Risk perception method and device for global supply chain enterprise, and medium

Abstract

The embodiment of the invention discloses a risk perception method and device for a global supply chain enterprise and a medium, and relates to the technical field of data analysis, and the method comprises the steps: receiving a risk perception request of a target enterprise for an enterprise perception object, the enterprise perception object comprises a target product and a target shipping port, the risk perception request comprises real-time perception and simulation interruption perception; based on the risk perception request, performing entity relationship extraction from a pre-constructed dynamic knowledge graph to form an evaluation sub-graph, and calculating a multi-dimensional quantitative toughness index through the evaluation sub-graph; and fusing the multi-dimensional quantitative toughness indexes to generate a comprehensive toughness index corresponding to the enterprise perception object, and carrying out risk perception on the enterprise perception object based on the comprehensive toughness index.

Description

Technical Field

[0001] This specification relates to the field of data analysis technology, and in particular to a risk perception method, device and medium for global supply chain enterprises. Background Technology

[0002] In the field of global supply chain management, existing risk perception methods mainly rely on two technological approaches: one is internal data monitoring based on enterprise resource planning (ERP) systems, and the other is trend analysis based on macroeconomic trade statistics. Specifically, existing technologies typically construct supply chain risk assessment models by collecting internal inventory and order delivery data from enterprises and combining them with quarterly trade statistics released by customs. These methods identify potential supply disruption risks through regression analysis of historical data.

[0003] A thorough analysis of existing technologies reveals that their reliance on static internal enterprise data and lagging macroeconomic statistics, coupled with the disconnect between commercial and logistical data, leads to blind spots in risk perception. Furthermore, their dependence on retrospective analysis of historical trade data or static network models fails to capture ongoing dynamic changes in logistics, resulting in significant lag in risk perception. Secondly, existing assessment methods are mostly limited to a single data dimension, focusing either solely on internal business process data or macroeconomic trade volume data, lacking an analytical perspective that integrates business logic with the physical logistics status. In addition, existing risk assessment models often only provide post-event statistical analysis and cannot proactively simulate and extrapolate potential disruptions.

[0004] In summary, existing risk perception technologies suffer from fragmented data sources and static, one-sided analytical methods, which prevent them from meeting the real-time, multi-dimensional, and extrapolable risk perception needs of global supply chain companies for specific supply chains. Summary of the Invention

[0005] This specification provides one or more embodiments of a risk perception method, device, and medium for global supply chain enterprises, which addresses the following technical problem: Existing risk perception technologies suffer from fragmented data sources and static and one-sided analysis methods, making it impossible to meet the real-time, multi-dimensional, and predictable risk perception needs of global supply chain enterprises for specific supply chains.

[0006] One or more embodiments of this specification employ the following technical solutions: This specification provides one or more embodiments of a risk perception method for global supply chain enterprises. The method includes: receiving a risk perception request from a target enterprise for a perceived object, wherein the perceived object includes a target product and a target shipping port, and the risk perception request includes real-time perception and simulated interruption perception; based on the risk perception request, extracting entity relationships from a pre-constructed dynamic knowledge graph to form an evaluation subgraph, and calculating multi-dimensional quantitative resilience indicators through the evaluation subgraph; fusing the multi-dimensional quantitative resilience indicators to generate a comprehensive resilience index corresponding to the perceived object, and performing risk perception on the perceived object based on the comprehensive resilience index.

[0007] This specification provides one or more embodiments of a risk perception device for global supply chain enterprises, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the above-described method.

[0008] This specification provides one or more embodiments of a non-volatile computer storage medium storing computer-executable instructions configured to perform the above-described method.

[0009] The at least one technical solution adopted in the embodiments of this specification can achieve the following beneficial effects: Through the embodiments of this specification, by constructing a dynamic knowledge graph that integrates global shipping dynamics and enterprise internal supply chain data, the problem of the separation of commercial flow and logistics data in traditional methods is fundamentally solved. It can systematically reveal the complete path and impact of a single port congestion or supplier disruption transmitted to the final product through a complex network; it introduces a multi-dimensional quantitative resilience index system, and performs collaborative calculations from four interrelated and complementary dimensions: logistics efficiency, network topology, redundancy substitutability, and abnormal response recovery. It not only reveals the current operating efficiency of the supply chain through route delay rates, but also utilizes node centrality. It deeply analyzes the inherent structural characteristics, redundancy capabilities to withstand single failures, and recovery potential after being impacted; it generates a comprehensive resilience index through weighted fusion and supports both real-time perception and simulated interruption modes. The comprehensive index condenses multi-dimensional information into an intuitive quantitative scale, overcoming the problem of difficulty in making overall decisions based on multiple indicators. Simulated interruption perception allows enterprises to conduct stress tests in the digital space before risk events occur, proactively acquiring the potential impact and transmission path after the interruption of critical product dependence paths or core ports, transforming risk management from passively responding to historical events to proactively preventing future risks, and meeting the real-time, multi-dimensional, and predictable risk perception needs of global supply chain enterprises for specific enterprise supply chains. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A flowchart illustrating a risk perception method for global supply chain enterprises, provided as an embodiment of this specification; Figure 2 This is a schematic diagram of the structure of a risk perception device for a global supply chain enterprise, provided as an embodiment of this specification. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0012] This specification provides a risk perception method for global supply chain enterprises. It should be noted that the execution entity in this specification embodiment can be a server or any device with data processing capabilities. Figure 1 A flowchart illustrating a risk perception method for global supply chain enterprises, provided as an embodiment of this specification, is shown below. Figure 1 As shown, the main steps include the following: Step S101: Receive the risk perception request from the target enterprise regarding the perceived object of the enterprise.

[0013] The enterprise's perception objects include target products and target shipping ports, and the risk perception request includes real-time perception and simulated interruption perception. It should be noted that, in this embodiment, the target enterprise in receiving the risk perception request from the target enterprise regarding the enterprise's perceived object refers to a manufacturing enterprise or a large retail enterprise with a global supply chain layout. The enterprise's perceived object is a key element of enterprise supply chain risk management; the enterprise's perceived object can be a specific product of the enterprise or a target shipping port.

[0014] If the target product is involved, the supply risk can be assessed in real time, or a simulated risk assessment of a potential disruption to the target product can be conducted. Specifically, a target product refers to a specific finished product or key raw material / component manufactured by the company. When the assessment object is a target product, the focus will be on the complete supply chain network of that product, from raw material supply to finished product delivery. Companies can request real-time assessment of the product's supply chain to understand its current operational status and risk level; they can also request simulated disruption assessment, which simulates the impact on the overall supply chain by virtually interrupting a certain link in the product's supply chain.

[0015] If the target shipping port is essential for the company's product logistics, the assessment can be either real-time perception of the logistical risks associated with the port's supply of the corresponding products, or a simulated risk perception of the company should the port be disrupted. Specifically, a target shipping port refers to a seaport that serves as a critical hub in the company's supply chain logistics network. These ports are typically core nodes for the import and export of bulk goods, and their operational status directly affects the smoothness of the supply chain. When the target shipping port is the object of perception, the company can request real-time perception of its operational status to assess its immediate impact on the logistics of related products; or it can request simulated disruption perception, by virtually closing or restricting the port's functions, to proactively assess the potential scope and extent of the disruption event's impact on the company's supply chain.

[0016] In one embodiment of this specification, the aforementioned risk perception request is received through a unified interface. The request must explicitly specify the type of the perception object (product / port), specific identifiers (such as product SKU code, port UN / LOCODE code), and perception mode (real-time / simulation). Based on the request type, subsequent differentiated data processing and analysis procedures are triggered, providing precise data support for the enterprise's supply chain risk management.

[0017] Step S102: Based on the risk perception request, entity relationships are extracted from the pre-built dynamic knowledge graph to form an evaluation subgraph, so as to calculate multi-dimensional quantitative resilience indicators through the evaluation subgraph.

[0018] Based on this risk perception request, before extracting entity relationships from a pre-constructed dynamic knowledge graph to form an evaluation subgraph, the method further includes: acquiring global ship AIS data streams, port basic data, ship static data, and the target company's internal supply chain data, including supplier master data, product bill of materials (BOM) data, and purchase order data; performing data cleaning on the global ship AIS data stream to remove latitude and longitude outliers and imput missing values, and then performing map matching algorithm processing on the cleaned AIS data stream to associate ship trajectories with predefined sea routes to determine ship continuity. Trajectory data; based on the port's basic data, the ship's static data, and the internal supply chain data, multiple entities are identified, including port entities, ship entities, enterprise entities, and product entities; through cleaned AIS data streams, the ship's continuous trajectory data, and the internal supply chain data, entity relationships among these multiple entities are defined, including berthing relationships, connection relationships, supply relationships, and production relationships; based on these multiple entities and their relationships, an initial knowledge graph is constructed by storing them in a graph database, and a dynamic knowledge graph is constructed by dynamically updating the ship's position attributes and port status attributes by consuming real-time AIS data streams.

[0019] Based on the cleaned AIS data stream, the berthing relationships between ship entities and port entities are defined, and ship berthing events are detected using a geofencing algorithm to generate berthing relationship attributes, including start time, end time, and berthing duration. Using the continuous ship trajectory data, the connection relationships between port entities are defined to calculate the route distance and historical freight frequency between ports, generating connection relationship attributes. Based on the purchase order data in the supply chain data, the port basic data, and the ship static data, the transportation relationships between ship entities and product entities are defined. The supply relationships between enterprise entities are defined according to the internal supply chain data, and based on the supplier-customer relationships in the purchase order data and the product BOM data in the internal supply chain data, the supply relationship attributes are determined, including supplied product attributes and dependency strength attributes. Based on the product-manufacturer correspondence in the product BOM data, the production relationships between enterprise entities and product entities are generated.

[0020] In one embodiment of this specification, the construction process of a dynamic knowledge graph begins with the collaborative acquisition and deep fusion processing of multi-source heterogeneous data, and the specific implementation process is as follows: The system continuously acquires AIS data streams from global ship broadcasts via satellite receiving stations and ground base station networks. This data stream includes dynamic fields such as the ship's unique MMSI identifier, real-time latitude and longitude coordinates, speed over land, heading over land, and estimated time of arrival. Simultaneously, it extracts port basic data (including port UN / LOCODE codes, latitude and longitude coordinates, number of berths, and cargo throughput capacity) and ship static data (including the mapping relationship between ship MMSI and physical parameters such as ship type, deadweight tonnage, length, and beam) from public maritime databases. Furthermore, it synchronously acquires the target company's internal supply chain data through standardized data interfaces of the target company's ERP system. This includes supplier master data recording the geographical location and qualifications of suppliers, product bill of materials (BOM) data describing the composition of product components, and purchase order data recording actual logistics needs.

[0021] First, the raw AIS data stream is preprocessed to improve data quality. An anomaly detection algorithm based on statistical laws is used to identify and remove latitude and longitude anomalies (such as coordinate drift points that deviate from the regular route). A time series interpolation algorithm is used to fill in the trajectory breaks caused by signal loss. Then, a map matching algorithm is used to associate the cleaned ship trajectory points with a predefined digital model of maritime routes. This algorithm transforms the discrete GPS point sequence into structured continuous ship trajectory data by calculating the spatial topological relationship between trajectory points and candidate routes. Each trajectory segment is associated with a defined origin port, destination port, and sailing time attribute.

[0022] Structured information is extracted from different data sources during entity extraction. Port entities are created based on port basic data and injected with port identifier, name, geographical location and design capacity attributes. Ship entities are created based on ship static data and injected with MMSI identifier, ship type classification and deadweight tonnage attributes. Enterprise entities (injecting enterprise unified social credit code, name and main business classification) and product entities (injecting product SKU code, name and industry classification code) are created based on enterprise internal supply chain data.

[0023] Semantic links between entities are constructed through multi-source data association analysis. Based on cleaned AIS data streams, a geofencing algorithm is used to construct electronic fence areas for ports on a digital map. When a ship's trajectory point continuously enters the fenced area and its speed is below a threshold, a berthing event detection is automatically triggered, generating a berthing relationship between the ship entity and the port entity. This relationship is then assigned start timestamps, end timestamps, and berthing duration attributes calculated based on the time difference. Based on continuous ship trajectory data, direct navigation patterns between ports are identified through trajectory sequence analysis, establishing connections between port entities. Route distances and cargo frequency within a statistical period are calculated by aggregating historical voyage data as connection relationship attributes.

[0024] Based on purchase order data and product BOM data from the supply chain data, the logistics route information of the target product is determined. This logistics route information includes the originating region, destination region, and product type. According to the logistics route corresponding to the product entity in the supply relationship, the port of departure and port of destination are determined. Vessel entities with berthing relationships with the ports of departure and destination are identified, and their historical navigation information is obtained to confirm the matching degree between their route coverage and the product logistics route. When the matching degree condition is met, a transportation relationship is established between the corresponding vessel entity and the product entity. Based on purchase order records in the enterprise's internal supply chain data, the correspondence between supplier numbers and customer numbers in the orders is analyzed to establish supply relationships between enterprise entities. Simultaneously, combined with the material composition relationships recorded in the product BOM data, the supply relationship is injected with the supplied product type and dependency strength attributes calculated based on order amount and delivery frequency. Based on the mapping relationship between product numbers and manufacturer numbers in the product BOM data, the production relationship between enterprise entities and product entities is directly established.

[0025] The aforementioned entities and relationships are imported into a graph database (such as Neo4j) to construct an initial knowledge graph, where entities are stored as nodes and relationships as edges. All attributes are attached to the corresponding nodes and edges in the form of key-value pairs. At the same time, a real-time data consumption pipeline is established to continuously process the incoming AIS data stream. The latitude and longitude attributes of ship entities are dynamically updated through a streaming computing engine. The number of ships queuing and the expected congestion level of port entities are updated in real time based on the latest berthing event data. Incremental updates of the knowledge graph are triggered through an event-driven mechanism to ensure that the graph state remains synchronized with the physical world.

[0026] The aforementioned technical solution systematically integrates internal supply chain data with external dynamic AIS data streams and static port and vessel data, constructing a dynamic knowledge graph. This allows risk perception to move beyond static snapshots, continuously updating vessel positions, port status, and route relationships by consuming real-time AIS data streams. This shifts perception capabilities from post-event statistics to in-event awareness, enabling immediate response to sudden logistics disruptions. Furthermore, algorithms imbue relationships with dynamic and quantifiable attributes. For example, geofencing algorithms accurately extract vessel berthing events and their duration from AIS data, quantify port connectivity and route characteristics based on continuous vessel trajectories, and combine internal order and BOM data to precisely characterize supply chain dependencies between enterprises. This provides a refined, weighted entity relationship network, offering data granularity and accuracy for subsequent multi-dimensional quantitative analysis. This transforms supply chain risk assessment from a traditional model relying on experience and single indicators to one based on a dynamic global network, multi-dimensional quantifiability, and real-time diagnostics and simulation, significantly enhancing enterprises' understanding and decision-making capabilities regarding complex supply chain risks.

[0027] Based on the risk perception request, entity relationships are extracted from a pre-constructed dynamic knowledge graph to form an evaluation subgraph. Specifically, this includes: parsing the risk perception request and extracting the target product identifier and / or the target shipping port identifier; starting with the product entity corresponding to the target product identifier and / or the port entity corresponding to the target shipping port identifier, performing graph traversal in the dynamic knowledge graph to extract all related enterprise entities, product entities, port entities, ship entities and their supply relationships, production relationships, transportation relationships, berthing relationships and connection relationships, and determining the evaluation subgraph.

[0028] In one embodiment of this specification, upon receiving a risk perception request submitted by a target enterprise, the request is first parsed to extract key identification information from the request message. Specifically, the type of enterprise perception object specified in the request is identified. If the object is a target product, its unique product identifier is extracted, which typically corresponds to the SKU code or product number in the enterprise's internal material management system. If the object is a target shipping port, its standardized port identifier is extracted, which uses the internationally recognized UN / LOCODE encoding system to ensure global uniqueness. In mixed request scenarios, both product identifiers and port identifiers can be extracted simultaneously, establishing a correlation mapping for subsequent processing. In mixed-type perception request scenarios, both product identifiers and port identifiers can be extracted in parallel, and a logical correlation mapping between them can be established, laying the foundation for subsequent correlation analysis.

[0029] For product-oriented query paths, the query starts from the product entity node and executes a multi-level depth-first traversal algorithm in the graph database. The traversal process first traces backward along the production relationship edges to locate all enterprise entity nodes that produce the product, including finished product manufacturers and key component suppliers. Then, it expands forward along the supply relationship edges, recursively traversing all upstream raw material supplier enterprise entities to construct a complete product supply chain network. During this process, the transportation relationships associated with these enterprise entities are extracted simultaneously, connecting them to the ship entities undertaking logistics tasks through transportation relationship edges, and then linking them to the relevant port entities through the ship entities' berthing relationship edges, ultimately forming a complete product association subgraph. This subgraph not only includes the product's commercial supply relationships but also extends to the physical logistics level, achieving full-link coverage of commercial and logistical flows.

[0030] For port-oriented query paths, the system starts with port entity nodes and performs a multi-dimensional graph traversal based on the spatial network. The traversal process first retrieves all vessel entities related to the port in the current and historical periods through berthing relationship edges, obtaining the vessels' static attribute information and dynamic location data. Then, it discovers other port entities connected to the port by shipping routes through connection relationship edges, constructing the port network topology. Based on this, it further associates vessel entities with product entities being transported through transportation relationship edges, and then traces back to related enterprise entities through the supply and production relationship edges of the product entities, forming a complete port-related subgraph. This subgraph starts from physical logistics nodes and traces back to commercial supply chain entities, achieving a comprehensive mapping from logistics status to commercial impact.

[0031] During the subgraph fusion phase, when a request involves both products and ports, the product-related subgraph and the port-related subgraph are organically integrated. The fusion process is based on a predefined entity association rule base. It uses spatial matching algorithms to associate the geographical location of enterprise entities with the service area of ​​port entities, product-ship matching algorithms to match product characteristics with ship transportation capacity, and time-series analysis algorithms to align supply chain plans with logistics execution status. The fusion engine automatically identifies and eliminates duplicate entities, maintains data consistency, preserves the semantic relationships between all entities, and establishes cross-subgraph association indexes, ultimately generating a unified evaluation subgraph that contains both the complete supply chain business logic and reflects the entire logistics status. This evaluation subgraph serves as the foundational data model for subsequent quantitative analysis, providing comprehensive, accurate, and timely data support for precise risk perception.

[0032] Compared to traditional supply chain risk assessment methods, this technical solution significantly improves assessment accuracy and efficiency through an intelligent subgraph extraction mechanism based on dynamic knowledge graphs. The real-time constructed assessment subgraph accurately captures the latest state and complex relationships within the supply chain network. Dual-path traversal, using both product-oriented and port-oriented approaches, effectively addresses the disconnect between commercial and logistical analysis in traditional methods. Intelligent subgraph fusion organically unifies the business logic of the product supply chain with the logistical status of the port network, enabling accurate analysis of the transmission path and impact of logistical disruptions on business operations, providing a more comprehensive perspective for risk management. By configuring different traversal strategies and fusion rules, it can flexibly address various complex risk assessment scenarios, maintaining consistent accuracy and reliability from single-product supply chain analysis to global port network risk assessment.

[0033] This assessment subgraph is used to calculate multi-dimensional quantitative resilience indicators, specifically including: calculating the rate of change of average port vessel berthing time and average sailing time on key routes based on the berthing relationship data in the assessment subgraph, generating a logistics efficiency dimension indicator; calculating node betweenness centrality, degree centrality, and network clustering coefficient using graph algorithms based on the topology of the assessment subgraph, generating a network topology dimension indicator; calculating path redundancy and the number of alternative suppliers between nodes using path data and supply relationship data in the assessment subgraph, generating a redundancy substitution dimension indicator; and calculating the propagation range of the impact of the outage event and the time required for network performance to recover to the baseline level based on pre-acquired historical outage event records and real-time AIS data, generating an anomaly response recovery dimension indicator.

[0034] In one embodiment of this specification, resilience metrics of different dimensions are processed based on entity and relationship data contained in the evaluation subgraph.

[0035] In calculating logistics efficiency indicators, the system first extracts all berthing relationship data from the evaluation subgraph, including the start time, end time, and berthing duration attributes of each berthing event. The system then uses a time-series analysis algorithm to calculate the average berthing time of vessels at each port node. This calculation process first groups berthing events at each port by time window, then calculates the average berthing duration within each time window, and finally detects abnormal fluctuation patterns through time series analysis. Simultaneously, the system calculates the average sailing time change rate for key routes. First, it identifies the main shipping routes in the evaluation subgraph, then calculates the historical average sailing time for each route based on continuous vessel trajectory data, then obtains the current sailing time through real-time AIS data stream, and finally calculates the deviation of the current value from the baseline value. The collaborative calculation of these two logistics efficiency indicators comprehensively reflects the operational efficiency status of the supply chain logistics network, providing fundamental performance data for resilience assessment.

[0036] In the calculation of network topology dimensional indicators, the evaluation subgraph is first converted into an adjacency list structure that can be processed by the graph computing engine, and then a parallel graph algorithm computing cluster is launched. First, node betweenness centrality is calculated using the Brandes algorithm on a distributed graph computing framework. This algorithm identifies key hub nodes carrying the most traffic by traversing the shortest paths between all node pairs and counting the number of times each node acts as a bridge. Next, degree centrality is calculated, distinguishing core nodes from edge nodes by counting the number of connected edges for each node. Finally, the network clustering coefficient is calculated based on the triangle counting principle, assessing the degree of local clustering by analyzing the ratio of actual connections to potential connections between each node's neighbors. The results of these topology indicator calculations collectively reveal the structural characteristics of the supply chain network, providing a quantitative basis for evaluating network connectivity and robustness.

[0037] When calculating the redundancy substitution dimension indicators, a multi-path analysis algorithm is used to calculate the path redundancy between nodes. First, key product flow paths are selected in the evaluation subgraph. Then, the Yen algorithm is used to find K shortest paths. The path redundancy level is assessed by comparing the overlap and cost differences of these paths. Simultaneously, the number of alternative suppliers is calculated using the supplier network analysis module. A supplier-product bipartite graph is constructed based on supply relationship data. Then, the diversity of supply sources for each key product is analyzed, and the number of alternative suppliers with the same supply capacity is counted. These two redundancy indicators assess the backup capability of the supply chain from both path and source dimensions, providing important redundancy reference data for resilience assessment.

[0038] When calculating the metrics for anomaly response and recovery, a historical disruption event knowledge base is first established, integrating historical disruption records from external event data sources. When a new disruption event is detected, a graph propagation algorithm is used to simulate the impact range. This algorithm starts from the disruption node and performs a breadth-first traversal along connection and supply relationships, counting the number of affected enterprises and products. Simultaneously, a performance recovery monitoring module calculates the network recovery speed, continuously tracking the recovery process of various efficiency indicators after the disruption, and recording the time required to recover from the abnormal state to the baseline level. The calculation of these dynamic response metrics enables the assessment of the supply chain's resilience and recovery capabilities in the face of disruption events, enriching the time dimension of resilience assessment.

[0039] Compared to traditional supply chain assessment methods, this technical solution achieves a fundamental improvement in assessment depth and accuracy through a multi-dimensional quantitative indicator calculation framework. By coordinating calculations across four dimensions, it comprehensively characterizes the resilience features of the supply chain from multiple perspectives, including efficiency, structure, redundancy, and responsiveness. Through specially optimized calculation algorithms and a parallel processing architecture, it effectively solves the performance bottleneck problem of traditional methods when handling large-scale complex networks. Using a distributed graph computing engine and real-time stream processing technology, it can perform full calculations on supply chain networks with hundreds of millions of nodes, ensuring the accuracy and timeliness of the assessment results. Each calculated indicator corresponds to a clear management meaning and improvement direction, enabling enterprises to formulate targeted optimization strategies based on the assessment results. This assessment method based on multi-dimensional quantitative indicators transforms the originally abstract concept of supply chain resilience into a concrete and measurable management object, significantly improving the scientific nature and effectiveness of supply chain risk management compared to traditional assessment methods that rely on experience-based judgment.

[0040] Step S103: Integrate multi-dimensional quantitative resilience indicators to generate a comprehensive resilience index corresponding to the enterprise's perceived object, and use the comprehensive resilience index to perceive risks in the enterprise's perceived object.

[0041] The multi-dimensional quantitative resilience indicators are integrated to generate a comprehensive resilience index corresponding to the enterprise's perceived object. Specifically, this includes: standardizing the logistics efficiency dimension indicator, the network topology dimension indicator, the redundancy substitution dimension indicator, and the anomaly response recovery dimension indicator to generate standardized indicator values; calculating the weight coefficients of each dimension indicator using the entropy weight method, where the input of the entropy weight method is the historical data sequence of each dimension indicator; and calculating the comprehensive resilience index based on the standardized indicator values ​​and weight coefficients using a weighted summation method.

[0042] In one embodiment of this specification, a comprehensive resilience index is generated based on three core stages: standardization, weight calculation, and weighted summation.

[0043] First, data normalization was performed on the four dimensions of quantitative resilience indicators. For the logistics efficiency dimension indicators, a min-max normalization method was used, which linearly transformed the original indicator values ​​to a unified numerical range based on the minimum and maximum values ​​of each indicator according to historical data. Negative indicators such as average port vessel berthing time were reversed to ensure that all indicators contributed to the overall index in the same direction. For the network topology dimension indicators, a Z-score normalization method was used, which linearly transformed the values ​​based on the mean and standard deviation of each indicator to eliminate differences in magnitude between indicators with different centralities. For the redundancy substitution dimension indicators, a hybrid method combining logarithmic transformation and min-max normalization was used. First, a logarithmic transformation was performed on indicators with long-tailed distributions, such as path redundancy, to improve the distribution pattern, followed by range normalization. For the anomaly response recovery dimension indicators, a quantile normalization method was used, which nonlinearly transformed the values ​​based on quantile statistics from historical data to effectively reduce the impact of outliers on the normalization results.

[0044] The entropy weighting method is used to automatically determine the weight coefficients of each dimension of the indicators. First, a historical indicator data matrix is ​​constructed by extracting sufficient time-span sequence data of each dimension from the historical indicator database to form a complete indicator observation dataset. Then, the information entropy value of each indicator is calculated. Specifically, this process includes calculating the proportion of each indicator in all observation samples and calculating the information entropy of each indicator based on information theory principles. The magnitude of the entropy value reflects the indicator's discriminative ability in the evaluation process. Indicators with lower information entropy indicate greater fluctuations across different samples and contain more information, thus requiring a greater weight in the comprehensive evaluation. The difference coefficient of each indicator is calculated through mathematical derivation, and weight values ​​are allocated proportionally according to the magnitude of the difference coefficient, ensuring that the sum of all weight coefficients is a fixed value. The entire weight calculation process is entirely data-driven, avoiding the influence of subjective factors on weight allocation. Furthermore, historical data samples are updated periodically, and weight coefficients are recalculated to adapt to changes in the characteristics of the supply chain network.

[0045] In the weighted summation stage, the standardized index values ​​are first multiplied by their corresponding weight coefficients using a scalar multiplication operation to obtain the weighted score for each index. Then, a linear weighted summation model is used to accumulate the weighted scores of all indicators to obtain a preliminary comprehensive index value. The preliminary calculation results are then adjusted for range and smoothed to ensure that the final output comprehensive resilience index falls within a preset numerical range and maintains smoothness and stability over time.

[0046] To ensure the transparency and traceability of the calculation process, the system will fully record the intermediate results of each calculation step, including the standardized values, weight coefficients, weighted scores and other detailed information of each indicator. This data can be used for verification of the calculation process and as the basis for subsequent analysis. It can efficiently process large-scale indicator data and automatically start backup calculation schemes when some data is abnormal, ensuring the real-time performance and reliability of the comprehensive resilience index calculation.

[0047] Compared to traditional supply chain assessment index construction methods, this technical solution significantly improves the objectivity and accuracy of assessment results through a scientific indicator fusion mechanism. It employs the entropy weight method to automatically calculate weights based on the characteristics of the data itself, ensuring the objectivity and scientific nature of weight allocation and making assessment results from different times and objects comparable. This solution effectively solves the problems of inconsistent dimensions and distribution characteristics of multiple indicators in traditional methods through a multi-level standardization process. Based on the data characteristics of each dimension indicator, the most suitable standardization method is adopted, fully preserving the information characteristics of the original data and laying a solid foundation for subsequent weight calculation and index synthesis. Furthermore, by continuously collecting historical indicator data, the weight coefficients can be automatically adjusted to adapt to changes in supply chain network characteristics. This data-driven weight update mechanism ensures that the assessment model remains consistent with reality, better capturing the dynamic evolution of the supply chain network and providing more accurate quantitative basis for risk management decisions, significantly improving the timeliness and practicality of supply chain resilience assessment.

[0048] Based on this comprehensive resilience index, risk perception is performed on the enterprise's perceived objects. Specifically, when the risk perception request is real-time, the comprehensive resilience index is compared with a preset risk level threshold to generate an initial risk level. When the comprehensive resilience index is detected to be lower than the preset threshold, key risk paths and single failure points are identified from the assessment sub-graph. A risk warning report is generated based on the identification results, and the risk is displayed through a visual interface.

[0049] In one embodiment of this specification, after obtaining the comprehensive resilience index, when the risk perception request is real-time perception, the automatic identification, analysis and early warning of supply chain risks are realized based on preset risk assessment rules and intelligent analysis algorithms.

[0050] First, the calculated comprehensive resilience index is input into the risk level classification module. This module has a built-in multi-level risk threshold system. These thresholds are dynamic parameters obtained through machine learning training on historical risk event data, and can adaptively adjust according to different industry characteristics and corporate risk preferences. The risk level classification module uses a fuzzy logic algorithm to handle the relationship between the comprehensive resilience index and the threshold boundaries. It calculates the degree to which the current index belongs to each risk level through a membership function, thereby generating an initial risk level judgment result including confidence. It not only considers the current absolute index value but also combines the index's time-varying trend for a comprehensive judgment. Through analysis of the index's first and second derivatives, it identifies potential risk states where the current index is within a safe range but shows a rapid deterioration trend.

[0051] When the overall resilience index is detected to be below a preset threshold or shows a significant deterioration trend, in-depth root cause analysis is performed. First, a critical path identification algorithm is executed in the evaluation subgraph, using an improved Dijkstra algorithm to find the most vulnerable transmission paths in the supply chain network. These paths are characterized by bottlenecks that perform poorly across multiple dimensions, including logistics efficiency, network topology, and redundancy substitution. Simultaneously, a node criticality assessment model from complex network theory is used to calculate the impact of removing each node on network connectivity. This, combined with the node's topological location and actual traffic load, accurately pinpoints individual failure points. The analysis process comprehensively considers real-time node status data, such as current port congestion levels, ship punctuality performance, and supplier capacity utilization, ensuring the timeliness and accuracy of identified risk points and critical paths. After risk identification and analysis, a structured risk warning report is automatically generated using natural language generation technology and visualization analysis components. The report generation process first organizes the basic framework of the report, including standard modules such as executive summary, risk overview, impact analysis, and response recommendations. Then, specific analysis results, indicator data, trend charts, and other content are embedded in the appropriate locations.

[0052] Finally, risks are displayed through an integrated visualization interface. This interface adopts a multi-view collaborative design concept. In the main view area, the physical distribution of risks is displayed using a WebGIS-based map component, presenting the risk intensity of different geographical areas in the form of a heat map, and using icons of different colors and sizes to mark key risk points and vulnerable paths. In the auxiliary view area, the logical structure of the supply chain network is displayed in the form of a topology map, highlighting identified single failure points and key risk paths, and using animation effects to show the transmission path and impact range of risks. Users can also click on any risk node to view detailed indicator data and influencing factors, achieving a comprehensive risk investigation from the macro situation to the micro root causes. All visualization components support real-time data updates; when the underlying data changes, the interface display automatically refreshes to ensure that users always have access to the latest risk status information.

[0053] Compared to traditional supply chain risk monitoring systems, this technical solution significantly improves risk management efficiency and effectiveness through intelligent risk perception and early warning mechanisms. Utilizing a dynamic threshold system and machine learning algorithms, it automatically adapts to risk assessment needs in different scenarios, greatly reducing system maintenance complexity and the frequency of manual intervention. This solution addresses the limitation of traditional methods' single analytical perspective through multi-dimensional fusion analysis technology. By combining topology analysis, path analysis, and real-time status monitoring, it reveals the deep-seated mechanisms and potential impacts of risks. Intelligent report generation and an interactive visualization interface greatly enhance the efficiency of risk information transmission and user experience, ensuring not only the timeliness of information but also the quality and consistency of content through standardized templates.

[0054] Based on the comprehensive resilience index, risk perception is performed on the enterprise's perceived objects. Specifically, this includes: when the risk perception request is a simulated interruption perception, removing or disabling the target entities and / or entity relationships corresponding to the enterprise's perceived objects in the assessment subgraph, and generating a simulated assessment subgraph after the interruption; based on the simulated assessment subgraph, multi-dimensional quantitative resilience indicators are calculated, and a simulated resilience index is generated through weighted fusion; based on the relationship between the simulated resilience index and the comprehensive resilience index, an impact assessment result is generated, and the impact path and impact degree are displayed through a visualization interface.

[0055] In one embodiment of this specification, when the risk perception request is a simulated interruption perception, the system first receives an interruption simulation command input by the user through a visual interface. This command is obtained through an interactive operation interface. The user can select the interruption target by directly clicking on a specific node or connecting edge in the supply chain network diagram, or specify the scope and extent of the interruption through a parameter configuration interface. The interruption command input by the user is semantically parsed and undergoes compliance checks to ensure that the interruption operation conforms to physical logic and business rules, for example, avoiding contradictory scenarios where a certain enterprise entity and its sole supplier are simultaneously interrupted. After parsing, a copy of the evaluation subgraph is created in memory as the basic data environment for the interruption simulation.

[0056] Next, entity and relationship removal or disabling operations are performed, a process implemented through the graph data manipulation engine. For node interruptions, the specified entity node is deleted from the evaluation subgraph copy, and all relation edges connected to that node are automatically removed. Simultaneously, the attributes and connection information of the removed entities are recorded for subsequent recovery and comparative analysis. For relationship interruptions, the specified semantic relationship is disabled, but the entity nodes at both ends are retained. This mode is suitable for simulating situations such as transportation route disruptions or temporary suspension of supply relationships. During the execution of removal or disabling operations, an impact scope pre-assessment mechanism is initiated. A fast graph traversal algorithm is used to pre-determine the potential scope of entities affected by the interruption operation and to warn of possible network segmentation risks.

[0057] After generating the simulated evaluation subgraph after the disruption, resilience metrics are recalculated, but with specific optimizations for the simulation scenario. In terms of logistics efficiency, the transport time and port turnaround efficiency of affected routes are recalculated, considering detour times and node congestion effects caused by route changes. In terms of network topology, the connectivity and node importance of the network after the disruption are re-analyzed, identifying newly emerging critical nodes and vulnerable paths. In terms of redundancy and alternatives, the diversity of remaining paths and the accessibility of alternative suppliers are assessed, analyzing the impact of the disruption event on the supply chain's backup capabilities. In terms of anomaly response and recovery, recovery time and recovery paths are predicted under the current disruption scenario based on recovery patterns from similar historical disruption events. All metric recalculations employ incremental calculation techniques, re-evaluating only the affected parts of the network, significantly improving computational efficiency.

[0058] After recalculating the indicators, they are fused as described above to generate a simulated resilience index. A comparative analysis framework is established between the simulated resilience index and the original comprehensive resilience index. A difference calculation model quantifies the degree of resilience loss caused by outage events. This model considers not only the difference in absolute values ​​but also analyzes the changing patterns and trend characteristics of indicators across various dimensions. Through a root cause analysis algorithm, changes in the resilience index are decomposed into specific network paths and business processes, identifying the key factors and transmission mechanisms leading to a decline in resilience.

[0059] Finally, the impact assessment results are presented through a specially designed simulation results visualization interface. This interface employs a multi-view comparison layout: the left side displays the supply chain network status and resilience indicators before the disruption, the right side displays the simulation results after the disruption, and the central area uses a dynamic flowchart to illustrate the transmission path and degree of impact. Color coding and animation effects visually demonstrate the risk transmission process, using different colors to highlight different levels of impact, and flow animations to show the diffusion path of the disruption's impact within the supply chain network. Simultaneously, a quantitative analysis panel of the impact degree is provided, detailing key indicators such as the number of affected entities, product types, estimated transportation delays, and additional logistics costs. Furthermore, natural language processing technology automatically generates a simulation analysis report, summarizing the main impact characteristics of the disruption event and providing response recommendations, offering comprehensive decision support for users to formulate supply chain resilience enhancement strategies.

[0060] Compared with traditional supply chain risk assessment methods, this solution represents a significant shift from passive response to proactive prevention. Enterprises can identify vulnerable links and develop contingency plans before actual risks occur, significantly improving the initiative and foresight of supply chain risk management. Through refined simulation modeling and rapid recalculation mechanisms, this solution addresses the inefficiencies and inaccuracies of traditional simulation methods. Based on a complete dynamic knowledge graph and full index recalculation, it can accurately simulate the transmission process and impact range of disruption events in the supply chain network, providing more reliable data support for decision-making.

[0061] Through the embodiments described in this specification, a dynamic knowledge graph integrating global shipping dynamics and enterprise internal supply chain data is constructed, fundamentally solving the problem of the separation of commercial and logistics data in traditional methods. It can systematically reveal the complete path and impact of a single port congestion or supplier disruption transmitted through a complex network to the final product. A multi-dimensional quantitative resilience index system is introduced, performing collaborative calculations from four interrelated and complementary dimensions: logistics efficiency, network topology, redundancy substitutability, and anomaly response recovery. This not only reveals the current operational efficiency of the supply chain (e.g., through route delay rates) but also deeply analyzes its inherent structural characteristics (e.g., node centrality) and its ability to withstand single [unspecified events]. The system assesses the redundancy of failures (such as path diversity) and the recovery potential after impacts. It generates a comprehensive resilience index through weighted fusion and supports both real-time perception and simulated interruption modes. The comprehensive index condenses multi-dimensional information into an intuitive quantitative scale, overcoming the problem of difficulty in coordinating decision-making among multiple indicators. Simulated interruption perception allows enterprises to conduct stress tests in the digital space before risk events occur, proactively acquiring the potential impact and transmission path of interruptions in critical product dependency paths or core ports. This transforms risk management from passively responding to historical events to proactively preventing future risks, meeting the real-time, multi-dimensional, and predictable risk perception needs of global supply chain enterprises for specific supply chains.

[0062] This specification also provides an embodiment of a risk perception device for global supply chain enterprises, such as... Figure 2 As shown, the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method.

[0063] This specification also provides a non-volatile computer storage medium storing computer-executable instructions configured to perform the above-described method.

[0064] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0065] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A risk perception method for global supply chain enterprises, characterized in that, The method includes: Receive risk perception requests from target enterprises for enterprise perception objects, wherein the enterprise perception objects include target products and target shipping ports, and the risk perception requests include real-time perception and simulated interruption perception; Based on the risk perception request, entity relationships are extracted from a pre-built dynamic knowledge graph to form an evaluation subgraph, which is then used to calculate multi-dimensional quantitative resilience indicators. The multi-dimensional quantitative resilience indicators are fused to generate a comprehensive resilience index corresponding to the enterprise perception object, so as to perceive the risk of the enterprise perception object based on the comprehensive resilience index.

2. The risk perception method for global supply chain enterprises according to claim 1, characterized in that, Before extracting entity relationships from a pre-built dynamic knowledge graph to form an evaluation subgraph based on the risk perception request, the method further includes: Acquire global ship AIS data streams, port basic data, ship static data, and the target company's internal supply chain data, wherein the internal supply chain data includes supplier master data, product bill of materials (BOM) data, and purchase order data; The global ship AIS data stream is cleaned to remove latitude and longitude outliers and imput missing values. The cleaned AIS data stream is then processed by a map matching algorithm to associate ship trajectories with predefined sea routes to determine continuous ship trajectory data. Based on the port basic data, the ship static data, and the internal supply chain data, multiple entities are identified, including port entities, ship entities, enterprise entities, and product entities. The entity relationships between the multiple entities are defined by the cleaned AIS data stream, the continuous ship trajectory data, and the internal supply chain data. These entity relationships include berthing relationships, connection relationships, transportation relationships, supply relationships, and production relationships. An initial knowledge graph is constructed by storing the multiple entities and their relationships in a graph database, and a dynamic knowledge graph is constructed by dynamically updating the ship location attributes and port status attributes by consuming real-time AIS data streams.

3. A risk perception method for global supply chain enterprises according to claim 2, characterized in that, The entity relationships between the multiple entities are defined using the cleaned AIS data stream, the continuous ship trajectory data, and the internal supply chain data, specifically including: Based on the cleaned AIS data stream, the berthing relationship between ship entities and port entities is defined, and the berthing event of the ship is detected by the geofencing algorithm to generate the berthing relationship attribute corresponding to the berthing relationship. The berthing relationship attribute includes start time, end time and berthing duration. Using the continuous ship trajectory data, the connection relationships between port entities are defined to calculate the route distance and historical cargo frequency between ports, and to generate the connection relationship attributes corresponding to the connection relationships. Based on the purchase order data in the supply chain data, the port basic data, and the ship static data, the transportation relationship between the ship entity and the product entity is defined. The supply relationships between enterprise entities are defined based on the internal supply chain data, and the supply relationship attributes corresponding to the supply relationships are determined based on the supplier-customer relationships in the purchase order data and the product BOM data of the internal supply chain data. The supply relationship attributes include the supplied product attributes and the dependence strength attributes. Based on the product-manufacturer correspondence in the product BOM data, the production relationship between the enterprise entity and the product entity is generated.

4. A risk perception method for global supply chain enterprises according to claim 1, characterized in that, Based on the risk perception request, entity relationships are extracted from a pre-constructed dynamic knowledge graph to form an evaluation subgraph, specifically including: Parse the risk perception request and extract the target product identifier and / or target shipping port identifier; Starting with the product entity corresponding to the target product identifier and / or the port entity corresponding to the target shipping port identifier, a graph traversal is performed in the dynamic knowledge graph to extract all related enterprise entities, product entities, port entities, ship entities and their supply relationships, production relationships, transportation relationships, berthing relationships and connection relationships, and to determine the evaluation subgraph.

5. A risk perception method for global supply chain enterprises according to claim 1, characterized in that, The evaluation subgraph is used to calculate a multi-dimensional quantitative resilience index, specifically including: Based on the berthing relationship data in the evaluation subgraph, the change rate of the average port berthing time of port vessels and the average sailing time of key routes is calculated to generate logistics efficiency dimension indicators. Based on the topology of the evaluation subgraph, the graph algorithm is used to calculate the node betweenness centrality, degree centrality, and network clustering coefficient to generate network topology dimension indexes. By using the path data and supply relationship data in the evaluation subgraph, the path redundancy between nodes and the number of alternative suppliers are calculated to generate redundancy and alternative dimension indicators. Based on pre-acquired historical outage event records and real-time AIS data, the scope of the outage event's impact and the time required for network performance to recover to the baseline level are calculated, generating anomaly response recovery dimension indicators.

6. A risk perception method for global supply chain enterprises according to claim 5, characterized in that, The multi-dimensional quantitative resilience indicators are fused to generate a comprehensive resilience index corresponding to the enterprise's perceived object, specifically including: The logistics efficiency dimension index, the network topology dimension index, the redundancy replacement dimension index, and the anomaly response recovery dimension index are standardized respectively to generate standardized index values. The entropy weight method is used to calculate the weight coefficients of each dimension indicator, where the input of the entropy weight method is the historical data sequence of each dimension indicator; The comprehensive resilience index is calculated using a weighted summation method based on standardized index values ​​and weighting coefficients.

7. A risk perception method for global supply chain enterprises according to claim 1, characterized in that, Based on the comprehensive resilience index, risk perception is conducted on the enterprise-aware entity, specifically including: When the risk perception request is real-time perception, the comprehensive resilience index is compared with a preset risk level threshold to generate an initial risk level; When the overall resilience index is detected to be below a preset threshold, key risk paths and single failure points are identified from the evaluation sub-graph. Risk warning reports are generated based on the identification results, and the risks are displayed through a visual interface.

8. A risk perception method for global supply chain enterprises according to claim 1, characterized in that, Based on the comprehensive resilience index, risk perception is conducted on the enterprise-aware entity, specifically including: When the risk perception request is a simulated interruption perception, the target entity and / or entity relationship corresponding to the enterprise perception object is removed or disabled in the evaluation subgraph to generate a simulated evaluation subgraph after the interruption. Based on the simulated evaluation subgraph, multi-dimensional quantitative resilience indicators are calculated, and a simulated resilience index is generated by weighted fusion. The impact assessment results are generated by analyzing the relationship between the simulated resilience index and the comprehensive resilience index, and the impact path and impact level are displayed through a visualization interface.

9. A risk perception device for global supply chain enterprises, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-8.

10. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are configured to perform the method as described in any one of claims 1-8.

Images