Business data processing method and system based on supply chain management

By collecting and integrating node degree, cargo flow and inventory levels in supply chain management, grid hierarchy division and importance score are carried out, combined with dual-deep Q network modeling, and generating dynamic scheduling strategies, the supply chain management in traditional methods is solved The problem of insufficient accuracy and flexibility improves the accuracy of data processing and the efficiency of resource allocation.

CN118428775BInactive Publication Date: 2024-11-05共幸科技(深圳)有限公司
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202410879329.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-02
Publication Date
2024-11-05
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional supply chain management methods are difficult to deal with market fluctuations, logistics interruptions and inventory management in real time, and lack effective hierarchical division and comprehensive scoring mechanisms, resulting in insufficient accuracy and reliability of node priority ranking and decision support, and lack of intelligent data processing. Capabilities and dynamic scheduling strategies.

Method used

By collecting and integrating the node degree, cargo flow and inventory levels in the supply chain network, performing grid-level division, compute comprehensive node importance scores, conducting node sorting and network index changes analysis, and using dual-deep Q network for multi-layer network perception Model and generate dynamic supply chain scheduling strategies.

Benefits of technology

It improves the accuracy of business data processing in supply chain management, enhances the flexibility and resilience of the supply chain, and optimizes resource allocation and scheduling decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118428775B_ABST
    Figure CN118428775B_ABST
Patent Text Reader

Abstract

The present application relates to the field of data processing technology, and discloses a business data processing method and system based on supply chain management. The method includes: collecting and integrating the node degree, cargo flow and inventory level of multiple nodes in the supply chain network, obtaining a node feature data set and performing grid level division to obtain a hierarchical structure diagram; performing node feature processing to obtain a network feature description and a level identification; performing importance comprehensive score calculation to obtain a node importance list; performing node removal simulation and network indicator change analysis to obtain network indicator change data; performing fuzzy hierarchical analysis to obtain an indicator selection evaluation result and using a double-depth Q network modeling to obtain a multi-layer network perception model; obtaining a real-time status data set and performing supply chain scheduling strategy analysis through a multi-layer network perception model, and outputting a dynamic supply chain scheduling strategy. The present application improves the accuracy of business data processing for supply chain management.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a business data processing method and system based on supply chain management. Background Art

[0002] In modern supply chain management, with the development of global trade and the continuous changes in market demand, the supply chain network has become increasingly complex and dynamic. This complexity and dynamism requires that the supply chain management method not only needs to efficiently process massive business data, but also needs to be highly flexible and responsive. However, traditional supply chain management methods usually rely on static data analysis and decision-making models, and cannot respond to various emergencies and changes in the supply chain network in real time. This has led to significant deficiencies in the supply chain in dealing with market fluctuations, logistics disruptions, and inventory management.

[0003] Current supply chain management systems often lack effective hierarchical division and comprehensive scoring mechanisms when processing and integrating business data from multiple nodes, making it difficult to accurately evaluate and optimize the importance of each node in the entire network. Traditional methods rely on a single indicator or simple statistical analysis, ignoring the complex relationships and dynamic changes between nodes, resulting in insufficient accuracy and reliability in node prioritization and decision support. In addition, existing supply chain management tools usually lack intelligent data processing capabilities and dynamic scheduling strategies, making it difficult for supply chains to achieve optimized management and efficient resource allocation. Summary of the invention

[0004] The present application provides a business data processing method and system based on supply chain management, which are used to improve the accuracy of business data processing based on supply chain management.

[0005] In a first aspect, the present application provides a business data processing method based on supply chain management, and the business data processing method based on supply chain management includes:

[0006] Collecting and integrating node degrees, cargo flows and inventory levels of multiple nodes in a supply chain network to obtain a node feature data set, and dividing the supply chain network into grid levels according to the node feature data set to obtain a hierarchical structure diagram;

[0007] Perform node feature processing on the multiple nodes according to the hierarchical structure diagram to obtain a network feature description and a hierarchical identifier of each node;

[0008] Calculating the importance comprehensive scores of the multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and sorting the multiple nodes according to the importance comprehensive score to obtain a node importance list;

[0009] Based on the node importance list, the supply chain network is simulated for node removal and analyzed for network indicator changes to obtain network indicator change data;

[0010] Performing fuzzy hierarchical analysis on the network indicator change data to obtain indicator selection evaluation results, and using a double-depth Q network to perform multi-layer network perception modeling on the indicator selection evaluation results to obtain a multi-layer network perception model;

[0011] A real-time status data set of the supply chain network is obtained, and the real-time status data set is input into the multi-layer network perception model to perform supply chain scheduling strategy analysis, and a dynamic supply chain scheduling strategy is output.

[0012] In a second aspect, the present application provides a business data processing system based on supply chain management, and the business data processing system based on supply chain management includes:

[0013] A collection module, used to collect and integrate the node degrees, cargo flows and inventory levels of multiple nodes in the supply chain network to obtain a node feature data set, and to divide the supply chain network into grid levels according to the node feature data set to obtain a hierarchical structure diagram;

[0014] A processing module, used to perform node feature processing on the multiple nodes according to the hierarchical structure diagram to obtain a network feature description and a hierarchical identifier of each node;

[0015] A calculation module, configured to calculate the importance comprehensive scores of the multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and to sort the multiple nodes according to the importance comprehensive score to obtain a node importance list;

[0016] An analysis module, used to perform node removal simulation and network indicator change analysis on the supply chain network based on the node importance list to obtain network indicator change data;

[0017] A modeling module is used to perform fuzzy hierarchical analysis on the network indicator change data to obtain an indicator selection evaluation result, and use a double-depth Q network to perform multi-layer network perception modeling on the indicator selection evaluation result to obtain a multi-layer network perception model;

[0018] The output module is used to obtain the real-time status data set of the supply chain network, input the real-time status data set into the multi-layer network perception model to perform supply chain scheduling strategy analysis, and output a dynamic supply chain scheduling strategy.

[0019] The third aspect of the present application provides a business data processing device based on supply chain management, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the business data processing device based on supply chain management executes the above-mentioned business data processing method based on supply chain management.

[0020] A fourth aspect of the present application provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned business data processing method based on supply chain management.

[0021] In the technical solution provided by the present application, a node feature data set is formed by collecting and integrating the node degree, cargo flow and inventory level of multiple nodes in the supply chain network. By dividing the node feature data set into grid levels, a hierarchical structure diagram is generated, and node feature processing is performed on multiple nodes to obtain a network feature description and a hierarchy identifier. According to the network feature description and hierarchy identifier, a comprehensive importance score calculation is performed on multiple nodes, and a node importance list is generated. By introducing a variety of evaluation indicators, the importance of the nodes is comprehensively evaluated. Based on the node importance list, the supply chain network is simulated for node removal and the network indicator change analysis is performed to generate network indicator change data, and the impact of node removal on the supply chain network is pre-evaluated to help discover key nodes and potential weak links, and optimize the supply chain structure and resource allocation. By performing fuzzy hierarchical analysis on the network indicator change data, the indicator selection evaluation results are obtained, and a dual-depth Q network is used for multi-layer network perception modeling to generate a multi-layer network perception model, which can more accurately capture the dynamic change characteristics of the supply chain network and improve the perception ability and prediction accuracy of the supply chain status. By obtaining the real-time status data set of the supply chain network and inputting it into the multi-layer network perception model for supply chain scheduling strategy analysis, a dynamic supply chain scheduling strategy is output. It can quickly respond to changes in the supply chain network, optimize scheduling decisions, improve the flexibility and adaptability of the supply chain, and thus improve the accuracy of business data processing in supply chain management. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0023] Figure 1 This is a schematic diagram of an embodiment of a business data processing method based on supply chain management in an embodiment of the present application;

[0024] Figure 2 This is a schematic diagram of an embodiment of a business data processing system based on supply chain management in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The embodiment of the present application provides a business data processing method and system based on supply chain management. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the business data processing method based on supply chain management includes:

[0027] Step S101, collecting and integrating the node degrees, cargo flows and inventory levels of multiple nodes in the supply chain network to obtain a node feature data set, and dividing the supply chain network into grid levels according to the node feature data set to obtain a hierarchical structure diagram;

[0028] It is understandable that the execution subject of the present application may be a business data processing system based on supply chain management, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0029] Specifically, a network topology model is established. By modeling the connectivity of multiple nodes in the supply chain network, the network topology structure is obtained. On this basis, the node degree of each node is weighted and the weighted node degree value set is obtained considering the importance and influence of the node, which helps to determine the relative importance of each node in the supply chain network. The cargo flow of multiple nodes is analyzed by time series analysis method, and a cargo flow time series model is established. By extracting the periodic changes of time series data, periodic cargo flow patterns can be identified, which reflect the cargo flow laws of nodes in different time periods. The inventory levels of multiple nodes are analyzed by cluster analysis method to form an inventory level clustering model. According to the model, the inventory level is hierarchically divided to obtain a hierarchical inventory level set. The weighted node degree value set, the periodic cargo flow pattern and the hierarchical inventory level set are fused to form a comprehensive node feature data set. The data set includes the connectivity and influence information of the node, as well as the cargo flow laws and inventory level characteristics of the node. According to the comprehensive node feature data set, the entire supply chain network is divided into grid levels. The network is divided into multiple levels, each level represents different node characteristics and attributes. The grid level division results are displayed through visualization means to form a hierarchical structure.

[0030] Step S102: performing node feature processing on multiple nodes according to the hierarchical structure diagram to obtain a network feature description and a hierarchical identifier of each node;

[0031] Specifically, the hierarchical structure graph is segmented to obtain a set of segmented hierarchical structure subgraphs. Boundary features are extracted for each hierarchical structure subgraph to form a boundary feature set. These boundary features can help identify the feature boundaries of subgraphs at different levels. The boundary feature set is classified to obtain a hierarchical identification set, which is used to annotate node features at different levels. The hierarchical structure graph is converted into a weighted directed graph model, which can better reflect the relationship and weight between nodes. Laplace matrix feature extraction is performed on the weighted directed graph model to obtain a Laplace feature matrix. The Laplace feature matrix can help the system understand the overall and local characteristics of the graph structure. The Laplace feature matrix is ​​spectrally decomposed to obtain a feature spectrum set, and the intrinsic structure of the weighted directed graph model and the relationship between nodes are understood. Spectral clustering is performed on the weighted directed graph model to obtain a node clustering result. Spectral clustering is an effective graph clustering method. By decomposing the spectrum of the graph, nodes are grouped according to their characteristics. Density peak clustering is performed on the node clustering results to obtain a density peak clustering model. Density peak clustering is a density-based clustering method that can effectively identify high-density areas in the data set and use these areas as the center of the cluster. According to the density peak clustering model, the center vector of each node is extracted to obtain a set of node center vectors. The node center vector can represent the centrality and importance of each node in the network. The node center vector set and the level identification set are merged to obtain a network feature description set for each node. Based on the network feature description set and the level identification set, the network feature description and level identification of each node are generated.

[0032] Step S103: Calculate the importance comprehensive scores of multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and sort the multiple nodes according to the importance comprehensive score to obtain a node importance list;

[0033] Specifically, the network feature description is standardized to obtain a standardized network feature vector, eliminating the influence of different feature dimensions and magnitudes, so that different features can be compared on the same scale. According to the standardized network feature vector, a node importance evaluation hierarchical model is constructed, in which the level identifier is introduced as one of the evaluation criteria. The elements of each level of the node importance evaluation hierarchical model are compared pairwise to construct a judgment matrix. The judgment matrix is ​​checked for consistency to ensure the logical consistency of the judgment matrix and avoid deviations caused by the subjectivity of human judgment. After the consistency test, the weight vector of each evaluation criterion is calculated according to the judgment matrix, and the weight vector reflects the relative importance of each criterion in the node importance evaluation. The initial importance score of the node is obtained by weighted summation of the weight vector, and the preliminary importance score of each node is obtained by comprehensively considering the influence of each evaluation criterion on the importance of the node. The initial importance score of the node is analyzed by gray correlation to obtain a gray correlation coefficient matrix. Gray correlation analysis is an analysis method based on the similarity of data sequences. By calculating the correlation between the initial importance score of the node and the ideal importance score, the importance of the node can be better reflected. According to the grey correlation coefficient matrix, the initial importance scores of the nodes are corrected to obtain the comprehensive importance scores of the nodes. The initial importance scores are adjusted according to the grey correlation coefficient to more accurately reflect the actual importance of each node. The comprehensive importance scores of the nodes are sorted to obtain a list of node importance, which effectively evaluates the importance of each node in the supply chain network.

[0034] Step S104: Perform node removal simulation and network indicator change analysis on the supply chain network based on the node importance list to obtain network indicator change data;

[0035] Specifically, according to the node importance list, the node set with the lowest importance is selected, and a complex network model is established for the supply chain network. According to the complex network model, the initial network topology index set is calculated. The topology indexes include node degree, clustering coefficient, network efficiency and average path length, etc. These indexes can fully reflect the topological structure characteristics of the network. The node set with the lowest importance is gradually removed for simulation, and after each node is removed, the network topology index set is recalculated according to the complex network model to observe the impact of each node removal on the overall topological structure of the network. By comparing the initial network topology index set with the recalculated network topology index set, a topology index change matrix is ​​obtained, which reflects the changes of each topological index before and after the node is removed. According to the topology index change matrix, the changes of key topological indicators are extracted. The key indicators may include network connectivity, the shortest path change between nodes and network robustness, etc., and an index change model is established for the changes of key topological indicators. By modeling the changes of key indicators, the specific impact of node removal on the network structure can be understood. All the index change models are integrated to obtain a comprehensive network index change model. The model integrates the changes of various key topological indicators and can fully reflect the impact of node removal on the overall structure of the supply chain network. Finally, based on the comprehensive network indicator change model, the network indicator change data is output to help the system understand the changes in the network under different node removal situations.

[0036] Step S105: Perform fuzzy hierarchical analysis on the network indicator change data to obtain the indicator selection evaluation results, and use a double-depth Q network to perform multi-layer network perception modeling on the indicator selection evaluation results to obtain a multi-layer network perception model;

[0037] Specifically, the network indicator change data is standardized to obtain a standardized indicator change data set, eliminating the dimensional differences between different indicators. Based on the standardized indicator change data set, a network indicator hierarchical weight matrix is ​​constructed, which is used to describe the relative importance of each network indicator in the entire network evaluation process. According to the standardized indicator change data set and the network indicator hierarchical weight matrix, a fuzzy hierarchical analysis model is established. The fuzzy hierarchical analysis model can effectively deal with the uncertainty and fuzziness in multi-indicator decision-making, and comprehensively evaluate each indicator through fuzzy mathematics methods. Based on the fuzzy hierarchical analysis model, the weight value of each network indicator is calculated to obtain a network indicator weight set, which reflects the importance of each indicator in the network evaluation process. According to the network indicator weight set, the standardized indicator change data set is weighted and fused to obtain weighted indicator change data. The influence of each indicator is combined to make the obtained indicator data more comprehensively reflect the changes in the network. The weighted indicator change data is input into the first layer Q network of the dual deep Q network to obtain the first layer network output. The first layer Q network extracts preliminary network features by deep learning the input data. The first layer network output is input into the second layer Q network to obtain the second layer network output. The output of the second layer network is feature extracted to obtain a set of multi-layer network feature vectors, which can fully reflect the characteristics of the network at different levels. Based on the multi-layer network feature vector set and the dual-depth Q network, a multi-layer network perception model is generated. This model uses a multi-layer deep learning method to fully perceive and analyze the changes in the network, and realize the dynamic evaluation and optimization of the supply chain network.

[0038] Step S106: Acquire a real-time status data set of the supply chain network, input the real-time status data set into a multi-layer network perception model to perform supply chain scheduling strategy analysis, and output a dynamic supply chain scheduling strategy.

[0039] Specifically, real-time node data, real-time cargo flow data and real-time inventory data are collected from the supply chain network. These data can fully reflect the current operating status of the supply chain network. Data cleaning and outlier processing are performed on the real-time node data to ensure the accuracy and consistency of the data, and a purified real-time node data set is obtained. The real-time cargo flow data is decomposed into time series to obtain a decomposed real-time cargo flow component set. The time series decomposition can reveal the periodic and trend changes in the cargo flow data. At the same time, the inventory turnover rate is calculated for the real-time inventory data to obtain a real-time inventory turnover rate data set. The inventory turnover rate is an important indicator for evaluating inventory management efficiency. The purified real-time node data set, the decomposed real-time cargo flow component set and the real-time inventory turnover rate data set are fused to obtain a real-time status data set of the supply chain network. The real-time status data set is input into a multi-layer network perception model to obtain a network perception output vector. The output vector reflects the overall characteristics and potential problems of the supply chain network in the current state. The network perception output vector is decoded to obtain a scheduling strategy encoding, and the network perception output is converted into a specific supply chain scheduling strategy. According to the scheduling strategy coding, the supply chain scheduling objective function is determined, and the supply chain scheduling objective function is constrained to obtain a set of constraints. The objective function and constraints together constitute the mathematical description of the supply chain scheduling problem. The heuristic algorithm is used to perform multi-objective optimization on the scheduling objective function under the constraint set. The heuristic algorithm has a strong solving ability and can find a near-optimal solution in a complex search space. According to the multi-objective optimization solution results, the initial supply chain scheduling strategy is generated, and the initial supply chain scheduling strategy is dynamically adjusted to output a dynamic supply chain scheduling strategy. The dynamic adjustment can timely correct and optimize the scheduling strategy according to the changes in real-time status data to ensure the efficient operation of the supply chain.

[0040] In the embodiment of the present application, a node feature data set is formed by collecting and integrating the node degree, cargo flow and inventory level of multiple nodes in the supply chain network. By dividing the node feature data set into grid levels, generating a hierarchical structure diagram, and performing node feature processing on multiple nodes, a network feature description and a hierarchy identifier are obtained. According to the network feature description and hierarchy identifier, a comprehensive importance score calculation is performed on multiple nodes, and a node importance list is generated. By introducing a variety of evaluation indicators, the importance of the nodes is comprehensively evaluated. Based on the node importance list, the supply chain network is simulated for node removal and the network indicator change analysis is performed to generate network indicator change data, and the impact of node removal on the supply chain network is pre-evaluated to help discover key nodes and potential weak links, and optimize the supply chain structure and resource allocation. By performing fuzzy hierarchical analysis on the network indicator change data, the indicator selection evaluation result is obtained, and a dual-depth Q network is used for multi-layer network perception modeling to generate a multi-layer network perception model, which can more accurately capture the dynamic change characteristics of the supply chain network and improve the perception ability and prediction accuracy of the supply chain status. By obtaining the real-time status data set of the supply chain network and inputting it into the multi-layer network perception model for supply chain scheduling strategy analysis, a dynamic supply chain scheduling strategy is output. It can quickly respond to changes in the supply chain network, optimize scheduling decisions, improve the flexibility and adaptability of the supply chain, and thus improve the accuracy of business data processing in supply chain management.

[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0042] (1) Model the network topology structure of multiple nodes in the supply chain network to obtain a network topology model;

[0043] (2) According to the network topology model, the node degrees of multiple nodes are weighted and a set of weighted node degree values ​​is obtained;

[0044] (3) Perform time series analysis on the freight flow of multiple nodes to obtain a freight flow time series model;

[0045] (4) According to the freight flow time series model, the periodic changes of freight flow are extracted to obtain the periodic freight flow pattern;

[0046] (5) Perform cluster analysis on the inventory levels of multiple nodes to obtain an inventory level clustering model;

[0047] (6) Based on the inventory level clustering model, the inventory level is divided into levels to obtain a set of hierarchical inventory levels;

[0048] (7) The weighted node degree value set, the periodic cargo flow pattern, and the hierarchical inventory level set are integrated to obtain the node feature data set;

[0049] (8) Based on the node feature data set, the supply chain network is divided into grid levels to obtain the grid level division results, and the grid level division results are visualized to obtain a hierarchical structure diagram.

[0050] Specifically, the network topology structure of multiple nodes in the supply chain network is modeled. Assume that a supply chain network contains several nodes (for example, suppliers, manufacturers, distributors, and retailers), which are connected to each other through edges (i.e., the paths of goods flow). A graph in graph theory is used to represent this network, where nodes represent entities in the supply chain and edges represent the connection relationships between them. Mathematically, Represents this graph, where Represents a collection of nodes. represents a set of edges. For example, if the supplier To manufacturers Provide raw materials, and you will have belong . According to the network topology model, the node degrees of multiple nodes are weighted. The node degree refers to the number of edges connected to a node. In the weighted calculation, each edge not only considers its existence, but also its weight (such as the transportation volume of goods). Weighted node degree It can be expressed as:

[0051] ;

[0052] in, Representation and Node The set of adjacent nodes, Represents edge For example, if the node With Node and connected, and , , then the node The weighted node degree . Perform time series analysis on the cargo flow of multiple nodes to capture the dynamic changes of cargo flow. Time series analysis can use methods such as ARIMA (Autoregressive Integrated Moving Average) model. Assume that the node In time The cargo flow is , the time series model can be expressed as:

[0053] ;

[0054] in, are model parameters, is a white noise term. By fitting historical data, these parameters are estimated and future cargo flow is predicted. According to the cargo flow time series model, the periodic changes of cargo flow are extracted. By performing frequency domain analysis (such as Fourier transform) on the time series, the periodic components are identified. Assume that the node There is a clear cycle in the flow of goods , then the periodic cargo flow pattern can be expressed as In each cycle Cluster analysis is performed on the inventory levels of multiple nodes to identify nodes with similar inventory behavior. Cluster analysis can use methods such as K-means or hierarchical clustering. Assume that the nodes , and The inventory levels are , and , through cluster analysis, these nodes can be grouped. For example, if the node and have similar inventory levels and can be grouped together, while nodes May belong to another category. According to the inventory level clustering model, the inventory level is hierarchically divided to obtain a hierarchical inventory level set. Through hierarchical clustering, different levels of inventory levels are determined, for example, high inventory level, medium inventory level and low inventory level. The weighted node degree value set, the periodic cargo flow pattern and the hierarchical inventory level set are fused to obtain a node feature data set, and the feature of each node can be represented as a vector. According to the node feature data set, the supply chain network is grid-level divided to obtain a grid-level division result. Hierarchical division can be achieved using community discovery algorithms (such as the Girvan-Newman algorithm). These algorithms identify community structures in the network by maximizing modularity. For example, the supply chain network is divided into several levels, each level represents a subnetwork in which nodes have similar characteristics. The grid-level division result is visualized through a visualization tool to obtain a hierarchical structure diagram.

[0055] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0056] (1) Segment the hierarchical structure graph to obtain a set of segmented hierarchical structure subgraphs;

[0057] (2) According to the segmented hierarchical structure subgraph set, the boundary features of each hierarchical structure subgraph are extracted to obtain a boundary feature set;

[0058] (3) Classify the boundary feature set to obtain a hierarchical identification set;

[0059] (4) Convert the hierarchical structure graph into a weighted directed graph model, and perform Laplace matrix feature extraction on the weighted directed graph model to obtain a Laplace feature matrix;

[0060] (5) Perform spectral decomposition on the Laplace characteristic matrix to obtain a characteristic spectrum set, and perform spectral clustering on the weighted directed graph model based on the characteristic spectrum set to obtain the node clustering result;

[0061] (6) Perform density peak clustering on the node clustering results to obtain a density peak clustering model;

[0062] (7) According to the density peak clustering model, the center vector of each node is extracted to obtain a set of node center vectors;

[0063] (8) Merge the node center vector set and the level identifier set to obtain the network feature description set of each node;

[0064] (9) Based on the network feature description set and the level identification set, generate the network feature description and level identification of each node.

[0065] Specifically, the hierarchical structure graph is segmented to obtain a set of segmented hierarchical structure subgraphs. Suppose there is a hierarchical structure graph of a supply chain network, in which nodes represent supply chain entities at different levels and edges represent the connection relationship between them. Through a community discovery algorithm (such as the Girvan-Newman algorithm or the Louvain algorithm), this hierarchical structure graph is segmented into several subgraphs, each of which represents a relatively independent community. The community discovery algorithm identifies the community structure in the network by maximizing the modularity. For example, suppose there is a hierarchical structure graph containing ten nodes. Through a community discovery algorithm, it is segmented into two subgraphs: subgraphs Figure 1 Contains nodes 1 to 5, children Figure 2 Contains nodes 6 to 10. Based on the segmented hierarchical subgraph set, the boundary features of each hierarchical subgraph are extracted to obtain a boundary feature set. Boundary features refer to the nodes in a subgraph connected to other subgraphs and their connection relationships. By identifying the boundary nodes of each subgraph, the features of these nodes are extracted. For example, in the subgraph Figure 1 If nodes 3 and 5 are connected to the child Figure 2 , then nodes 3 and 5 are child nodes. Figure 1 The boundary features of the subgraph nodes include their connection relationships with other subgraph nodes. The boundary feature set is classified to obtain a set of hierarchical identifiers. A machine learning algorithm (such as a support vector machine or a decision tree) is used to classify the boundary features and identify each subgraph as a different level. Figure 1 Identified as "Supplier Level", sub Figure 2Identified as "manufacturer hierarchy". The hierarchical structure graph is converted into a weighted directed graph model, and the Laplace matrix feature extraction is performed on the weighted directed graph model to obtain the Laplace feature matrix. The weighted directed graph model means that each edge in the graph has a direction and weight (such as cargo flow). The Laplace matrix is ​​a matrix that represents the structure of the graph. By calculating the Laplace matrix, the global structural characteristics of the graph are obtained. The Laplace feature matrix is ​​spectrally decomposed to obtain a feature spectrum set. Spectral decomposition refers to decomposing the Laplace matrix into eigenvalues ​​and eigenvectors. The feature spectrum set refers to the set of eigenvalues ​​and eigenvectors. By analyzing the feature spectrum, the structural characteristics of the graph can be understood. According to the feature spectrum set, spectral clustering is performed on the weighted directed graph model to obtain node clustering results. Spectral clustering divides nodes into several groups by spatial clustering of eigenvectors, and each group represents a clustering result. Density peak clustering is performed on the node clustering results to obtain a density peak clustering model. Density peak clustering is a density-based clustering method that divides nodes into several clusters by calculating the local density of each node and the minimum distance from the density peak. For example, assuming that nodes 1 and 2 are density peak nodes, by calculating the distance and density from other nodes to these two nodes, nodes 3, 4, and 5 are classified as cluster 1, and nodes 6, 7, 8, 9, and 10 are classified as cluster 2. According to the density peak clustering model, the center vector of each node is extracted to obtain a set of node center vectors. The center vector refers to the feature vector of each cluster center node. By extracting the center vector, the center characteristics of each node in the cluster are obtained. The node center vector set and the hierarchical identifier set are fused to obtain a set of network feature descriptions for each node. The network feature description set refers to the combination of the center vector and the hierarchical identifier of each node. By fusion of these two sets, a comprehensive feature description of each node is obtained. By matching the network feature description and the hierarchical identifier of each node, the final feature description and hierarchical identifier of each node are obtained.

[0066] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0067] (1) Standardize the network feature description to obtain a standardized network feature vector, and construct a node importance evaluation hierarchical model based on the standardized network feature vector;

[0068] (2) In the node importance evaluation hierarchical model, the level identifier is introduced as one of the evaluation criteria;

[0069] (3) Compare the elements of each level of the node importance evaluation hierarchical model pairwise and construct a judgment matrix;

[0070] (4) Perform consistency check on the judgment matrix and calculate the weight vector of each evaluation criterion based on the judgment matrix;

[0071] (5) The initial importance score of the node is obtained by weighted summation through the weight vector, and the initial importance score of the node is analyzed by grey correlation to obtain the grey correlation coefficient matrix;

[0072] (6) According to the grey correlation coefficient matrix, the initial importance scores of the nodes are modified to obtain the comprehensive scores of the node importance. The comprehensive scores of the node importance are sorted to obtain a list of node importance.

[0073] Specifically, the network feature description is standardized, and the features of different dimensions are converted to the same scale for subsequent analysis. Based on the standardized network feature vector, a node importance evaluation hierarchical model is constructed. The evaluation hierarchy model is a hierarchical structure used to represent the relationship between different features and criteria. In this model, the level identifier is introduced as one of the evaluation criteria to ensure that the comprehensive evaluation of node importance is more comprehensive. For example, the importance of a node can be determined by its connectivity, cargo flow, inventory level and level. The elements of each level of the node importance evaluation hierarchy model are compared pairwise to construct a judgment matrix. The judgment matrix is ​​used to represent the relative importance of different elements. Assuming there are three evaluation criteria: connectivity, cargo flow and inventory level, these criteria are compared pairwise to construct a judgment matrix ,in Representation Guidelines Relative to the criteria The elements of the judgment matrix can be determined by expert scoring or by empirically based proportional scaling. For example, if the importance of connectivity is twice that of cargo flow, and the importance of cargo flow is three times that of inventory level, then the judgment matrix can be expressed as:

[0074] ;

[0075] Conduct consistency check on the judgment matrix to ensure the consistency of the judgment matrix and avoid errors caused by subjective judgment. The formula is:

[0076] ;

[0077] in, is the maximum eigenvalue of the judgment matrix, is the order of the judgment matrix. By calculating the consistency ratio ,if If it is less than 0.1, the consistency of the judgment matrix is ​​acceptable, otherwise the judgment matrix needs to be adjusted. According to the consistency test results, the weight vector of each evaluation criterion is calculated. The weight vector represents the relative importance of each evaluation criterion in the overall evaluation. The weight vector is obtained by solving the eigenvector of the judgment matrix. The initial importance score of the node is obtained by weighted summation through the weight vector. Grey correlation analysis is performed on the initial importance score of the node to obtain the grey correlation coefficient matrix. Grey correlation analysis is used to measure the importance similarity between different nodes. Assume that there are two nodes and , whose initial importance scores are and , grey correlation coefficient The formula is:

[0078] ;

[0079] in, is the resolution coefficient, which is usually taken as 0.5. By calculating the grey correlation coefficient, the grey correlation coefficient matrix is ​​constructed. According to the grey correlation coefficient matrix, the initial importance score of the node is corrected to obtain the comprehensive score of node importance. The correction process can be done by using the weighted summation method, using the grey correlation coefficient as the weight to adjust the initial importance score. The overall score is , the formula is:

[0080] ;

[0081] in, Is a node and nodes The grey correlation coefficient of Is a node The initial importance score of the nodes is sorted by the comprehensive importance scores of the nodes to obtain a node importance list, which reflects the priority of each node in the supply chain network.

[0082] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0083] (1) According to the node importance list, select the node set with the lowest importance and establish a complex network model for the supply chain network;

[0084] (2) According to the complex network model, the initial network topology index set is calculated, and the node set with the lowest importance is gradually removed for simulation;

[0085] (3) After each node is removed, the network topology indicator set is recalculated according to the complex network model;

[0086] (4) Compare the initial network topology index set with the recalculated network topology index set to obtain the topology index change matrix;

[0087] (5) According to the topological index change matrix, extract the key topological index changes and establish an index change model for the key topological index changes;

[0088] (6) All indicator change models are integrated to obtain a comprehensive network indicator change model, and network indicator change data is output based on the comprehensive network indicator change model.

[0089] Specifically, according to the node importance list, the node set with the lowest importance is selected. Assume that there is a supply chain network in which the importance of the nodes has been calculated and sorted by the above method. Select the nodes with the lowest ranking to form the node set with the lowest importance. Assume that there are 10 nodes, numbered from 1 to 10, and their importance scores are 0.9, 0.85, 0.8, 0.75, 0.7, 0.65, 0.6, 0.55, 0.5 and 0.45 respectively. Select the three nodes with the lowest scores (node ​​8, node 9 and node 10) as the node set with the lowest importance. Establish a complex network model for the supply chain network. The complex network model is a graph structure in which nodes represent entities in the supply chain (such as suppliers, manufacturers, distributors and retailers) and edges represent the connection relationship between them (such as the flow of goods). According to the complex network model, calculate the initial network topology index set. Common network topology indicators include node degree, clustering coefficient, average path length and network diameter. These indicators can describe the structural characteristics of the network. Perform a step-by-step removal simulation on the node set with the lowest importance. Select a node from the set of nodes with the lowest importance, remove it from the network, and recalculate the set of topological indicators of the network. Recalculate the set of network topological indicators after removing the node, and calculate the node degree, clustering coefficient and average path length again. After removing a node each time, recalculate the set of network topological indicators according to the complex network model, and compare these results with the initial set of network topological indicators to obtain the topological indicator change matrix. According to the topological indicator change matrix, extract the key topological indicator changes, and establish an indicator change model for the key topological indicator changes. Key topological indicator changes refer to indicator changes that have a greater impact on the overall structure of the network, such as a significant decrease in node degree or a significant increase in average path length. By analyzing these changes, a key indicator change model is established to describe the impact of these changes on the network. All indicator change models are integrated to obtain a comprehensive network indicator change model. The comprehensive network indicator change model integrates various key indicator change models to form an overall model that can fully reflect the impact of node removal on the network structure. By integrating the results of multiple node removal simulations, a comprehensive network indicator change model is obtained. According to the comprehensive network indicator change model, network indicator change data is output. The comprehensive network index change model is used to predict the impact of different node removals on the network structure and provide a basis for decision-making. For example, by gradually removing nodes 8, 9, and 10 and analyzing the changes in the network structure after each removal, a comprehensive model is generated to describe the impact of these nodes on the overall network.

[0090] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0091] (1) Standardize the network indicator change data to obtain a standardized indicator change data set and construct a network indicator hierarchical weight matrix;

[0092] (2) Establish a fuzzy hierarchical analysis model based on the standardized indicator change data set and the network indicator hierarchy weight matrix;

[0093] (3) According to the fuzzy hierarchical analysis model, the weight value of each network indicator is calculated to obtain the network indicator weight set;

[0094] (4) According to the network indicator weight set, the standardized indicator change data set is weighted and fused to obtain the weighted indicator change data;

[0095] (5) Input the weighted index change data into the first layer Q network of the dual-depth Q network to obtain the first layer network output;

[0096] (6) Input the output of the first layer network into the second layer Q network to obtain the output of the second layer network;

[0097] (7) Extract features from the output of the second layer network to obtain a set of multi-layer network feature vectors;

[0098] (8) Generate a multi-layer network perception model based on the multi-layer network feature vector set and the dual deep Q network.

[0099] Specifically, the network indicator change data is standardized and the features of different dimensions are converted to the same scale for subsequent analysis. A network indicator hierarchical weight matrix is ​​constructed. The weight matrix is ​​used to represent the relative importance of different network indicators in the overall evaluation. Assuming there are three network indicators: node degree change, clustering coefficient change, and average path length change, these indicators are compared pairwise to construct a judgment matrix ,in Indicator Relative to the indicator The elements of the judgment matrix can be determined by expert scoring or empirically based proportional scaling. For example, if the importance of node degree change is twice that of clustering coefficient change, and the importance of clustering coefficient change is three times that of average path length change, then the judgment matrix can be expressed as:

[0100] ;

[0101] A fuzzy hierarchical analysis model is established based on the standardized indicator change data set and the network indicator hierarchical weight matrix. The fuzzy hierarchical analysis model combines fuzzy mathematics and hierarchical analysis method to deal with uncertainty and ambiguity in multi-indicator decision-making. By fuzzifying the judgment matrix, a fuzzy judgment matrix can be obtained, and the fuzzy weight value of each indicator can be calculated. According to the fuzzy hierarchical analysis model, the weight value of each network indicator is calculated to obtain the network indicator weight set. The weight value represents the relative importance of each network indicator in the overall evaluation. By solving the fuzzy judgment matrix, the weight set of network indicators is obtained. ,in Indicates The weights of network indicators are calculated according to the fuzzy hierarchical analysis model, and the weight value of each network indicator is obtained to obtain the network indicator weight set. The weight value indicates the relative importance of each network indicator in the overall evaluation. According to the network indicator weight set, the standardized indicator change data set is weighted and fused to obtain the weighted indicator change data. The weighted indicator change data is input into the first layer Q network of the double deep Q network to obtain the first layer network output. The double deep Q network is a reinforcement learning algorithm that uses deep neural networks for decision optimization. The first layer Q network extracts preliminary network features by deep learning the input data. The first layer network output is input into the second layer Q network to obtain the second layer network output. The second layer Q network extracts deeper network features based on further processing of the first layer output. Feature extraction is performed on the second layer network output to obtain a multi-layer network feature vector set. The feature vector set can fully reflect the characteristics of the network at different levels. Through feature extraction, richer network feature information can be obtained. Based on the multi-layer network feature vector set and the double deep Q network, a multi-layer network perception model is generated. The multi-layer network perception model combines multi-layer deep learning and reinforcement learning methods, which can comprehensively perceive and analyze changes in the network, thereby realizing dynamic evaluation and optimization of the supply chain network.

[0102] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0103] (1) Collect real-time node data, real-time cargo flow data, and real-time inventory data from the supply chain network;

[0104] (2) Perform data cleaning and outlier processing on the real-time node data to obtain a purified real-time node data set;

[0105] (3) Performing time series decomposition on the real-time cargo flow data to obtain a set of decomposed real-time cargo flow components;

[0106] (4) Calculate the inventory turnover rate based on the real-time inventory data to obtain a real-time inventory turnover rate data set;

[0107] (5) The purified real-time node data set, the decomposed real-time cargo flow component set, and the real-time inventory turnover rate data set are integrated to obtain the real-time status data set of the supply chain network;

[0108] (6) Input the real-time status data set into the multi-layer network perception model to obtain the network perception output vector;

[0109] (7) Decode the network-aware output vector to obtain the scheduling strategy code;

[0110] (8) According to the scheduling strategy code, determine the supply chain scheduling objective function, and construct constraints for the supply chain scheduling objective function to obtain a set of constraints;

[0111] (9) Use heuristic algorithms to solve multi-objective optimization problems of the scheduling objective function under a set of constraints;

[0112] (10) Based on the multi-objective optimization solution results, an initial supply chain scheduling strategy is generated, and the initial supply chain scheduling strategy is dynamically adjusted to output a dynamic supply chain scheduling strategy.

[0113] Specifically, real-time node data, real-time cargo flow data and real-time inventory data are collected from the supply chain network. Node data includes the status information of each node, such as production capacity, storage capacity, etc. Cargo flow data includes the flow of goods between nodes in different time periods. Inventory data includes the current inventory level of each node. Data cleaning and outlier processing are performed on the real-time node data to obtain a purified real-time node data set. The purpose of data cleaning is to remove noise and erroneous data to ensure the accuracy of the data. Outlier processing is to identify and process data points that deviate from the normal range. For example, if the production capacity of a node suddenly shows a negative value or is much higher than the historical highest value, these data need to be marked as outliers and processed. Time series decomposition is performed on the real-time cargo flow data to obtain a decomposed real-time cargo flow component set. Time series data is decomposed into different components, such as trends, seasonality and random fluctuations. Methods such as seasonal-trend decomposition or empirical mode decomposition are used. For example, assuming that there is cargo flow data for a period of time, the long-term trend, cyclical fluctuations and short-term fluctuations of the cargo flow are obtained through time series decomposition. These components can help the system better understand and predict changes in cargo flow. The inventory turnover rate is calculated for the real-time inventory data to obtain a real-time inventory turnover rate data set. Inventory turnover rate is an important indicator for measuring inventory management efficiency. The purified real-time node data set, the decomposed real-time cargo flow component set and the real-time inventory turnover rate data set are integrated to obtain the real-time status data set of the supply chain network. The real-time status data set is input into the multi-layer network perception model to obtain a network perception output vector. The multi-layer network perception model is a deep learning model that processes the input data through a multi-layer neural network to extract high-level feature information. The network perception output vector is decoded to obtain a scheduling strategy code. The network perception output vector is converted into a specific scheduling strategy. For example, assuming that the network perception output vector contains the prediction results and optimization suggestions for the supply chain status, specific scheduling operations such as adjusting the production plan and optimizing the transportation route are obtained through decoding. According to the scheduling strategy code, the supply chain scheduling objective function is determined, and the supply chain scheduling objective function is constrained to construct a constraint condition set. The scheduling objective function represents the goal to be optimized, such as minimizing inventory costs and maximizing service levels. The constraint condition set is the conditions that must be met during the scheduling process, such as production capacity constraints and transportation capacity constraints. Heuristic algorithms are used to solve multi-objective optimization of the scheduling objective function under a set of constraints. Heuristic algorithms are an optimization method based on experience and heuristic rules, which are suitable for solving complex optimization problems. Common heuristic algorithms include genetic algorithms and particle swarm optimization. Through heuristic algorithms, a near-optimal scheduling solution is found under the premise of satisfying the constraints.According to the multi-objective optimization solution results, the initial supply chain scheduling strategy is generated, and the initial supply chain scheduling strategy is dynamically adjusted to output the dynamic supply chain scheduling strategy. The initial scheduling strategy is generated based on the current state and optimization results, but due to the dynamic changes in the supply chain environment, the strategy needs to be adjusted in real time to cope with emergencies and changes. For example, if the inventory of a certain node suddenly drops, the production plan and transportation route need to be adjusted to ensure the continuity and stability of the supply chain.

[0114] The above describes the business data processing method based on supply chain management in the embodiment of the present application. The following describes the business data processing system based on supply chain management in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the business data processing system based on supply chain management includes:

[0115] The collection module 201 is used to collect and integrate the node degrees, cargo flows and inventory levels of multiple nodes in the supply chain network to obtain a node feature data set, and to divide the supply chain network into grid levels according to the node feature data set to obtain a hierarchical structure diagram;

[0116] The processing module 202 is used to perform node feature processing on multiple nodes according to the hierarchical structure diagram to obtain a network feature description and a hierarchical identifier of each node;

[0117] The calculation module 203 is used to calculate the importance comprehensive score of multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and sort the multiple nodes according to the importance comprehensive score to obtain a node importance list;

[0118] An analysis module 204 is used to perform node removal simulation and network indicator change analysis on the supply chain network based on the node importance list to obtain network indicator change data;

[0119] Modeling module 205, used to perform fuzzy hierarchical analysis on network indicator change data to obtain indicator selection evaluation results, and use a dual-depth Q network to perform multi-layer network perception modeling on the indicator selection evaluation results to obtain a multi-layer network perception model;

[0120] The output module 206 is used to obtain the real-time status data set of the supply chain network, input the real-time status data set into the multi-layer network perception model to perform supply chain scheduling strategy analysis, and output a dynamic supply chain scheduling strategy.

[0121] Through the cooperation of the above components, the node degree, cargo flow and inventory level of multiple nodes in the supply chain network are collected and integrated to form a node feature data set. By dividing the node feature data set into grid levels, a hierarchical structure diagram is generated, and node feature processing is performed on multiple nodes to obtain network feature descriptions and hierarchical identifications. According to the network feature description and hierarchical identification, the importance of multiple nodes is calculated and a node importance list is generated. By introducing a variety of evaluation indicators, the importance of nodes is comprehensively evaluated. Based on the node importance list, the supply chain network is simulated for node removal and the network indicator change analysis is performed to generate network indicator change data, and the impact of node removal on the supply chain network is pre-evaluated to help discover key nodes and potential weak links, and optimize the supply chain structure and resource allocation. By performing fuzzy hierarchical analysis on the network indicator change data, the indicator selection evaluation results are obtained, and a dual-depth Q network is used for multi-layer network perception modeling to generate a multi-layer network perception model, which can more accurately capture the dynamic change characteristics of the supply chain network and improve the perception ability and prediction accuracy of the supply chain status. By obtaining the real-time status data set of the supply chain network and inputting it into the multi-layer network perception model to analyze the supply chain scheduling strategy, a dynamic supply chain scheduling strategy is output. It can quickly respond to changes in the supply chain network, optimize scheduling decisions, improve the flexibility and adaptability of the supply chain, and thus improve the accuracy of business data processing in supply chain management.

[0122] The present application also provides a business data processing device based on supply chain management, and the business data processing device based on supply chain management includes a memory and a processor, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the steps of the business data processing method based on supply chain management in the above-mentioned embodiments.

[0123] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the business data processing method based on supply chain management.

[0124] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0125] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0126] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A business data processing method based on supply chain management, characterized in that: The business data processing method based on supply chain management includes: Collecting and integrating node degrees, cargo flows and inventory levels of multiple nodes in a supply chain network to obtain a node feature data set, and dividing the supply chain network into grid levels according to the node feature data set to obtain a hierarchical structure diagram; According to the hierarchical structure graph, node feature processing is performed on the multiple nodes to obtain a network feature description and a hierarchical identification of each node; specifically, the processing includes: segmenting the hierarchical structure graph to obtain a hierarchical structure subgraph set after segmentation; according to the hierarchical structure subgraph set after segmentation, boundary feature extraction is performed on each hierarchical structure subgraph to obtain a boundary feature set; classifying the boundary feature set to obtain a hierarchical identification set; converting the hierarchical structure graph into a weighted directed graph model, and performing Laplace matrix feature extraction on the weighted directed graph model to obtain a Laplace feature matrix; The Lass feature matrix is ​​spectrally decomposed to obtain a feature spectrum set, and the weighted directed graph model is spectrally clustered according to the feature spectrum set to obtain a node clustering result; the node clustering result is subjected to density peak clustering to obtain a density peak clustering model; according to the density peak clustering model, the center vector of each node is extracted to obtain a node center vector set; the node center vector set and the level identification set are merged to obtain a network feature description set of each node; according to the network feature description set and the level identification set, a network feature description and a level identification of each node are generated; Calculating the importance comprehensive scores of the multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and sorting the multiple nodes according to the importance comprehensive score to obtain a node importance list; Based on the node importance list, the supply chain network is subjected to node removal simulation and network indicator change analysis to obtain network indicator change data; specifically comprising: selecting the lowest importance node set according to the node importance list, and establishing a complex network model for the supply chain network; calculating the network initial topology indicator set according to the complex network model, and performing a step-by-step removal simulation on the lowest importance node set; after each node is removed, recalculating the network topology indicator set according to the complex network model; comparing the network initial topology indicator set with the recalculated network topology indicator set to obtain a topology indicator change matrix; extracting key topology indicator changes according to the topology indicator change matrix, and establishing an indicator change model for the key topology indicator changes; fusing all indicator change models to obtain a comprehensive network indicator change model, and outputting network indicator change data according to the comprehensive network indicator change model; Performing fuzzy hierarchical analysis on the network indicator change data to obtain indicator selection evaluation results, and using a double-depth Q network to perform multi-layer network perception modeling on the indicator selection evaluation results to obtain a multi-layer network perception model; A real-time status data set of the supply chain network is obtained, and the real-time status data set is input into the multi-layer network perception model to perform supply chain scheduling strategy analysis, and a dynamic supply chain scheduling strategy is output.

2. The business data processing method based on supply chain management according to claim 1 is characterized in that: The node degrees, cargo flows and inventory levels of multiple nodes in the supply chain network are collected and integrated to obtain a node feature data set, and the supply chain network is grid-level divided according to the node feature data set to obtain a hierarchical structure diagram, including: Model the network topology structure of multiple nodes in the supply chain network to obtain a network topology model; According to the network topology model, performing weighted calculation on the node degrees of the multiple nodes to obtain a weighted node degree value set; Performing time series analysis on the cargo flows of the multiple nodes to obtain a cargo flow time series model; According to the cargo flow time series model, extracting periodic changes of the cargo flow to obtain a periodic cargo flow pattern; Performing cluster analysis on the inventory levels of the multiple nodes to obtain an inventory level clustering model; According to the inventory level clustering model, the inventory level is hierarchically divided to obtain a hierarchical inventory level set; Merging the weighted node degree value set, the periodic cargo flow pattern and the hierarchical inventory level set to obtain a node feature data set; The supply chain network is divided into grid levels according to the node feature data set to obtain a grid level division result, and the grid level division result is visualized to obtain a hierarchical structure diagram.

3. The business data processing method based on supply chain management according to claim 1 is characterized in that: The calculating of the importance comprehensive scores of the multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and sorting the multiple nodes according to the importance comprehensive score to obtain a node importance list includes: Standardizing the network feature description to obtain a standardized network feature vector, and constructing a node importance evaluation hierarchical model based on the standardized network feature vector; In the node importance evaluation hierarchical model, the level identifier is introduced as one of the evaluation criteria; Compare the elements of each level of the node importance evaluation hierarchical model in pairs to construct a judgment matrix; Performing a consistency check on the judgment matrix, and calculating a weight vector of each evaluation criterion according to the judgment matrix; Performing weighted summation through the weight vector to obtain the initial importance score of the node, and performing grey correlation analysis on the initial importance score of the node to obtain a grey correlation coefficient matrix; According to the grey correlation coefficient matrix, the initial importance scores of the nodes are corrected to obtain a comprehensive importance score of the nodes, and the comprehensive importance scores of the nodes are sorted to obtain a node importance list.

4. The business data processing method based on supply chain management according to claim 1 is characterized in that: The fuzzy hierarchical analysis is performed on the network indicator change data to obtain the indicator selection evaluation result, and the double-depth Q network is used to perform multi-layer network perception modeling on the indicator selection evaluation result to obtain a multi-layer network perception model, including: Standardizing the network indicator change data to obtain a standardized indicator change data set, and constructing a network indicator hierarchical weight matrix; Establishing a fuzzy hierarchical analysis model according to the standardized indicator change data set and the network indicator hierarchical weight matrix; According to the fuzzy hierarchical analysis model, the weight value of each network indicator is calculated to obtain a network indicator weight set; According to the network indicator weight set, weighted fusion is performed on the standardized indicator change data set to obtain weighted indicator change data; Inputting the weighted index change data into the first layer Q network of the dual-depth Q network to obtain the first layer network output; Input the output of the first layer network into the second layer Q network to obtain the output of the second layer network; Performing feature extraction on the output of the second layer network to obtain a multi-layer network feature vector set; A multi-layer network perception model is generated according to the multi-layer network feature vector set and the dual-depth Q network.

5. The business data processing method based on supply chain management according to claim 1 is characterized in that: The step of acquiring a real-time status data set of the supply chain network, inputting the real-time status data set into the multi-layer network perception model to perform supply chain scheduling strategy analysis, and outputting a dynamic supply chain scheduling strategy includes: Collect real-time node data, real-time cargo flow data and real-time inventory data from the supply chain network; Performing data cleaning and outlier processing on the real-time node data to obtain a purified real-time node data set; Performing time series decomposition on the real-time cargo flow data to obtain a decomposed real-time cargo flow component set; Calculating the inventory turnover rate of the real-time inventory data to obtain a real-time inventory turnover rate data set; Merging the purified real-time node data set, the decomposed real-time cargo flow component set and the real-time inventory turnover rate data set to obtain a real-time status data set of the supply chain network; Inputting the real-time status data set into the multi-layer network perception model to obtain a network perception output vector; Decoding the network-aware output vector to obtain a scheduling strategy code; Determine the supply chain scheduling objective function according to the scheduling strategy code, and construct constraint conditions for the supply chain scheduling objective function to obtain a constraint condition set; Using a heuristic algorithm to perform multi-objective optimization on the scheduling objective function under the set of constraints; According to the multi-objective optimization solution, an initial supply chain scheduling strategy is generated, and the initial supply chain scheduling strategy is dynamically adjusted to output a dynamic supply chain scheduling strategy.

6. A business data processing system based on supply chain management, characterized in that: Used to execute the business data processing method based on supply chain management according to any one of claims 1 to 5, the business data processing system based on supply chain management comprises: A collection module, used to collect and integrate the node degrees, cargo flows and inventory levels of multiple nodes in the supply chain network to obtain a node feature data set, and to divide the supply chain network into grid levels according to the node feature data set to obtain a hierarchical structure diagram; A processing module, used to perform node feature processing on the multiple nodes according to the hierarchical structure diagram to obtain a network feature description and a hierarchical identifier of each node; A calculation module, configured to calculate the importance comprehensive scores of the multiple nodes according to the network feature description and the level identifier to obtain the importance comprehensive score of each node, and to sort the multiple nodes according to the importance comprehensive score to obtain a node importance list; An analysis module, used to perform node removal simulation and network indicator change analysis on the supply chain network based on the node importance list to obtain network indicator change data; A modeling module is used to perform fuzzy hierarchical analysis on the network indicator change data to obtain an indicator selection evaluation result, and use a double-depth Q network to perform multi-layer network perception modeling on the indicator selection evaluation result to obtain a multi-layer network perception model; The output module is used to obtain the real-time status data set of the supply chain network, input the real-time status data set into the multi-layer network perception model to perform supply chain scheduling strategy analysis, and output a dynamic supply chain scheduling strategy.

7. A business data processing device based on supply chain management, characterized in that: The business data processing device based on supply chain management includes: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the supply chain management-based business data processing device to execute the supply chain management-based business data processing method according to any one of claims 1 to 5.

8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the business data processing method based on supply chain management as described in any one of claims 1 to 5 is implemented.

Citation Information

Patent Citations

  • Intelligent decision-making-based judicial library management method, system and device, and storage medium

    CN117670066A

  • Data management method and system for supply chain

    CN118115000A

  • Distributed Additive Manufacturing Platform for Value Chain Networks

    US20230080545A1