Supply chain risk assessment system based on multi-source data fusion

Abstract

The application relates to the technical field of supply chain risk assessment, in particular to a supply chain risk assessment system based on multi-source data fusion, which comprises a data acquisition module, a multi-source data acquisition end is configured, a binding relationship between the acquisition end and each node in a database is established, wherein the database is used for storing multi-source data related to supply chain risk assessment; a data fusion module is used for constructing a knowledge graph based on the multi-source data, wherein the nodes of the knowledge graph correspond to risk influence factors, and the edges represent the correlation between the factors; a model construction module is used for constructing a risk assessment model based on the knowledge graph; a risk assessment module is used for performing risk assessment on the supply chain based on the risk assessment model to obtain a risk assessment result; through the fusion of the multi-source data, the change of the multi-source data is monitored in real time, the dynamic change in the supply chain is responded in real time, the evaluation strategy is adjusted in a timely manner, and the accuracy and timeliness of the evaluation result are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supply chain risk management, in particular to a supply chain risk assessment system based on multi-source data fusion. BACKGROUND

[0002] A supply chain refers to a functional network structure organization that links raw material suppliers, product manufacturers, logistics transportation enterprises, warehousing managers, product distributors and end consumers into a whole around a core enterprise. Supply chain financial risk refers to the possibility of loss of participating subjects caused by previously unpredictable uncertainties or external disturbances within the supply chain during supply chain financial financing. In supply chain management, risk assessment is a key link to ensure the stable operation of the supply chain and reduce potential risks. With the increasing complexity of the supply chain, the data sources involved become increasingly diverse, including supplier information, logistics data, market demand data, etc. Traditional supply chain risk assessment methods often rely on a single data source or simple statistical models, which are difficult to comprehensively and accurately reflect the actual risk situation of the supply chain, affecting the efficiency and accuracy of the assessment.

[0003] To solve the above problems, we propose a supply chain risk assessment system based on multi-source data fusion. SUMMARY

[0004] The purpose of the present application is to provide a supply chain risk assessment system based on multi-source data fusion to solve the problems raised in the background.

[0005] To achieve the above purpose, the present application provides the following technical solution: a supply chain risk assessment system based on multi-source data fusion, comprising:

[0006] A data acquisition module is configured with a multi-source data acquisition end, and a binding relationship between the acquisition end and each node in the database is established, wherein the database is used to store multi-source data related to supply chain risk assessment; the multi-source data is classified and stored in a database composed of multiple sub-databases according to the category; the sub-database is bound to the acquisition end of the corresponding category basic data;

[0007] A data fusion module obtains the classified basic data and the corresponding sub-database from the database; extracts the features of the basic data as risk influence factors, obtains the correlation relationship as the correlation relationship between factors, sets the knowledge graph nodes and edges, and constructs a knowledge graph based on the nodes and edges;

[0008] A model construction module defines an association area including a main association area and a secondary association area with the knowledge graph node as the center; sets a virtual boundary layer, establishes a communication channel between the acquisition end and the node, and establishes a driving relationship between the sensing point and the risk assessment model; determines the change node and dynamically adjusts the association area; evaluates the association area and trains the evaluation model based on the comprehensive results;

[0009] The risk assessment module obtains a knowledge graph fused from multi-source data, determines an associated region, performs risk assessment on the supply chain based on a risk assessment model, and obtains a risk assessment result.

[0010] Preferably, the data collection module comprises:

[0011] The collection unit registers a collection end corresponding to the collection point, collects basic data for supply chain risk assessment based on the collection end, and obtains multi-source data for supply chain risk assessment.

[0012] The storage unit classifies the multi-source data to obtain multi-class basic data, and stores the multi-class basic data in the database according to the classes, wherein the database comprises a plurality of sub-databases, and each sub-database corresponds to a class of basic data.

[0013] The establishment unit binds the sub-database and the collection end of the basic data of the corresponding class.

[0014] Preferably, the data fusion module comprises:

[0015] The first acquisition unit acquires and extracts the classified basic data and the sub-database corresponding to the basic data from the database.

[0016] The data fusion unit extracts the data features of a class of basic data and the corresponding sub-database as a risk impact factor, obtains a plurality of risk impact factors based on the multi-class basic data, and obtains the association relationship between each class of basic data as the association relationship between factors.

[0017] The graph construction unit sets a plurality of nodes of the knowledge graph corresponding to the plurality of risk impact factors, sets a plurality of edges of the knowledge graph corresponding to the association relationship between the factors, and constructs the knowledge graph based on the plurality of nodes and the plurality of edges.

[0018] Preferably, the model construction module comprises:

[0019] The region determination unit takes each node in the knowledge graph as a center point, takes the node as a target node, defines all nodes and edges directly associated with the target node as an associated region of the target node, and the associated region comprises a main associated region and a secondary associated region.

[0020] The boundary construction unit sets a virtual boundary layer with sensing points in the knowledge graph, wherein the sensing points correspond one-to-one to the nodes, establishes a communication channel between the collection end and the corresponding node, wherein each communication channel corresponds to a sensing point, and establishes a driving relationship between the sensing point and the risk assessment model.

[0021] The regional adjustment unit determines a changed node as a changed node, and dynamically adjusts the associated region based on the changed node;

[0022] The model construction unit performs individual risk assessment on each associated region, identifies and eliminates the associated region with repeated calculation, integrates the assessment results of each associated region to obtain the risk assessment result corresponding to the knowledge graph, and trains the evaluation model based on the risk assessment results corresponding to multiple historical knowledge graphs.

[0023] Preferably, the step of defining all nodes and edges directly associated with the target node as the associated region of the target node includes:

[0024] The corresponding risk impact factor is set to an impact threshold value;

[0025] The risk impact factor exceeding the impact threshold value is taken as a core risk factor;

[0026] The risk impact factor not exceeding the impact threshold value is taken as an auxiliary risk factor; the node corresponding to the core risk factor is taken as a main node, and the node corresponding to the auxiliary risk factor is taken as a secondary node; the associated region corresponding to the plurality of main nodes and secondary nodes is obtained, and the plurality of main associated regions before adjustment and the plurality of secondary associated regions before adjustment are obtained.

[0027] Preferably, the step of determining the changed node as a changed node and dynamically adjusting the main associated region and the secondary associated region based on the changed node includes:

[0028] Based on the sensing point, the changed node is obtained as an active node, and the time interval between two consecutive changes of each active node is recorded;

[0029] For each active node, it is judged whether the latest change interval exceeds a preset time threshold;

[0030] The active node corresponding to the latest change interval exceeding the preset time threshold is marked as a changed node, and the data at the last change of the node is taken as the basic data corresponding to the node; the associated region corresponding to the changed node is taken as the adjusted main associated region;

[0031] The node not changed is taken as a non-changed node, and the associated region corresponding to the non-changed node is taken as the adjusted secondary associated region.

[0032] Preferably, the risk assessment module includes:

[0033] The second acquisition unit acquires the multi-source data for supply chain risk assessment from the database, and fuses them into a knowledge graph;

[0034] The risk assessment unit determines the associated area in the knowledge graph, and performs risk assessment on the associated area based on an evaluation model to obtain a risk assessment result.

[0035] Preferably, the risk assessment unit comprises:

[0036] The comparison subunit is configured to count the number of change nodes, calculate the proportion of change nodes in the knowledge graph based on the number of change nodes to obtain a change node proportion, and compare the change node proportion with a preset proportion threshold.

[0037] The initial evaluation subunit takes the number of change nodes corresponding to the change node proportion exceeding the preset proportion threshold as the number of abnormal change nodes, wherein the change nodes corresponding to the number of abnormal change nodes are referred to as abnormal change nodes, and performs risk assessment on the abnormal change nodes based on the main associated area before adjustment and the secondary associated area before adjustment.

[0038] The secondary evaluation subunit takes the number of change nodes corresponding to the change node proportion not exceeding the preset proportion threshold as the number of normal change nodes, wherein the change nodes corresponding to the number of normal change nodes are referred to as normal change nodes, and performs risk assessment on the normal change nodes based on the main associated area after adjustment and the secondary associated area before adjustment.

[0039] Compared with the prior art, the present application has the following advantages:

[0040] 1. By establishing the binding relationship between the collection end and each node in the database, the multi-source data related to the supply chain risk assessment can be comprehensively and accurately collected, the multi-source data is integrated, the risk influence factors and the correlation between the factors are intuitively presented, the associated area is dynamically adjusted according to the change nodes, the dynamic changes in the supply chain can be responded in real time, the evaluation strategy can be adjusted in time, and the accuracy and timeliness of the evaluation result are improved.

[0041] 2. In the definition process of the associated area, the influence threshold is set for the risk influence factors, the risk influence factors exceeding the influence threshold are taken as the core risk factors, and the risk influence factors not exceeding the influence threshold are taken as the auxiliary risk factors, the key risks are highlighted, the risk assessment is more targeted, the evaluation efficiency and accuracy are improved, and the evaluation results are ensured to be comprehensive and accurate. BRIEF DESCRIPTION OF DRAWINGS

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a system structure block diagram of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] For examples, please refer to [link / reference]. Figure 1 This invention provides a technical solution for a supply chain risk assessment system based on multi-source data fusion: the supply chain risk assessment system based on multi-source data fusion includes:

[0046] The data acquisition module configures multi-source data acquisition terminals and establishes binding relationships between the acquisition terminals and various nodes within the database. The database is used to store multi-source data related to supply chain risk assessment. The multi-source data is classified and stored in a database composed of multiple sub-databases according to their categories. The sub-databases are then bound to the acquisition terminals of the corresponding categories of basic data.

[0047] The data acquisition module includes:

[0048] Collection Unit: Collection points are deployed at the data sources of multi-source data, and collection terminals are registered at the corresponding collection points. Based on the collection terminals, basic data for supply chain risk assessment are collected to obtain multi-source data for supply chain risk assessment.

[0049] Storage unit: Classify multi-source data to obtain multiple types of basic data, and store the multiple types of basic data in the database according to the category. The database includes multiple sub-databases, and each sub-database corresponds to one type of basic data.

[0050] Establishing a unit: Binding the sub-database to the collection terminal of the corresponding category of basic data;

[0051] Specifically, multi-source data can include internal supply chain data (such as operational and financial data), external data (such as market and policy data), supplier data (such as supplier performance and financial status), and logistics data (such as transportation status and warehousing data). Collection terminals are deployed according to the source of the basic data, and various types of basic data are collected through these terminals. Preprocessing of the basic data includes normalization; for example, if different systems use different names for the same data, a mapping table can be established to convert aliases into standard field names. Normalization processing includes at least processing of naming, units, format, and value range. This is achieved by analyzing the field naming, units, and values ​​of each data source. To address format differences, corresponding mapping rules were established and batch conversions were performed to ensure consistency across multiple data sources, facilitating subsequent data use. After normalization, the processed multi-source data was categorized, with each type corresponding to a separate sub-database for storage. This facilitates the use of risk impact factors for nodes when constructing the knowledge graph. Binding the data collection end to the corresponding sub-database and establishing a communication channel between them allows for updates to the basic data in the sub-database when the data collected by the end changes. The activation of the communication channel determines whether basic data updates have occurred, thereby indicating whether nodes in the knowledge graph have changed.

[0052] The data fusion module retrieves the categorized basic data and corresponding sub-databases from the database; extracts the features of the basic data as risk impact factors; obtains the correlation relationships as the correlation relationships between factors; sets knowledge graph nodes and edges; and constructs a knowledge graph based on the nodes and edges.

[0053] The data fusion module includes:

[0054] The first acquisition unit: retrieves and extracts the categorized basic data and the corresponding sub-databases from the database;

[0055] The data fusion unit extracts the data features of one type of basic data and its corresponding sub-database as a risk impact factor, and obtains multiple risk impact factors based on multiple types of basic data; it also obtains the correlation between each type of basic data and uses the correlation between each type of basic data as the correlation between factors.

[0056] The knowledge graph construction unit sets multiple nodes for multiple risk impact factors and sets multiple edges for the knowledge graph based on the relationships between factors. The knowledge graph is constructed based on multiple nodes and multiple edges.

[0057] Specifically, the preprocessed multi-source data is categorized, with each category corresponding to a separate sub-database for storage (e.g., operational data sub-database, financial data sub-database, etc.). The data acquisition terminal is bound to the corresponding sub-database, and a communication channel is established between them for subsequent data updates. The categorized basic data and its corresponding sub-databases are retrieved and extracted from the database. Data features of one type of basic data (e.g., "order volume" feature in operational data) and its corresponding sub-database (e.g., operational data sub-database) are extracted as a risk impact factor. Multiple risk impact factors (e.g., "order volume," "sales revenue," "inventory turnover rate," etc.) are obtained based on multiple types of basic data. The correlation between each type of basic data is analyzed (e.g., the positive correlation between "order volume" and "sales revenue"). The correlation between each type of basic data is used as the correlation between factors. Multiple nodes in a knowledge graph are set for multiple risk impact factors (e.g., "high order volume," "sales revenue growth," "low inventory turnover rate," etc.). Multiple edges in the knowledge graph are set for the correlation between factors (e.g., a positive correlation edge between "high order volume" and "sales revenue growth"). A knowledge graph is constructed based on multiple nodes and edges to display different risk factors and their relationships.

[0058] Specifically, for example, suppose there is a supply chain management system that needs to construct the following knowledge graph: Nodes: Order Quantity_High, Order Quantity_Low, Sales Revenue_Increase, Sales Revenue_Decrease, Inventory Turnover_High, Inventory Turnover_Low, Supplier Performance_Excellent, Supplier Performance_Poor; Edges: Order Quantity_High—Sales Revenue_Increase (indicating that when order quantity is high, sales revenue tends to increase), Inventory Turnover_Low—Order Quantity_Low (indicating that when inventory turnover is low, order quantity tends to be low as well), Supplier Performance_Excellent—Order Quantity_High (indicating that when supplier performance is excellent, order quantity tends to be high); Construct a graph containing the above nodes and edges to show different risk influencing factors and their relationships. Through the knowledge graph, potential risk points in the supply chain can be intuitively identified (e.g., "Inventory Turnover_Low" may lead to "Order Quantity_Low"). When the data collected by the acquisition end changes, the basic data in the sub-database can be automatically updated, and the knowledge graph nodes can be judged through the communication channel to achieve real-time monitoring and early warning.

[0059] The model building module, centered on knowledge graph nodes, defines associated regions including primary and secondary associated regions; sets up a virtual boundary layer; establishes a communication channel between the data acquisition end and nodes; establishes a driving relationship between sensing points and the risk assessment model; determines change nodes; dynamically adjusts associated regions; evaluates associated regions; and trains the evaluation model based on the comprehensive results.

[0060] The model building module includes:

[0061] Region determination unit: Taking each node in the knowledge graph as the center point, the node as the center point is taken as the target node; all nodes and edges directly associated with the target node are defined as the associated region of the target node, where the associated region includes the primary associated region and the secondary associated region;

[0062] Boundary Construction Unit: A virtual boundary layer with sensing points is set in the knowledge graph, where each sensing point corresponds to a node; a communication channel is established between the data acquisition terminal and the corresponding node, where each communication channel corresponds to a sensing point; and a driving relationship is established between the sensing points and the risk assessment model.

[0063] Regional adjustment unit: Identify the nodes that have changed to obtain the changed nodes, and dynamically adjust the associated regions based on the changed nodes;

[0064] Model building unit: Performs separate risk assessment for each associated region, identifies and eliminates associated regions that are repeatedly calculated, integrates the assessment results of each associated region to obtain the risk assessment result corresponding to the knowledge graph, and trains the assessment model based on the risk assessment results corresponding to multiple historical knowledge graphs;

[0065] It should be noted that a separate risk assessment is performed on each associated region, identifying and eliminating duplicate associated regions. The assessment results of each associated region are then combined to obtain the specific content of the risk assessment result corresponding to the knowledge graph. The assessment results of each associated region are then integrated to obtain the risk assessment result for the entire knowledge graph. The integration method can be a simple averaging, weighted averaging, or a more complex fusion algorithm, depending on the assessment objectives and data characteristics. Based on the integrated assessment results, a first assessment model is constructed. The model can be a mathematical formula, a decision tree, a neural network, or other forms of model, depending on the complexity and requirements of the assessment task. For example, suppose there is a supply chain risk knowledge graph containing the following nodes and edges: Nodes: A (supplier performance), B (order volume), C (inventory turnover rate), D (sales revenue); Edges: AB (supplier performance affects order volume), BC (order volume affects inventory turnover rate), CD (inventory turnover rate affects sales revenue); Related regions centered on node A: {A,B,AB}; Related regions centered on node B: {B,A,C,AB,BC} (note the overlap with the related regions of A); Related regions centered on node C: {C,B,D,BC,CD}; Related regions centered on node D: {D,C,CD}; A machine learning algorithm is used to assess the risk of each related region, resulting in an assessment score for each region. When integrating the assessment results, duplicate calculated related regions are removed (e.g., edge AB is calculated once in the related regions of A and B). The assessment scores of each related region are then weighted and averaged to obtain the risk assessment result for the entire knowledge graph. Based on the integrated assessment results, a neural network model is constructed as the primary assessment model. This model can take new data as input and output a risk assessment score, enabling real-time early warning and monitoring of supply chain risks.

[0066] The specific content of training an assessment model based on risk assessment results from multiple historical knowledge graphs includes: definition and extraction of associated regions: knowledge graphs. ,in It is a set of nodes. It is a set of edges; setting the association depth k (e.g., first-order neighbors); set of associated regions. ,in Based on nodes A subgraph centered on the node; Its associated region Defined as: ,in, It represents a specific node in the knowledge graph and is the central node that defines the associated regions. Represented by node The central associated area, this area contains the related Distance not exceeding All nodes and the edges between these nodes. Represents any node in the knowledge graph, used to connect with the central node. Compare and determine whether it belongs to The associated region. : Represents a node and The shortest path length between two nodes is used to measure the distance between them. This represents the association depth, a preset threshold used to determine which nodes belong to the association. The associated region. Only with The shortest path length does not exceed Only nodes that meet the above criteria will be included. middle. It represents the set of all nodes in a knowledge graph. Indicates the connection node and If the edge, and All belong to Then this edge will also be included. middle. Let represent the set of all edges in the knowledge graph; the formula means that for any node in the knowledge graph... Its associated region It is composed of all and The shortest path length does not exceed The subgraph consisting of the nodes and the edges between these nodes;

[0067] Assess the risk of associated regions separately: Input associated regions eigenvectors (Includes node attributes, edge weights, etc.). Output risk score. . formula: ,in It is a risk assessment function (such as a machine learning model). These are feature weights (which can be determined through entropy weighting or expert weighting).

[0068] Deduplication: Related Region Sets The set after deduplication .

[0069] Structural deduplication: Compare subgraph isomorphisms (e.g., using the VF2 algorithm) and remove structurally similar related regions. Node deduplication: Merge duplicate nodes (e.g., deduplicate by node ID). Formula (Example: based on node set) Similarity threshold): For any two related regions and ,calculate ;

[0070] like Then remove one of the related regions. This represents the set of associated regions, which is the initial set before the deduplication operation and contains multiple associated regions. This represents the set of related regions after deduplication, obtained through a series of deduplication operations (such as structural deduplication, node deduplication, etc.). Obtained from [the source]. Represents the set of related regions The first in A region is a specific subgraph that contains all nodes and edges directly associated with a central point. Represents the set of related regions The first in Each related region is also a specific subgraph, and... Similar, but may have different nodes and edges. : Indicates the associated region The set of nodes in, i.e. The set of all nodes contained therein. Indicates the associated region The set of nodes in, i.e. The set of all nodes contained therein. Indicates the associated region and Between Similarity is a metric used to measure the similarity between two sets. Here, it is used to measure the similarity between the sets of nodes of two related regions (i.e., two subgraphs). express The similarity threshold is a preset similarity threshold used to determine whether two related regions are so similar that one of them needs to be removed. If Then it is believed and If the similarity is too high, one of them needs to be removed to avoid duplicate calculations. The entire formula means: for a set of related regions... Any two related regions and Calculate the Jaccard similarity between them. If the similarity exceeds a preset similarity threshold... If the two related regions are considered to have too high similarity, one of them needs to be removed to obtain a deduplicated set of related regions. ;

[0071] Overall assessment results: Set of risk scores after deduplication Output global risk score Formula (weighted average method): ,in It is a related region The weights (which can be determined based on node degree, centrality, etc.) This represents the set of risk scores after deduplication, that is, the set of risk scores remaining after removing risk scores from related regions that have been counted repeatedly or have too high similarity. Represents the set of risk scores after deduplication. The first in Each risk score corresponds to a risk assessment result for a specific associated region (after removing duplicates). Represents the set of risk scores after deduplication. The total number of medium-risk scores, i.e. the number of associated regions (after removing duplicates). The global risk score is an overall risk assessment result obtained by combining the risk scores of each related region after deduplication through a certain fusion method (such as weighted average). This indicates that in the set of deduplicated risk scores, the i-th Each associated region (corresponding to) The weight of the associated region is determined based on various factors, such as the node degree of the associated region (i.e., the number of connections between nodes within the associated region) and centrality (i.e., the importance of the associated region in the overall knowledge graph). The weight is used to adjust the influence of each associated region on the global risk score when synthesizing the risk scores of various associated regions. The entire formula means: Global Risk Score It is obtained by combining the risk scores of each deduplicated related region using a weighted average method. The risk score for each related region... Multiply by its corresponding weight Then, add all the products together to get the global risk score. This method allows us to adjust the impact of each associated region on the global risk score based on its importance, collecting risk assessment results from multiple historical knowledge graphs, including the risk score for each associated region and the global risk score. A suitable machine learning model (such as a neural network or decision tree) is selected as the evaluation model. The evaluation model is trained using historical data, and the model parameters are adjusted to minimize prediction error. The model performance is evaluated using a validation set to ensure its accuracy and generalization ability. An effective evaluation model can be trained based on the risk assessment results from multiple historical knowledge graphs, enabling risk assessment of new knowledge graphs.

[0072] Specifically, the associated region here refers to extracting other nodes related to the central node, taking the node as the center point, and using the other nodes related to the central node and the region formed by the node as the associated region of the corresponding central node. Each node is treated as a central point, and the risk of the associated region corresponding to the node is calculated separately. Finally, the risks of multiple associated regions are comprehensively calculated. In the comprehensive calculation, it is necessary to remove the associated regions that are calculated repeatedly, so as to complete the risk assessment of the knowledge graph. This can facilitate the subsequent direct extraction of the risk assessment of the associated regions of secondary parts, reduce the complexity of the overall assessment calculation, and thus improve the efficiency of supply chain risk assessment.

[0073] The steps of defining all nodes and edges directly associated with the target node as the associated region of the target node, where the associated region includes primary associated regions and secondary associated regions, include: setting impact thresholds for corresponding risk impact factors, designating risk impact factors exceeding the impact thresholds as core risk factors; designating risk impact factors not exceeding the impact thresholds as auxiliary risk factors; designating nodes corresponding to core risk factors as primary nodes; and obtaining the associated regions corresponding to multiple primary and secondary nodes to obtain multiple primary and secondary associated regions before adjustment.

[0074] The steps for identifying changed nodes and dynamically adjusting the primary and secondary related regions based on these changed nodes include: identifying active nodes based on sensor data collection of changed nodes, and recording the time interval between two consecutive changes for each active node; for each active node, determining whether the latest change interval exceeds a preset time threshold; marking the active node corresponding to the latest change interval exceeding the preset time threshold as a changed node, and using its last change data as the base data for that node; using the related region corresponding to the changed node as the adjusted primary related region; and using nodes that have not changed as non-changed nodes, and using the related region corresponding to non-changed nodes as the adjusted secondary related region.

[0075] Specifically, active nodes are nodes that have recently changed. Change interval duration is the time difference between two consecutive changes. Preset duration threshold is a user-defined time criterion (e.g., 12 days). Changed nodes are active nodes whose change interval exceeds the threshold. Representative data is a snapshot of the data at the time of the last change of the changed node. By monitoring the change interval and comparing it to the threshold, nodes that have not been updated for a long time or are abnormally active (i.e., changed nodes) can be quickly located, avoiding indiscriminate processing of all nodes and improving analysis efficiency. Using the data at the time of the last change as representative data ensures that the latest node status is used, avoiding decision-making bias due to outdated data. The preset duration threshold can be flexibly adjusted according to business needs (e.g., in supply chain risk monitoring, the threshold can be set to "supplier delivery cycle + safety buffer period") to adapt to different scenarios.

[0076] Specifically, nodes correspond to sub-databases, and sensing points correspond to nodes. Therefore, when the basic data collected at the acquisition end is transmitted to the corresponding sub-database through the communication channel, it is equivalent to transmitting it to the corresponding node. When the acquisition end collects changed basic data, it transmits it to the corresponding node. At this time, the sensing point will detect that new basic data has been transmitted in the communication channel, indicating that the node on the knowledge graph has changed. Therefore, the evaluation result will also change, and a re-risk assessment of the knowledge graph is required. There are two types of re-risk assessment. One is to divide the risk impact factors corresponding to the node into core risk factors and auxiliary risk factors, with the core risk factors as the main part and the auxiliary risk factors as the secondary part. The two parts are evaluated separately according to the first assessment model and then combined. The other is to adjust the main and secondary parts based on the first assessment model when subsequent nodes change, and evaluate according to the adjusted main and secondary parts. Before conducting supply chain risk assessment, the frequency of node changes is determined first. When the frequency of change is within the time threshold, the risk assessment model is not activated. When it exceeds the time threshold, the assessment is conducted. For example, a re-assessment is conducted after a certain period of no change.

[0077] The risk assessment module acquires a knowledge graph formed by fusing data from multiple sources; identifies relevant regions; and conducts risk assessments of the supply chain based on the risk assessment model to obtain the risk assessment results.

[0078] The risk assessment module includes:

[0079] The second acquisition unit: retrieves multi-source data from the database for supply chain risk assessment and integrates it into a knowledge graph;

[0080] Risk assessment unit: Identify the related regions in the knowledge graph, and conduct risk assessment based on the corresponding related regions of the assessment model to obtain the risk assessment result. The related regions in the knowledge graph include the related regions before adjustment and the related regions after adjustment.

[0081] Specifically, the related regions in the knowledge graph include related regions before adjustment and related regions after adjustment. The related regions before adjustment refer to the primary and secondary related regions divided according to the initial core risk factors and auxiliary risk factors. The related regions after adjustment refer to the regions where the nodes corresponding to the change nodes that have not exceeded the preset conditions are taken as the center points, and the positions of the primary related regions are adjusted using the center points as the core risk factors.

[0082] The risk assessment unit includes:

[0083] Comparison Subunit: Used to count the number of changed nodes, calculate the proportion of changed nodes in the knowledge graph based on the number of changed nodes, and compare the changed node proportion with a preset proportion threshold;

[0084] Initial assessment subunit: The number of change nodes corresponding to the proportion of change nodes exceeding the preset ratio threshold is taken as the number of abnormal change nodes. Among them, the change nodes corresponding to the number of abnormal change nodes are recorded as abnormal change nodes. For abnormal change nodes, risk assessment is carried out based on the primary related area and the secondary related area before the adjustment.

[0085] Secondary assessment sub-unit: The number of change nodes corresponding to the proportion of change nodes that does not exceed the preset ratio threshold is taken as the number of normal change nodes. Among them, the change nodes corresponding to the number of normal change nodes are recorded as normal change nodes. For normal change nodes, risk assessment is carried out based on the adjusted main related area and the original secondary related area.

[0086] Specifically, when the percentage of changed nodes exceeds a preset threshold, a global reassessment is required. This involves using the original primary and secondary related regions, and employing the data from the node changes (the latest data collected by the data acquisition end) as the basis for each node's data, thus completing a global reassessment of the entire knowledge graph. Conversely, when the percentage of changed nodes does not exceed a preset threshold, a separate risk assessment is necessary. This uses the adjusted primary and secondary related regions. For example, if the knowledge graph has 10 nodes, and the data of one changed node has changed, but the percentage does not exceed the preset threshold, the adjusted primary and original secondary related regions are used. The primary related region then becomes the region centered on the changed node, while the related regions of other nodes become secondary. Therefore, only the primary related regions need to be reassessed, while the secondary related regions can use the initial assessment results. Combining the initial assessment results with the results of the adjusted primary related region assessment yields a new risk assessment after the node data changes, improving the efficiency of supply chain risk assessment.

[0087] This invention, by establishing a binding relationship between the data collection terminal and various nodes within the database, can comprehensively and accurately collect multi-source data related to supply chain risk assessment. It integrates this multi-source data, intuitively presenting risk impact factors and their interrelationships. Dynamically adjusting the associated regions based on changing nodes allows for real-time response to dynamic changes in the supply chain, timely adjustments to assessment strategies, and improved accuracy and timeliness of assessment results. In defining the associated regions, impact thresholds are set for each risk impact factor. Risk impact factors exceeding the threshold are designated as core risk factors, while those within the threshold are designated as auxiliary risk factors, highlighting key risks and making risk assessment more targeted, thus improving efficiency and accuracy. Assessments are conducted based on the associated regions before and after adjustment, ensuring the comprehensiveness and accuracy of the assessment results. It can integrate multiple data sources, including internal supply chain data, external data, supplier data, and logistics data, to conduct a comprehensive and accurate assessment of supply chain risks through data fusion.

[0088] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A supply chain risk assessment system based on multi-source data fusion, characterized in that, include: The data acquisition module configures multi-source data acquisition terminals and establishes binding relationships between the acquisition terminals and various nodes within the database. The database is used to store multi-source data related to supply chain risk assessment. The multi-source data is classified and stored in a database composed of multiple sub-databases according to their categories. The sub-databases are then bound to the acquisition terminals of the corresponding categories of basic data. The data fusion module retrieves the categorized basic data and corresponding sub-databases from the database; extracts the features of the basic data as risk impact factors; obtains the correlation relationships as the correlation relationships between factors; sets knowledge graph nodes and edges; and constructs a knowledge graph based on the nodes and edges. The model building module, centered on knowledge graph nodes, defines associated regions including primary and secondary associated regions; sets up a virtual boundary layer; establishes a communication channel between the data acquisition end and nodes; establishes a driving relationship between sensing points and the risk assessment model; determines change nodes; dynamically adjusts associated regions; evaluates associated regions; and trains the evaluation model based on the comprehensive results. The risk assessment module acquires a knowledge graph formed by fusing data from multiple sources; identifies relevant regions; and conducts risk assessments of the supply chain based on the risk assessment model to obtain the risk assessment results. The model building module includes: Region determination unit: Taking each node in the knowledge graph as the center point, the node as the center point is taken as the target node; all nodes and edges directly associated with the target node are defined as the associated region of the target node, where the associated region includes the primary associated region and the secondary associated region; Boundary Construction Unit: A virtual boundary layer with sensing points is set in the knowledge graph, where each sensing point corresponds to a node; a communication channel is established between the data acquisition terminal and the corresponding node, where each communication channel corresponds to a sensing point; and a driving relationship is established between the sensing points and the risk assessment model. Regional adjustment unit: Identify the nodes that have changed to obtain the changed nodes, and dynamically adjust the associated regions based on the changed nodes; Model building unit: Performs separate risk assessment for each associated region, identifies and eliminates associated regions that are repeatedly calculated, integrates the assessment results of each associated region to obtain the risk assessment result corresponding to the knowledge graph, and trains the assessment model based on the risk assessment results corresponding to multiple historical knowledge graphs; The steps of defining all nodes and edges directly associated with the target node as the associated region of the target node, where the associated region includes primary and secondary associated regions, are as follows: Set impact thresholds for each corresponding risk factor; Risk factors that exceed the impact threshold will be considered as core risk factors. Risk factors that do not exceed the impact threshold are used as auxiliary risk factors; nodes corresponding to core risk factors are used as primary nodes; nodes corresponding to auxiliary risk factors are used as secondary nodes; the associated regions corresponding to multiple primary and secondary nodes are obtained to obtain multiple primary associated regions and secondary associated regions before adjustment. The steps for identifying changed nodes and dynamically adjusting the primary and secondary related regions based on these changed nodes include: Active nodes are identified by collecting data from sensing points that show changes, and the time interval between two consecutive changes of each active node is recorded. For each active node, determine whether the interval between the latest changes exceeds a preset time threshold; The active nodes corresponding to the latest change interval that exceeds the preset time threshold are marked as changed nodes, and the data at the time of the last change is taken as the basic data corresponding to the node; the associated region corresponding to the changed node is taken as the adjusted main associated region. Nodes that have not changed are considered as unchanged nodes, and the associated regions corresponding to unchanged nodes are considered as adjusted secondary associated regions. The risk assessment module includes: The second acquisition unit: retrieves multi-source data from the database for supply chain risk assessment and integrates it into a knowledge graph; Risk assessment unit: Identify the related regions in the knowledge graph, conduct risk assessment based on the corresponding related regions of the assessment model to obtain the risk assessment result. The related regions in the knowledge graph include the related regions before adjustment and the related regions after adjustment. The risk assessment unit includes: Comparison Subunit: Used to count the number of changed nodes, calculate the proportion of changed nodes in the knowledge graph based on the number of changed nodes, and compare the changed node proportion with a preset proportion threshold; Initial assessment subunit: The number of change nodes corresponding to the proportion of change nodes exceeding the preset ratio threshold is taken as the number of abnormal change nodes. Among them, the change nodes corresponding to the number of abnormal change nodes are recorded as abnormal change nodes. For abnormal change nodes, risk assessment is carried out based on the primary related area and the secondary related area before the adjustment. Secondary assessment sub-unit: The number of change nodes corresponding to the proportion of change nodes that does not exceed the preset ratio threshold is taken as the number of normal change nodes. Among them, the change nodes corresponding to the number of normal change nodes are recorded as normal change nodes. For normal change nodes, risk assessment is carried out based on the adjusted main related area and the original secondary related area.

2. The supply chain risk assessment system based on multi-source data fusion according to claim 1, characterized in that: The data acquisition module includes: Collection Unit: Collection points are deployed at the data sources of multi-source data, and collection terminals are registered at the corresponding collection points. Based on the collection terminals, basic data for supply chain risk assessment are collected to obtain multi-source data for supply chain risk assessment. Storage unit: Classify multi-source data to obtain multiple types of basic data, and store the multiple types of basic data in the database according to the category. The database includes multiple sub-databases, and each sub-database corresponds to one type of basic data. Establishing a unit: Binding the sub-database to the collection terminal of the basic data of the corresponding category.

3. The supply chain risk assessment system based on multi-source data fusion according to claim 1, characterized in that: The data fusion module includes: The first acquisition unit: retrieves and extracts the categorized basic data and the corresponding sub-databases from the database; The data fusion unit extracts the data features of one type of basic data and its corresponding sub-database as a risk impact factor, and obtains multiple risk impact factors based on multiple types of basic data; it also obtains the correlation between each type of basic data and uses the correlation between each type of basic data as the correlation between factors. The knowledge graph construction unit sets multiple nodes for multiple risk impact factors and sets multiple edges for the relationships between factors, and constructs the knowledge graph based on multiple nodes and multiple edges.

Images