange the nodes of each transmission link in order from upstream to downstream. At the same time, record the causal pointing relationship and confidence level between each node. Integrate all the sorted node transmission sequences to form a sequence set containing complete node transmission sequence and relationship information, which is the complete upstream tracing chain of the abnormal fluctuation point.

[0051] By sorting out the transmission sequence of all nodes in the complete upstream traceability chain, the initial source node in each sequence that has no transmission influence from upstream nodes is identified. All identified initial source nodes are uniformly marked with source point labels, the label content of which includes the source point number and the corresponding abnormal transmission initiation attribute. At the same time, the identifier, hierarchical attribute, functional attribute, and transmission initiation information of all marked initial source nodes in the traceability chain are integrated to form a set of marked node information, which is the risk source location of the industrial chain.

[0052] Extract the node transmission sequence from each risk source point to the abnormal fluctuation point in the complete upstream traceability chain and the directed relationship between each node. In the topological space of the causal pointing graph, with the risk source point as the starting point and the abnormal fluctuation point as the ending point, draw trajectory lines with transmission direction for adjacent nodes in sequence according to the node transmission sequence. The style of the trajectory lines matches the credibility level of the corresponding relationship in the causal pointing graph. Mark the transmission relationship information and transmission sequence number between each node on the trajectory lines. Integrate all the drawn trajectories and record the node composition, transmission sequence, pointing relationship and credibility level of each trajectory. The resulting path set containing topological trajectories and detailed information is the risk propagation path of the industrial chain.

[0053] The beneficial effects include: accurately retrieving the direct upstream causal transmission nodes of abnormal fluctuation points, ensuring the accuracy and completeness of node information, providing a reliable starting point for subsequent assessments, quantitatively assessing the impact contribution of upstream adjacent nodes, screening core nodes with quantitative thresholds, improving the targeting and scientific nature of attribution node screening, focusing on the core of source tracing, continuously tracing back to the initial source node according to the topological relationship hierarchy, completely restoring the entire chain transmission node sequence of abnormal fluctuations, improving the coherence and comprehensiveness of the source tracing process, accurately identifying the risk initiation source node and completing unified marking and information integration, clarifying the core attributes of the risk source, achieving accurate positioning of the risk source in the industrial chain, drawing a visualized risk transmission trajectory and completing multi-dimensional information annotation, making the risk propagation path more intuitive, while ensuring the detail of path information records, and clearly restoring the entire risk transmission process.

[0054] S5. Based on the risk source location and the risk propagation path, perform counterfactual blocking inference in the causal pointing map to obtain the blocking data of the risk propagation path; In this embodiment of the invention, the step of performing counterfactual blocking inference on the causal pointing map based on the risk source location and the risk propagation path to obtain blocking data for the risk propagation path includes: The path nodes along the risk propagation path are selected based on their betweenness numbers to obtain a high betweenness number node group for the risk propagation path. The blocking effect of candidate nodes in the high betweenness number node group is simulated to obtain the simulated blocking data of the candidate node group; Based on the simulated blocking data, the blocking coverage rate and blocking implementation cost of the candidate nodes are comprehensively compared to obtain the optimal blocking node for the risk propagation path; The blocking effect of the optimal blocking node is verified to obtain blocking data that blocks the risk propagation path, including: The influence domain of the optimal blocking node is delineated to obtain the blocking influence area of ​​the optimal blocking node; Based on the area affected by the blockade, the dependent edges of the optimal blockade node are cut off to obtain the post-blockade dependency relationship of the industrial chain; Based on the post-blocking dependency relationship, a propagation blocking assessment is performed on the area affected by the blocking to obtain the blocking data of the risk propagation path.

[0055] Complete information on all path nodes in the risk propagation path is extracted, including node identifiers, hierarchical attributes, functional attributes, and all connections between nodes and other nodes in the causal pointing graph. Node betweenness is an indicator that characterizes the importance of a node as a transmission intermediary in the risk propagation path. The number of times each path node appears as an intermediate node in the transmission sequence of all nodes in the risk propagation path is counted. At the same time, the number of upstream transmission connections received by the node and the number of transmission connections sent downstream in the causal pointing graph are counted. The number of intermediate node occurrences is summed with the total number of transmission connections of the node. The result is then compared with the average of the same type of result for all nodes in the risk propagation path. Nodes with results higher than the average are classified as high betweenness nodes. All classified high betweenness nodes are integrated into a set, which is the high betweenness node group of the risk propagation path. The group contains the identifier, hierarchical attributes, functional attributes, and betweenness-related statistical data of each high betweenness node.

[0056] In the topological simulation environment corresponding to the causal pointing graph, the node connection relationship of the current industrial chain, the risk transmission characteristics of the risk source point, and the transmission trajectory of the risk propagation path are replicated. The candidate nodes are all nodes in the high betweenness node group. Then, each candidate node in the high betweenness node group is set to the blocking state in turn. The blocking state means cutting off all upstream and downstream transmission connections of the node. The transmission process of risk from the risk source point is simulated in the simulation environment. The number of nodes that the risk cannot be transmitted to when each candidate node is in the blocking state, the reduction ratio of the risk transmission range, and the time delay length of the risk transmission to the abnormal fluctuation point are recorded. At the same time, the number of nodes affected by the blocking of the candidate node in the industrial chain in other normal transmission paths is recorded. All the above simulation record data corresponding to each candidate node are integrated to form a structured data set, which is the simulation blocking data of the candidate node group. Each candidate node in the dataset corresponds to a unique set of simulation blocking records.

[0057] Blocking coverage rate is the proportion of nodes to which the risk cannot be transmitted after a candidate node is blocked, relative to the total number of nodes in the risk propagation path. Blocking implementation cost is the number of nodes that will affect other normal transmission paths in the industry chain after a candidate node is blocked. First, the blocking coverage rate of each candidate node is calculated based on simulation blocking data by dividing the number of nodes to which the risk cannot be transmitted after blocking the node by the total number of nodes in the risk propagation path. Then, the specific number of nodes corresponding to the blocking implementation cost of each candidate node is extracted based on the simulation blocking data. The weight of blocking coverage rate is set at 70%, and the weight of blocking implementation cost is set at 30%. The blocking coverage rate value of each candidate node is multiplied by the corresponding weight, and then the result of multiplying the blocking implementation cost value by the corresponding weight is subtracted from the result to obtain the comprehensive score of each candidate node. All candidate nodes are sorted from high to low according to the comprehensive score, and the candidate node with the highest comprehensive score is selected as the optimal blocking node in the risk propagation path. This node contains complete identification, hierarchical attributes, functional attributes, and calculation data related to the comprehensive score.

[0058] Extract all upstream and downstream connections of the optimal blocking node in the causal pointing graph, including the identifiers of directly connected upstream and downstream nodes and the identifiers of indirectly connected upstream and downstream nodes. Then, combine the functional attributes of the node in the industry chain to determine the range of industry chain links corresponding to the transmission function undertaken by the node. Taking the optimal blocking node as the core, integrate its directly connected upstream and downstream nodes, indirectly connected upstream and downstream nodes, and the range of corresponding industry chain transmission links to delineate a continuous transmission region. This region contains all nodes and industry chain transmission links affected by the blocking operation of the optimal blocking node. This delineated continuous transmission region is the blocking influence region of the optimal blocking node. The region records the identifiers, hierarchical attributes, and corresponding industry chain transmission link information of all affected nodes.

[0059] Identify all causal-directed dependency edges between the optimal blocking node and other nodes within the blocking influence area. Dependency edges are directed transmission lines with arrows between nodes in the causal-directed graph. Then, according to the requirements of the blocking operation, cut off all upstream input dependency edges and downstream output dependency edges of the optimal blocking node within the blocking influence area. Retain normal dependency edges outside the blocking influence area that are not directly related to the optimal blocking node. At the same time, record the remaining causal-directed connections between nodes within the blocking influence area after cutting off dependency edges, as well as the normal connections between nodes within the blocking influence area and nodes outside the area. Reintegrate the causal-directed connections between all nodes in the industrial chain after cutting off dependency edges. The resulting structured relationship set is the post-blocking dependency relationship of the industrial chain. This set contains the identifiers of all nodes, as well as the remaining connections and transmission direction information between nodes.

[0060] In the topological environment of the causal pointing graph, the transmission process of risk from the risk source point is re-simulated based on the dependency relationship after blocking. The transmission termination position of risk in the blocking influence area, the probability that the risk cannot break through the blocking influence area, and the actual transmission interruption length of the risk propagation path are recorded during the simulation. At the same time, the reduction ratio of risk fluctuation amplitude at the abnormal fluctuation point after the blocking operation and the probability of risk transmission to that point are statistically analyzed. The statistical data obtained from all the above assessments and simulation records are integrated to form a structured data set, which is the blocking data of the risk propagation path. This data includes the risk transmission termination characteristics, fluctuation amplitude change characteristics, and all information related to the transmission interruption after the blocking operation.

[0061] The beneficial effects are as follows: it accurately selects high-median-number nodes, laying a precise data foundation for the screening of blocking nodes; it simulates and records the blocking characteristic data of candidate nodes, providing detailed support for the selection of the optimal blocking node; it quantitatively and comprehensively scores to select the optimal blocking node, balancing effectiveness and cost to ensure the blocking effect; it accurately delineates the blocking impact area, clarifies the boundary of action, avoids misjudging irrelevant transmission links; it accurately reconstructs the node dependency relationship after blocking, providing a precise basis for the propagation blocking assessment; it evaluates the propagation blocking effect in all dimensions, and provides reliable data support for the compilation of early warning information.

[0062] S6. Compile the blocked data to obtain the industrial chain risk warning information; In this embodiment of the invention, the step of compiling the blocked data to obtain the supply chain risk warning information includes: The blocked data is semantically parsed to obtain the structured compiled fields of the blocked data; Based on the structured compiled fields, the risk propagation path is condensed into text to obtain a textual risk description of the industrial chain; The structured compiled fields and the textual risk description are encapsulated in a message to obtain the industry chain risk warning information.

[0063] All data elements in the blocking data are extracted, with each data element corresponding to a specific information dimension related to the blocking. A fixed field classification framework is established for the information display and usage needs of supply chain risk warning. This framework includes four core field categories: node information, regional information, relationship change, and blocking effect. Each category is further divided into sub-fields. The node information sub-field includes the blocking node number, node level, and node functional attributes. The regional information sub-field includes the list of nodes covered by the blocking impact area and the hierarchical distribution of nodes within the region. The relationship change sub-field includes the number of cut-off dependent edges, the upstream and downstream relationships retained after blocking, and the node pairings with changed dependent relationships. The blocking effect sub-field includes the risk propagation blocking rate, the degree of risk transmission attenuation, and the recovery trend of abnormal fluctuation points after blocking. Each extracted data element is matched to the corresponding sub-field according to its information attributes. The data source within each sub-field is standardized and organized to unify the data expression format and recording dimensions. All fields that have completed data matching and standardization are sorted and integrated by major category. The resulting information set with a fixed field structure and standardized data content is the structured compiled field.

[0064] Core data related to the risk propagation path is extracted from structured compiled fields, including risk source location identifiers, a list of nodes along the risk propagation path, the location of the optimal blocking node in the path, the transmission links severed after blocking, and the actual coverage of risk propagation after blocking. A textualized expression framework is established according to the information reading logic of supply chain risk warning. This framework arranges the expression content in the order of risk source, propagation process, blocking intervention, and blocking effect. The extracted core data is then converted into text according to the order of the expression framework, and the information corresponding to each data point is described in concrete text. This clarifies the level and function of the risk source location, the transmission sequence of nodes in the propagation path, the transmission link of the optimal blocking node in the path, the specific two nodes involved in the severed transmission link, the range of nodes where the risk cannot be propagated after blocking, and the range of nodes where slight transmission still exists. The converted textual content is logically sorted and sentences integrated, eliminating redundant expressions and ensuring the coherence and completeness of the information. The resulting textual content, presented in standardized language and according to fixed logic, describing the risk propagation path and all related information, constitutes the textualized risk description of the supply chain.

[0065] A standard message format for supply chain risk warning information is established. This format includes three parts: header, body, and footer. The header area is used to enter the generation time, warning number, and warning level of the warning information. The warning level is determined based on the risk propagation blocking rate in the blocking effect category of the structured compiled fields. A risk propagation blocking rate of less than 50% is a Level 1 warning, 50% to 80% is a Level 2 warning, and higher than 80% is a Level 3 warning. The body area is divided into a structured data area and a text description area. The structured data area fully incorporates all the content of the structured compiled fields, and the text description area fully incorporates all the content of the textual risk description. The footer area is used to enter the data source, compiler identification, and verification results. Following the established standard message format, the generation time of the warning information, the unique warning number, and the determined warning level are entered sequentially in the corresponding areas. Structured compilation fields and textual risk descriptions are embedded in the corresponding partitions of the message body. The source of the blocking data, the identification of the personnel who performed this information compilation, and the results of the integrity verification of the compiled content are entered in the tail area. All information in the message is format verified to ensure that there are no missing information in each area, the format is uniform, and the content is consistent. The standard message that has completed information entry and format verification is solidified. The resulting complete warning information carrier containing standardized structured data, standardized text descriptions, and uniform format is the industrial chain risk warning information.

[0066] The beneficial effects include: standardizing and blocking data, improving its readability, enabling accurate information classification, avoiding early warning bias, making abstract risk path data more concrete, improving readability, reducing the cost of understanding risk information, standardizing early warning information carriers, improving their standardization and completeness, classifying early warning levels, improving the efficiency of early warning transmission and execution, and ensuring that early warnings are implemented.

[0067] like Figure 2 The diagram shown is a functional block diagram of an industrial chain risk early warning system based on multi-source risk signals provided in an embodiment of the present invention.

[0068] The supply chain risk early warning system 100 based on multi-source risk signals described in this invention can be installed in an electronic device. Depending on the functions implemented, the supply chain risk early warning system 100 may include a risk anomaly detection module 101, a causal topology construction module 102, a causal effect assessment module 103, a risk transmission and tracing module 104, a risk blocking deduction module 105, and an early warning information compilation module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0069] In this embodiment, the functions of each module / unit are as follows: The risk anomaly detection module 101 is used to detect anomalies in the historical risk time series data and real-time risk time series data of the industrial chain, and to obtain the historical risk events and real-time abnormal fluctuation points of the industrial chain. The causal topology construction module 102 is used to construct a directed topology of the causal orientation of the industrial chain based on the upstream and downstream flow relationship in the historical risk event, so as to obtain the causal orientation map of the industrial chain. The causal effect assessment module 103 is used to assess the upstream attribution component of the abnormal fluctuation point based on the causal pointing map, and obtain the causal effect value of the abnormal fluctuation point. The risk transmission and tracing module 104 is used to trace the source of risk transmission of the abnormal fluctuation points based on the causal effect value and the causal pointing map, so as to obtain the risk source points and risk propagation paths of the industrial chain. The risk blocking simulation module 105 is used to perform counterfactual blocking simulation on the causal pointing map based on the risk source location and the risk propagation path to obtain blocking data of the risk propagation path; The early warning information compilation module 106 is used to compile the blocking data to obtain the industrial chain risk early warning information.

[0070] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0071] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0072] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0073] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0074] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A supply chain risk early warning method based on multi-source risk signals, characterized in that, The method includes: S1. Detect anomalies in the historical and real-time risk time series data of the industrial chain to obtain the historical risk events and real-time abnormal fluctuation points of the industrial chain. S2. Based on the upstream and downstream flow relationships in the historical risk events, a directed topology is constructed for the causal orientation of the industrial chain to obtain the causal orientation map of the industrial chain. S3. Based on the causal pointing map, perform upstream attribution component evaluation on the abnormal fluctuation points to obtain the causal effect value of the abnormal fluctuation points. S4. Based on the causal effect value and the causal pointing map, the source of risk transmission is traced at the abnormal fluctuation points to obtain the risk source points and risk propagation paths of the industrial chain; S5. Based on the risk source location and the risk propagation path, perform counterfactual blocking inference in the causal pointing map to obtain the blocking data of the risk propagation path; S6. Compile the blocked data to obtain the industrial chain risk warning information.

2. The supply chain risk early warning method based on multi-source risk signals as described in claim 1, characterized in that, The process of detecting anomalies in historical and real-time risk time-series data of the industrial chain to obtain historical risk events and real-time abnormal fluctuation points of the industrial chain includes: The operational log data of the nodes in the industry chain are continuously collected to obtain the risk time series data of the industry chain; The risk time series data is smoothed and denoised to obtain the purified risk data of the industrial chain; By detecting change points in the purified risk data, the discrete risk fluctuation points of the industrial chain can be obtained. According to the preset time-series classification rules, the initial abnormal fluctuation points are classified to obtain the historical fluctuation points and real-time fluctuation points of the industrial chain. By extracting events from the aforementioned historical fluctuation points, the historical risk events of the industrial chain can be obtained. The real-time fluctuation points are identified as anomalies to obtain the real-time abnormal fluctuation points of the industrial chain.

3. The supply chain risk early warning method based on multi-source risk signals as described in claim 1, characterized in that, The process of constructing a directed topology of the causal relationships of the industrial chain based on the upstream and downstream flow relationships in the historical risk events, to obtain a causal orientation graph of the industrial chain, includes: The flow of the historical risk events is identified to obtain the original flow relationship pairs of the industrial chain; Based on the sequence of occurrence of the historical risk events, the original flow relationship pairs are verified to obtain the refined causal pair of the industrial chain; The refined causal pairs are subjected to a reliable selection process to obtain a list of reliable causal pairs for the industrial chain. Based on the aforementioned confidence causal pointing list, the nodes in the industrial chain are hierarchically arranged to obtain the hierarchical node sequence of the industrial chain; Based on the hierarchical node sequence and the confidence causal pointing list, the upstream and downstream dependencies between the nodes are topologically reconstructed to obtain the causal pointing graph of the industrial chain.

4. The supply chain risk early warning method based on multi-source risk signals as described in claim 1, characterized in that, The process of evaluating the upstream attribution components of the anomalous fluctuation points based on the causal pointing map to obtain the causal effect value of the anomalous fluctuation points includes: Feature mining is performed on the fluctuation attributes of the abnormal fluctuation points to obtain the fluctuation amplitude characteristics and fluctuation time markers of the abnormal fluctuation points. Based on the topological connection relationship of the causal pointing graph, the upstream transmission path of the abnormal fluctuation point is traced to obtain the upstream node group of the abnormal fluctuation point. Historical load assessment is performed on the upstream node group to obtain the conduction strength parameters of the upstream node group; The path attenuation factor of the upstream transmission path is obtained by analyzing the upstream transmission path between the upstream node group and the abnormal fluctuation point. The causal effect value of the abnormal fluctuation point is obtained by attributing and quantifying the fluctuation amplitude characteristics, the transmission strength parameter, and the path attenuation factor.

5. The supply chain risk early warning method based on multi-source risk signals as described in claim 4, characterized in that, The specific formula for calculating the causal effect value is as follows: ； In the formula, where, The causal effect value is... The fluctuation amplitude characteristic, The number of upstream nodes in the upstream node group. For the first The transmission strength parameters from each upstream node to the abnormal fluctuation point. For the causal pointing graph, the first... The upstream transmission path between each upstream node and the abnormal fluctuation point. For the nodes in the upstream transmission path Path decay factor.

6. The supply chain risk early warning method based on multi-source risk signals as described in claim 1, characterized in that, The process of tracing the source of risk transmission at the abnormal fluctuation points based on the causal effect value and the causal pointing map to obtain the risk source points and risk propagation paths of the industrial chain includes: Based on the causal pointing graph, the upstream adjacency of the abnormal fluctuation point is traced to obtain the upstream adjacency node of the abnormal fluctuation point; Based on the causal effect value, the influence contribution of the upstream adjacent nodes is evaluated to obtain the upstream attribution node record of the abnormal fluctuation point. By performing hierarchical backtracking on the nodes in the upstream attribution node record along the causal pointing graph, the complete upstream tracing chain of the abnormal fluctuation point is obtained; By marking the source points of the complete upstream traceability chain, the risk source locations of the industrial chain can be obtained; Based on the node pointing relationship in the complete upstream traceability chain, the path from the risk source point to the abnormal fluctuation point is plotted to obtain the risk propagation path of the industrial chain.

7. The supply chain risk early warning method based on multi-source risk signals as described in claim 1, characterized in that, Based on the risk source locations and the risk propagation path, counterfactual blocking inference is performed on the causal pointing map to obtain blocking data for the risk propagation path, including: The path nodes along the risk propagation path are selected by betweenness selection to obtain a high betweenness node group of the risk propagation path; The blocking effect of candidate nodes in the high betweenness number node group is simulated to obtain the simulated blocking data of the candidate node group; Based on the simulated blocking data, the blocking coverage rate and blocking implementation cost of the candidate nodes are comprehensively compared to obtain the optimal blocking node for the risk propagation path; The blocking effect of the optimal blocking node is verified to obtain blocking data that blocks the risk propagation path.

8. The supply chain risk early warning method based on multi-source risk signals as described in claim 7, characterized in that, The step of verifying the blocking effect of the optimal blocking node to obtain blocking data that blocks the risk propagation path includes: The influence domain of the optimal blocking node is delineated to obtain the blocking influence area of ​​the optimal blocking node; Based on the area affected by the blockade, the dependent edges of the optimal blockade node are cut off to obtain the post-blockade dependency relationship of the industrial chain; Based on the post-blocking dependency relationship, a propagation blocking assessment is performed on the area affected by the blocking to obtain the blocking data of the risk propagation path.

9. The supply chain risk early warning method based on multi-source risk signals as described in claim 1, characterized in that, The process of compiling the blocked data to obtain the supply chain risk warning information includes: The blocked data is semantically parsed to obtain the structured compiled fields of the blocked data; Based on the structured compiled fields, the risk propagation path is condensed into text to obtain a textual risk description of the industrial chain; The structured compiled fields and the textual risk description are encapsulated in a message to obtain the industry chain risk warning information.

10. A supply chain risk early warning system based on multi-source risk signals, characterized in that, The system is used to implement the supply chain risk early warning method based on multi-source risk signals as described in claim 1, the system comprising: The risk anomaly detection module is used to detect anomalies in the historical and real-time risk time series data of the industrial chain, and to obtain the historical risk events and real-time abnormal fluctuation points of the industrial chain. The causal topology construction module is used to construct a directed topology of the causal orientation of the industrial chain based on the upstream and downstream flow relationships in the historical risk events, so as to obtain the causal orientation map of the industrial chain. The causal effect assessment module is used to assess the upstream attribution components of the abnormal fluctuation points based on the causal pointing map, and obtain the causal effect value of the abnormal fluctuation points. The risk transmission and tracing module is used to trace the source of risk transmission based on the causal effect value and the causal orientation map, so as to obtain the risk source location and risk propagation path of the industrial chain. The risk blocking simulation module is used to perform counterfactual blocking simulation on the causal pointing map based on the risk source location and the risk propagation path to obtain blocking data of the risk propagation path; The early warning information compilation module is used to compile the blocked data to obtain the industrial chain risk early warning information.

Images