CN-121766788-B - Causal reasoning-based supply chain risk prediction method apparatus, device and medium

CN121766788BCN 121766788 BCN121766788 BCN 121766788BCN-121766788-B

Abstract

The application relates to the technical field of supply chain risk prediction based on causal reasoning, and discloses a supply chain risk prediction method, device, equipment and medium based on causal reasoning, wherein the method comprises the steps of realizing real-time fusion and characterization of ground-edge regulation and logistics operation multi-source heterogeneous data by constructing a double-layer knowledge graph of a regulation layer knowledge graph and a logistics physical layer knowledge graph, improving perception instantaneity and coverage dimension of a risk event, automatically generating a multi-stage causal chain by introducing a large language model, overcoming the defect of insufficient hidden association mining in the traditional method, improving reasoning about risk tracing, and finally obtaining a risk prediction result by mapping causal chain link points to the logistics physical layer knowledge graph and carrying out probabilistic calculation. The risk prediction method has the beneficial effects that the risk prediction of the appointed supply chain is realized, and the accuracy of the risk prediction result is improved.

Inventors

WANG XUETENG
ZENG WEIJIA
CHEN DAWEI
XU LINGZI
HE ZHONGQING
XU KUNYANG
ZHAO SHAN
Xie Qiongbing

Assignees

深圳市明心数智科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260304

Claims (8)

1. A causal reasoning-based supply chain risk prediction method, the method comprising: Acquiring a plurality of initial logistics operation data and a plurality of initial ground-edge regulation data related to a designated supply chain in real time; Carrying out standardization processing on each initial logistics operation data and each initial ground edge regulation data to obtain standard logistics operation data corresponding to each initial logistics operation data and standard ground edge regulation data corresponding to each initial ground edge regulation data; Constructing a regulation layer knowledge graph based on each standard ground-edge regulation data, and constructing a physical layer knowledge graph based on each standard logistics operation data; Acquiring perceived risk events in the legal layer knowledge graph and the physical layer knowledge graph of the logistics; Inputting the risk event into a large language model to generate a multi-stage causal chain for conducting the risk event in a designated supply chain, wherein the multi-stage causal chain comprises at least one conducting node; mapping the multi-level causal chain to the physical layer knowledge graph of the logistics to predict risk of the designated supply chain; The step of constructing a legal layer knowledge graph based on each standard ground law data, and the step of constructing a physical layer knowledge graph based on each standard logistics operation data, wherein the standard logistics operation data comprises track data, port throughput data and transportation hub state data of an automatic ship identification system, and the step of constructing the legal layer knowledge graph based on each standard ground law data comprises the following steps: analyzing each initial earth-edge regulation data by adopting a preset cross-language pre-training model to obtain an analysis result, wherein the analysis result comprises an entity, an event and relation information; Based on the entity, the event and the relation information of the analysis result, constructing nodes and edges of the knowledge graph to obtain an initial knowledge graph; Generating a quantized tension index based on the analysis result, and integrating the tension index into the initial knowledge graph as a dynamic attribute of a corresponding entity or relationship to obtain a legal layer knowledge graph, wherein the tension index is generated in a manner of determining an influence factor, calculating an original score by adopting a regularized grading, supervised learning or mixed integration model, carrying out normalization processing on the original score, and designing a time attenuation function to reflect the dynamic property that the influence of an event decreases with time or rises due to the superposition of new events so as to obtain the tension index; The step of mapping the multi-stage causal chain to the physical layer knowledge graph of the logistics to perform risk prediction on the designated supply chain comprises: Associating the conducting nodes in the multi-stage causal chain with the entity nodes in the physical layer knowledge graph of the logistics to form a risk propagation network; calculating confidence probabilities of risk conduction along different paths in the risk propagation network by adopting a probability propagation algorithm, wherein the confidence probabilities are used for quantifying risk prediction results; And carrying out risk prediction on the appointed supply chain according to the confidence probability.
2. The causal reasoning-based supply chain risk prediction method of claim 1, wherein the standard logistics operation data comprises trajectory data, port throughput data and transportation hub status data of a ship automatic identification system, and wherein the step of constructing a regulatory layer knowledge graph based on each of the standard ground-edge regulation data and constructing a logistics physical layer knowledge graph based on each of the standard logistics operation data comprises: acquiring track data, port throughput data and transportation hub state data of the ship automatic identification system from the standard logistics operation data; and constructing a physical layer knowledge graph taking the logistics nodes as vertexes and taking the transportation route as edges based on the track data, the port throughput data and the transportation hub state data.
3. The causal inference-based supply chain risk prediction method of claim 1, wherein the step of inputting the risk event into a large language model to generate a multi-level causal chain in which the risk event is conducted in a designated supply chain, comprises: Inputting the risk event as a prompt into a large language model; the large language model is driven to infer and output a multi-level causal chain from the risk event source via at least one intermediate conductive link based on supply chain domain knowledge.
4. The causal inference-based supply chain risk prediction method of claim 1, wherein after the step of risk predicting the designated supply chain based on the confidence probabilities, further comprises: Acquiring new event data; Dynamically adjusting the influence weights of nodes and edges in the risk propagation network through a preset reinforcement learning frame based on the new event data to obtain an updated risk propagation network; Recalculating the confidence probability of the risk conducted along the path based on the updated risk propagation network; And performing risk prediction on the appointed supply chain based on the recalculated confidence probability.
5. The causal reasoning based supply chain risk prediction method of claim 1, wherein after the step of mapping the multi-level causal chain to the physical layer knowledge graph of the logistics to risk predict the designated supply chain, further comprises: Acquiring a risk prediction result of risk prediction of the appointed supply chain and a corresponding risk conduction path; and inputting the risk prediction result and the corresponding risk conduction path into a preset large language model to generate a classified coping strategy.
6. A causal reasoning-based supply chain risk prediction apparatus, the apparatus comprising: the first acquisition module is used for acquiring a plurality of initial logistics operation data and a plurality of initial ground-edge regulation data related to a designated supply chain in real time; The processing module is used for carrying out standardized processing on the initial logistics operation data and the initial ground edge regulation data to obtain standard logistics operation data corresponding to the initial logistics operation data and standard ground edge regulation data corresponding to the initial ground edge regulation data; The building module is used for building a rule layer knowledge graph based on the standard ground edge rule data and building a physical layer knowledge graph based on the standard logistics operation data; the second acquisition module is used for acquiring perceived risk events in the legal layer knowledge graph and the physical layer knowledge graph of the logistics; An input module for inputting the risk event into a large language model to generate a multi-stage causal chain for conducting the risk event in a designated supply chain, wherein the multi-stage causal chain comprises at least one conducting node; the mapping module is used for mapping the multi-stage causal chain to the physical layer knowledge graph of the logistics so as to predict the risk of the appointed supply chain; The standard logistics operation data comprise track data, port throughput data and transportation hub state data of the ship automatic identification system, and the construction module comprises: the analysis sub-module is used for analyzing the initial ground-edge regulation data by adopting a preset cross-language pre-training model to obtain analysis results, wherein the analysis results comprise entities, events and relationship information; The initial knowledge graph acquisition sub-module is used for constructing nodes and edges of the knowledge graph based on the entity, the event and the relation information of the analysis result so as to obtain an initial knowledge graph; The integration sub-module is used for generating a quantized tension index based on the analysis result, integrating the tension index into the initial knowledge graph as a dynamic attribute of a corresponding entity or relation to obtain a legal layer knowledge graph, wherein the generation mode of the tension index is that an influence factor is determined, an original score is calculated by adopting a regularized scoring, supervised learning or mixed integration model, the original score is normalized, and a time attenuation function is designed to reflect the dynamic property that the influence of an event decreases with time or rises due to the superposition of new events, so that the tension index is obtained; The mapping module comprises: The association sub-module is used for associating the conducting nodes in the multi-stage causal chain with the entity nodes in the physical layer knowledge graph of the logistics to form a risk propagation network; the risk prediction system comprises a first calculation sub-module, a second calculation sub-module and a third calculation sub-module, wherein the first calculation sub-module is used for calculating the confidence probabilities of risk conduction along different paths in a risk propagation network by adopting a probability propagation algorithm, and the confidence probabilities are used for quantifying a risk prediction result; And the first risk prediction sub-module is used for performing risk prediction on the appointed supply chain according to the confidence probability.
7. A computer readable storage medium, characterized in that a computer program is stored, which, when being executed by a processor, causes the processor to perform the steps of the causal reasoning based supply chain risk prediction method as defined in any of claims 1 to 5.
8. An electronic device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the causal reasoning-based supply chain risk prediction method of any of claims 1 to 5.

Description

Causal reasoning-based supply chain risk prediction method apparatus, device and medium Technical Field The invention relates to the technical field of supply chain risk prediction based on causal reasoning, in particular to a supply chain risk prediction method, a supply chain risk prediction device, a supply chain risk prediction equipment and a supply chain risk prediction medium based on causal reasoning. Background The conventional supply chain risk prediction system mainly relies on a static rule base and historical data to perform single-dimension analysis, and has significant limitations. Firstly, the knowledge graph and the risk rule base which are relied on by the system are usually static, and multi-source heterogeneous data such as territory rule events, public opinion and the like which emerge in real time cannot be fused and analyzed dynamically, so that a prediction model is lagged, and the error rate is high. Second, existing methods typically only evaluate single node risk (e.g., port congestion), lack modeling and reasoning capabilities for complex causal conductive chains (e.g., raw material interruption-production stagnation-delivery violations), resulting in serious delays in risk early warning. Thus, there is a need for an intelligent prediction scheme. Disclosure of Invention Based on this, it is necessary to provide a causal reasoning-based supply chain risk prediction method, device, equipment and medium for solving the existing causal reasoning-based supply chain risk prediction problem. A causal reasoning-based supply chain risk prediction method, the method comprising: Acquiring a plurality of initial logistics operation data and a plurality of initial ground-edge regulation data related to a designated supply chain in real time; Carrying out standardization processing on each initial logistics operation data and each initial ground edge regulation data to obtain standard logistics operation data corresponding to each initial logistics operation data and standard ground edge regulation data corresponding to each initial ground edge regulation data; Constructing a regulation layer knowledge graph based on each standard ground-edge regulation data, and constructing a physical layer knowledge graph based on each standard logistics operation data; Acquiring perceived risk events in the legal layer knowledge graph and the physical layer knowledge graph of the logistics; Inputting the risk event into a large language model to generate a multi-stage causal chain for conducting the risk event in a designated supply chain, wherein the multi-stage causal chain comprises at least one conducting node; mapping the multi-level causal chain to the physical layer knowledge graph of the logistics to make risk prediction for the designated supply chain. Further, the standard logistics operation data includes, but is not limited to, trajectory data, port throughput data and transportation hub status data of the ship automatic identification system, and in the step of constructing a regulatory layer knowledge graph based on each of the standard ground law data, and constructing a physical layer knowledge graph based on each of the standard logistics operation data, the step of constructing a regulatory layer knowledge graph based on each of the standard ground law data includes: analyzing each initial earth-edge regulation data by adopting a preset cross-language pre-training model to obtain an analysis result, wherein the analysis result comprises an entity, an event and relation information; Based on the entity, the event and the relation information of the analysis result, constructing nodes and edges of the knowledge graph to obtain an initial knowledge graph; and generating a quantized tension index based on the analysis result, and integrating the tension index serving as a dynamic attribute of a corresponding entity or relationship into the initial knowledge graph to obtain a legal layer knowledge graph. Further, the standard logistics operation data includes, but is not limited to, trajectory data, port throughput data and transportation hub status data of the automatic ship identification system, and in the step of constructing a regulatory layer knowledge graph based on each of the standard ground-edge regulation data and constructing a logistics physical layer knowledge graph based on each of the standard logistics operation data, the step of constructing a logistics physical layer knowledge graph based on each of the standard logistics operation data includes: acquiring track data, port throughput data and transportation hub state data of the ship automatic identification system from the standard logistics operation data; and constructing a physical layer knowledge graph taking the logistics nodes as vertexes and taking the transportation route as edges based on the track data, the port throughput data and the transportation hub state data. Further, the step of inputting the risk event i