Supply chain prediction method based on knowledge graph and industrial model

Abstract

The invention provides a supply chain prediction method based on a knowledge graph and an industrial model, and the method comprises the steps: collecting the features needed by supply chain prediction, and the needed features comprise historical order data and production plan data, industrial product purchasing data and collection information, logistics data and storage data; performing data cleaning, standardization and correlation analysis on the historical order data, the production plan data, the industrial product purchasing data, the collection information, the logistics data and the storage data to obtain target features; key entities and relationships in the target features are identified, and a knowledge graph is constructed based on the key entities and the relationships; the industrial model is used for identifying supply chain key nodes and node relations in the knowledge graph, the supply chain key nodes and the node relations are converted into numerical feature vectors used for supply chain prediction so as to carry out supply chain prediction, the relations among the nodes in the supply chain can be better captured by introducing the knowledge graph and the industrial model, and the supply chain prediction accuracy is improved. Therefore, the supply chain prediction precision is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of supply chain forecasting technology, and in particular to a supply chain forecasting method based on knowledge graph and industrial model. Background Art

[0002] In the field of supply chain management, accurate supply chain forecasting is of great significance. Traditional supply chain forecasting methods usually rely on historical data and simple statistical models. However, these methods have many limitations and defects. For example, inflexible feature selection and insufficient model optimization lead to poor results in dealing with supply chain complexity and uncertainty. In recent years, forecasting methods based on knowledge graphs and industrial large models have gradually attracted attention. As a structured representation method, knowledge graphs make up for the shortcomings of traditional methods in capturing complex supply chain relationships by integrating and associating multi-industry data.

[0003] However, most knowledge graph-based prediction models lack adaptability in dynamic environments and cannot effectively respond to real-time changes in the supply chain, resulting in poor accuracy in supply chain predictions. Summary of the invention

[0004] The present invention provides a supply chain forecasting method based on knowledge graph and industrial model to solve the defect of poor accuracy of supply chain forecasting in the prior art.

[0005] In a first aspect, the present invention provides a supply chain prediction method based on a knowledge graph and an industrial model, comprising:

[0006] Collecting features required for supply chain forecasting, including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data, and warehousing data;

[0007] Performing data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain target features;

[0008] Identify key entities and relationships in the target feature, and construct a knowledge graph based on the key entities and relationships;

[0009] The industrial model is used to identify key supply chain nodes and node relationships in the knowledge graph, and is converted into numerical feature vectors for supply chain forecasting to perform supply chain forecasting.

[0010] A supply chain prediction method based on a knowledge graph and an industrial model provided by the present invention also includes:

[0011] Collecting actual data and forecast data of the supply chain in real time for comparison;

[0012] Based on the results of the comparison, deviation analysis is performed and optimization suggestions are generated;

[0013] Based on the optimization suggestions, the industrial model is dynamically adjusted and optimized through a real-time feedback loop.

[0014] According to a supply chain forecasting method based on knowledge graph and industrial model provided by the present invention, the historical order data and production plan data in the features required for supply chain forecasting are collected, including:

[0015] Obtain historical order files and production plan files through the enterprise resource planning platform and manufacturing execution platform;

[0016] Through data aggregation, the historical order files and the production plan files are weightedly aggregated according to time series and dimensions to obtain historical order data and production plan data.

[0017] According to a supply chain forecasting method based on knowledge graph and industrial model provided by the present invention, the industrial product procurement data and centralized procurement information in the features required for supply chain forecasting are collected, including:

[0018] Obtain supplier data and material data through supply chain management system and procurement management system;

[0019] Input the supplier data and the material data into the supplier and material association matrix, and output industrial product procurement data and centralized procurement information.

[0020] According to a supply chain prediction method based on knowledge graph and industrial model provided by the present invention, the collection of logistics data and warehousing data in the features required for supply chain prediction includes:

[0021] Obtain logistics information and warehousing information through logistics management system and warehousing management system;

[0022] The logistics information and the warehousing information are input into the logistics and warehousing efficiency analysis model, and the efficiency index is output as logistics data and warehousing data.

[0023] According to a supply chain prediction method based on knowledge graph and industrial model provided by the present invention, the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data are subjected to correlation analysis to obtain target features, including:

[0024] Represent the historical order data and production plan data, the industrial product procurement data and centralized procurement information, and the logistics data and warehousing data after data cleaning and standardization as corresponding features respectively;

[0025] Determining the correlation between each of the features using a weighted Pearson correlation coefficient;

[0026] Based on the correlation between the various features, the target feature is determined.

[0027] According to a supply chain prediction method based on a knowledge graph and an industrial model provided by the present invention, the identifying key entities and relationships in the target features and constructing a knowledge graph based on the key entities and relationships include:

[0028] Based on the target features, identifying entities in the supply chain, the entities including suppliers, products, and logistics nodes;

[0029] Determine the relationship between the supplier, the product and the logistics node, and obtain the upstream and downstream relationship of the supply chain, the logistics path and transportation relationship, and the inventory management relationship;

[0030] The suppliers, the products, the logistics nodes, the upstream and downstream relationships of the supply chain, the logistics routes and transportation relationships, and the inventory management relationships are integrated into a multi-level knowledge graph, and a knowledge graph is constructed through structured integration, mathematical modeling, and graph reasoning.

[0031] According to a supply chain prediction method based on a knowledge graph and an industrial model provided by the present invention, the supplier, the product, the logistics node, the upstream and downstream relationship of the supply chain, the logistics path and transportation relationship, and the inventory management relationship are integrated into a multi-level knowledge graph, and a knowledge graph is constructed through structured integration, mathematical modeling, and graph reasoning, including:

[0032] The supplier, the product, the logistics node, the upstream and downstream relationship of the supply chain, the logistics path and transportation relationship and the inventory management relationship are formally represented by using graph theory;

[0033] Introducing cross-industry data sets, and modeling the probabilistic dependencies between the entities after the formal representation through Bayesian networks, to obtain a high-dimensional knowledge graph;

[0034] Through graph embedding, the high-dimensional knowledge graph is embedded into a low-dimensional space, retaining the topological structure and semantic relationship of the nodes to obtain a knowledge graph.

[0035] According to a supply chain prediction method based on a knowledge graph and an industrial model provided by the present invention, the industrial model is used to identify key supply chain nodes and node relationships in the knowledge graph and convert them into numerical feature vectors for supply chain prediction, including:

[0036] Using the graph convolutional network in the industrial model to identify the information of each node in the knowledge graph;

[0037] Aggregating information of each node and its neighboring nodes using neighbor information to generate a new data representation;

[0038] Based on the data representation, identifying supply chain prediction features using the graph convolutional network;

[0039] The supply chain prediction features are converted into numerical feature vectors.

[0040] According to a supply chain prediction method based on knowledge graph and industrial model provided by the present invention, after converting the supply chain prediction feature into a numerical feature vector, it also includes:

[0041] Based on the numerical feature vector, construct a supply chain scenario model;

[0042] Using the supply chain scenario model, simulate forecasting performance under different supply chain environments;

[0043] Based on the error of the forecast performance, the applicability and stability of the supply chain scenario model are determined.

[0044] In a second aspect, the present invention further provides a supply chain prediction device based on a knowledge graph and an industrial model, comprising:

[0045] A collection module is used to collect features required for supply chain forecasting, including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data, and warehousing data;

[0046] An analysis module is used to perform data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain target features;

[0047] A construction module, used to identify key entities and relationships in the target feature, and construct a knowledge graph based on the key entities and relationships;

[0048] The prediction module is used to use the industrial model to identify the key supply chain nodes and node relationships in the knowledge graph, and convert them into numerical feature vectors for supply chain prediction to perform supply chain prediction.

[0049] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a supply chain prediction method based on a knowledge graph and an industrial model as described above is implemented.

[0050] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a supply chain forecasting method based on a knowledge graph and an industrial model as described in any one of the above.

[0051] In a fifth aspect, the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the supply chain prediction methods based on knowledge graphs and industrial models as described above.

[0052] The present invention provides a supply chain prediction method based on knowledge graph and industrial model, including collecting features required for supply chain prediction, including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data; performing data cleaning, standardization and correlation analysis on historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data to obtain target features; identifying key entities and relationships in target features, and constructing knowledge graph based on key entities and relationships; using industrial model to identify key supply chain nodes and node relationships in knowledge graph, and converting them into numerical feature vectors for supply chain prediction, so as to perform supply chain prediction. By introducing knowledge graph and industrial model, the relationship between nodes in the supply chain can be better captured, thereby effectively improving the accuracy of supply chain prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 is a flow chart of a supply chain prediction method based on a knowledge graph and an industrial model provided in this embodiment;

[0055] Figure 2 is a schematic diagram of supply chain prediction correlation results provided by this embodiment;

[0056] Figure 3 is a result schematic diagram of the key entity and relationship structure diagram provided in this embodiment;

[0057] Figure 4 is a schematic diagram of the knowledge graph structure provided in this embodiment;

[0058] Figure 5 is a schematic diagram of the hybrid model structure provided in this embodiment;

[0059] Figure 6 is a schematic diagram of MSE error provided in this embodiment;

[0060] Figure 7 is a schematic diagram of the structure of a supply chain prediction device based on a knowledge graph and an industrial model provided in this embodiment;

[0061] Figure 8 It is a schematic diagram of the structure of the electronic device provided in this embodiment. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Figure 1 It is a flowchart of the supply chain prediction method based on knowledge graph and industrial model provided in this embodiment.

[0064] like Figure 1 As shown, the supply chain prediction method based on knowledge graph and industrial model provided by the embodiment of the present invention mainly includes the following steps:

[0065] 101. Collect the features required for supply chain forecasting, including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data.

[0066] In a specific implementation process, the acquisition of historical order data and production plan data is to ensure that the collected data is comprehensive and accurate and can reflect the actual operation of the enterprise. The acquisition of industrial product procurement data and centralized procurement information is the basis of data collection and provides a new perspective for supply chain management. Logistics data and warehousing data are to ensure that all links from transportation to inventory have sufficient data support to improve the overall efficiency of the supply chain.

[0067] The historical order data and production plan data are obtained as follows:

[0068] Obtain historical order files and production plan files through the enterprise resource planning platform and manufacturing execution platform. Access the company's ERP (enterprise resource planning) and MES (manufacturing execution system) platforms, and log in to the order management module and production planning module respectively. In these two modules, select the time range (for example, the past 12 months) and key data fields to be extracted, such as order number, order date, product number, product quantity, planned production date, production line number, etc. Then use the built-in query function of the module to filter out data related to specific product lines or production areas. You can set up multi-condition queries, such as filtering by order priority, customer type, or product type, to ensure that the extracted data can more accurately reflect the company's actual production and order execution.

[0069] Through data aggregation, historical order files and production plan files are weighted and aggregated according to time series and dimensions to obtain historical order data and production plan data. The formula for weighted aggregation is (1):

[0070]

[0071] Where Y t,d,k Represents the weighted aggregation result within the time window t (e.g., monthly), dimension d (e.g., a specific product number), and category k (e.g., high-priority orders), that is, the weighted total order quantity or production plan quantity. i,t,d Represents the i-th data point in time window t, dimension d, and category k, such as the quantity of an order or the quantity of a production plan. i,k is the weight factor of the ith data point in category k, which is used to adjust the weights of different data points according to their importance. For example, high-priority orders may be given higher weights. t ,d represents the total number of data points in the time window t and dimension d, that is, the total number of orders or production plans under this time period and product dimension. I(C i,k ) is an indicator function used to determine whether data point i satisfies a specific condition C i,k (For example, whether it is a high priority order). If the conditions are met, then I(C i,k )=1; otherwise I(C i,k )=0.

[0072] Through data aggregation, weighted time series data reflecting order demand fluctuations and production load change trends can be generated for subsequent trend analysis and forecasting. The aggregation algorithm can dynamically adjust the time window (for example, by week or month) and perform weighting and filtering according to specific conditions (such as order priority or customer type) to adapt to data analysis on different time scales and business needs.

[0073] As shown in Table 1, the filtered, integrated and aggregated order history and production plan data are displayed, including order number, order date, product number, product quantity, planned production date and production line number.

[0074]

[0075] Table 1

[0076] Therefore, it not only ensures the accuracy and completeness of historical order data and production plan data, but also provides a high-quality data foundation for subsequent supply chain forecasts.

[0077] Among them, the methods for obtaining industrial product procurement data and centralized procurement information are as follows:

[0078] Obtain supplier data and material data through the supply chain management system and procurement management system. Access the company's supply chain management system (SCM) or procurement management system, enter the purchase order management module and the centralized procurement project management module, select the time range to be extracted (for example, the past 12 months) and key data fields, such as purchase order number, purchase date, material number, purchase quantity, supplier information, purchase price, delivery date, centralized procurement project number, etc. Through the built-in query function, filter out data related to specific suppliers, material categories or centralized procurement projects. You can set up multi-condition queries, such as filtering by material priority, supplier rating, purchase amount or the scale of centralized procurement projects to ensure that the extracted data can accurately reflect the company's actual procurement behavior and centralized procurement activities.

[0079] In order to reveal deeper supply chain relationships in procurement data, supplier data and material data are input into the supplier-material association matrix, and industrial product procurement data and centralized procurement information are output. The supplier-material association matrix is ​​as shown in formula (2):

[0080]

[0081] In the formula, A s,m Represents the strength of association between supplier s and material m, that is, the proportion of the supplier's supply of a certain type of material. i,s,m Represents the quantity or amount of material m in the i-th order provided by supplier s. m Represents the total quantity or total amount of material m purchased by the enterprise during the entire time range. s Represents the total number of orders provided by supplier s. I(V i,s ) is an indicator function used to determine whether the order belongs to a high-quality supplier V i,s If yes, then I(V i,s )=1; otherwise I(V i,s )=0.

[0082] As shown in Table 2, an example of data calculated based on the supplier-material association matrix analysis method is presented, including supplier number, material number, supplier-material association strength, total purchase quantity, high-quality suppliers and other information.

[0083]

[0084] Table 2

[0085] Through the correlation matrix analysis, enterprises can identify which suppliers have a higher supply ratio in a specific material category, as well as the importance of high-quality suppliers in the supply of key materials. It can not only reflect the supply stability of suppliers, but also provide data support for the optimization of centralized procurement strategies.

[0086] Among them, the methods for obtaining logistics data and warehousing data are as follows:

[0087] Obtain logistics information and warehousing information through the logistics management system and warehouse management system. Access the company's logistics management system (TMS, transportation management system) and warehouse management system (WMS, warehouse management system), and log in to the transportation scheduling module and inventory management module respectively. Select the time range (for example, the past 12 months) and key data fields to be extracted, including transportation batch number, shipping date, transportation route, transportation time, transportation cost, storage location, inventory quantity, inventory turnover rate, etc. Through the built-in query function, filter out data related to specific transportation routes, storage locations or product types. You can set up multi-condition queries, such as filtering by transportation frequency, warehouse utilization, or product characteristics, to ensure that the extracted data can accurately reflect the company's operating efficiency and potential bottlenecks in the logistics and warehousing links.

[0088] In order to analyze the logistics and warehousing data more deeply, the logistics and warehousing efficiency analysis formula can be used to calculate the comprehensive index of transportation efficiency and warehousing efficiency, that is, input logistics information and warehousing information into the logistics and warehousing efficiency analysis model, and output the efficiency index as logistics data and warehousing data, as shown in formula (3):

[0089]

[0090] In the formula, E l,w,t Represents the comprehensive efficiency index under the transportation route l, storage location w and time window t. l Represents the unit transportation time of transportation route l, such as the average transportation time of each batch. l Represents the unit transportation cost of transportation route l, such as the average transportation cost per batch. w Represents the inventory turnover rate of storage location w, reflecting the efficiency of warehouse utilization. w,tRepresents the ideal inventory level or target inventory level of storage location w within time window t.

[0091] Through formula (3), enterprises can calculate the comprehensive efficiency under different time periods, transportation routes and storage locations, and then identify the most efficient link in the supply chain. As shown in Table 3, an example of data calculated according to the logistics and storage efficiency analysis formula is shown.

[0092]

[0093] Table 3

[0094] Therefore, enterprises can not only ensure the accuracy and completeness of logistics data and warehousing data, but also identify key links that need to be improved in the supply chain through comprehensive efficiency analysis, thereby improving the overall logistics and warehousing management level.

[0095] 102. Perform data cleaning, standardization and correlation analysis on historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data to obtain target features.

[0096] First, the historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data are cleaned and standardized. Specifically, the data is checked to identify and process outliers, duplicate values ​​and missing values ​​to ensure the integrity and accuracy of the data. Data from different sources are uniformly standardized to keep the data consistent in scale and format for subsequent analysis and comparison. After cleaning and standardization, the data is checked for consistency to ensure data quality and lay the foundation for subsequent analysis.

[0097] Then, data correlation analysis is performed, which aims to identify the linear relationship between different features, so as to find the variables that have the greatest impact on supply chain forecasting. In order to improve the accuracy and practicality of the analysis, the weighted Pearson correlation coefficient and weighted partial correlation coefficient are introduced to better reflect the actual relationship between different features when controlling the influence of other variables.

[0098] Perform correlation analysis to obtain target features, including:

[0099] The historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data after data cleaning and standardization are respectively represented as corresponding features. The weighted Pearson correlation coefficient is used to determine the correlation between the various features. The weighted Pearson correlation coefficient can assign different weights to different data points, so that the analysis results are more in line with the actual situation in the supply chain. For example, when analyzing the relationship between order history data and industrial product procurement data, key orders or large purchases can be given higher weights to highlight the importance of these data points in supply chain management, as shown in formula (4):

[0100]

[0101] In the formula, is the weighted Pearson correlation coefficient between features x and y, taking into account the weight w of each data point i .w i is the weight of the ith data point, reflecting the importance of the data point. The larger the weight, the greater the impact of the data point on the correlation coefficient. i and i are the observed values ​​of feature x and feature y in the i-th sample, respectively. and is the weighted mean of features x and y, defined as (5) and (6), respectively:

[0102]

[0103] In the formula, the molecular part and represents the weighted covariance, which reflects the joint change of feature x and feature y under weight distribution. The denominator is composed of the square roots of two weighted variances, which are and They are used to normalize the weighted covariances so that the correlation coefficients are between -1 and 1.

[0104] However, the relationship between different features in the supply chain may be affected by other variables, such as logistics and warehousing data. In this case, the true relationship between the two features may not be fully understood by the weighted Pearson correlation coefficient alone. Therefore, the weighted partial correlation coefficient is used to measure the feature relationship after controlling the influence of other variables (such as logistics data). It helps to reveal the relationship between the two main features by controlling the influence of other variables (such as logistics data). Calculating the weighted partial correlation coefficient can determine how strong the relationship between historical order data and production plan data and industrial product procurement data and centralized procurement information is after excluding the influence of other variables (such as logistics data). As shown in formula (7):

[0105]

[0106] Where: It is the weighted partial correlation coefficient between historical order data and production plan data (x) and industrial product procurement data and centralized procurement information (y) after controlling the influence of logistics and warehousing data (z). It is the weighted Pearson correlation coefficient between historical order data and production plan data and industrial product procurement data and centralized procurement information. This coefficient takes into account the importance of key orders and large purchases to supply chain forecasts. and They are the weighted correlation coefficients between historical order data and production plan data and logistics data and warehousing data, and between industrial product procurement data and centralized procurement information and logistics data and warehousing data.

[0107] Finally, based on the correlation between each feature, the target feature is determined. The most critical feature for supply chain forecasting is selected as the target feature. Figure 2 is a schematic diagram of the supply chain prediction correlation results provided by this embodiment, such as Figure 2 As shown, the value in each square represents the weighted correlation coefficient between the two features: the correlation between historical order data and production plan data and industrial product procurement data and centralized procurement information is 1, showing a very strong positive correlation, which means that the order history and production plan interact closely with the procurement of industrial products in the supply chain.

[0108] The correlation between historical order data and production plan data and logistics data and warehousing data is 1, which also shows a high positive correlation, indicating that there is a strong dependency between order and production planning and logistics management. The correlation between industrial product procurement data and centralized procurement information and logistics data and warehousing data is 1, showing a strong correlation between procurement and logistics, indicating that procurement quantity and logistics arrangements are closely related.

[0109] 103. Identify key entities and relationships in target features and build a knowledge graph based on key entities and relationships.

[0110] Specifically, we first define key entities and relationships. Based on the target characteristics from the results of the relationship analysis, we identify the entities, i.e. the core elements, in the supply chain and clarify the relationships between them, thus providing a solid foundation for the construction of the knowledge graph. Entities include suppliers, products, and logistics nodes, specifically:

[0111] (1) Suppliers: The importance of suppliers in the supply chain, such as their supply capacity, response time, and correlation with order history data, is determined through weighted Pearson correlation coefficient and weighted partial correlation coefficient analysis.

[0112] (2) Product: When defining product entities, it is necessary to consider key attributes such as production cycle and output demand reflected in production planning data, which have been identified as features that are highly correlated with supply chain performance.

[0113] (3) Logistics nodes: Logistics nodes include warehouses, transportation routes, distribution centers, etc. Through correlation analysis, we can clarify the relationship between these nodes and suppliers and products, as well as their impact on the overall supply chain efficiency.

[0114] Then determine the relationship between suppliers, products and logistics nodes, and obtain the upstream and downstream relationships of the supply chain, logistics paths and transportation relationships, and inventory management relationships, specifically:

[0115] (1) Upstream and downstream relationships in the supply chain: Based on the results of correlation analysis, clarify the upstream and downstream relationships between suppliers, products, and logistics nodes, such as the supply relationship between suppliers and products, and the flow paths of products at different logistics nodes.

[0116] (2) Logistics paths and transportation relationships: Define the transportation paths between logistics nodes and products and suppliers, and determine the efficiency and bottlenecks of each transportation route by analyzing logistics and warehousing data.

[0117] (3) Inventory management relationships: By analyzing inventory data, the inventory management relationships between products at various logistics nodes are defined, including inventory turnover rate, storage cost, etc. These relationships will help optimize the inventory management strategy in the supply chain.

[0118] Figure 3 It is a schematic diagram of the key entities and relationship structure diagram in the supply chain, such as Figure 3 As shown in the figure, the four core entities in the supply chain (suppliers, products, logistics nodes, and inventory) and their key relationships are shown. These relationships include the supply relationship between suppliers and products, the transportation path between products and logistics nodes, the inventory management relationship between logistics nodes and inventory, and the logistics distribution relationship between suppliers and logistics nodes. It intuitively expresses the core links in the supply chain and their interdependencies, providing a structured basis for further analysis and optimization.

[0119] Secondly, suppliers, products, logistics nodes, upstream and downstream relationships in the supply chain, logistics routes, transportation relationships, and inventory management relationships are integrated into a multi-level knowledge graph, and a knowledge graph is constructed through structured integration, mathematical modeling, and graph reasoning. After completing the definition of key entities and relationships, these elements are integrated into a multi-level knowledge graph, and by introducing cross-industry data sets, a knowledge framework that can fully capture the complexity of the supply chain is constructed. It not only includes structured integration of data, but also involves complex mathematical modeling and graph reasoning algorithms to ensure that the knowledge graph can play its maximum role in subsequent analysis and optimization. Specifically:

[0120] In the first step, the core entities, namely suppliers, products, logistics nodes, upstream and downstream relationships in the supply chain, logistics routes, transportation relationships, and inventory management relationships, are formally represented through graph theory. i can be represented as a node, and each relationship R ij Corresponding to the directed edges between entities, the following formula can be used to represent the relationship in the graph, such as formula (8):

[0121] G=(V,E),V={E1,E2,....,E n},E={R ij ∣i,j∈V} (8)

[0122] In the formula, G represents the overall structure of the knowledge graph; V is the set of all entity nodes; E is the set of all relations, and R ij Represents entity E i and entity E j The relationship between.

[0123] In the second step, in order to further enrich the expressive power of the knowledge graph, a cross-industry data set is introduced, and the probabilistic dependencies between entities are formally represented through Bayesian network modeling to obtain a high-dimensional knowledge graph. Bayesian networks can effectively handle uncertainty and complex dependencies, which is especially important for the randomness and uncertainty factors that often appear in the supply chain. i ∣E j ) can be expressed by the following formula (9):

[0124]

[0125] In practical applications, the inference algorithm of Bayesian network, such as the maximum expectation (EM) algorithm, can be used to calculate the probabilistic impact of a change in the state of a key node in the supply chain on other related nodes. For example, how will the change in the supplier's supply capacity E1 affect the product production plan E2 and inventory level E4.

[0126] In order to ensure the efficiency of the knowledge graph, the random walk algorithm is further introduced to analyze the importance of each node in the graph. This algorithm can be described by the following Markov Chain model, as shown in formula (10):

[0127] P t+1 =P t *M (10)

[0128] Where P tis the state vector of each node in the graph at time t; M is the transfer matrix, which describes the probability of transferring from one node to other nodes.

[0129] The random walk algorithm can help identify the most critical entities and paths in the supply chain, thereby guiding subsequent optimization decisions. For example, by analyzing the random walk of inventory nodes, it is possible to determine which logistics nodes are most critical to the overall efficiency of the supply chain.

[0130] The third step is to embed the high-dimensional knowledge graph into a low-dimensional space through graph embedding technology, so as to facilitate the use of subsequent machine learning models, retain the topological structure and semantic relationship of the nodes in the original graph, and obtain the knowledge graph. The embedding process can be expressed as formula (11):

[0131] f:V→R d (11)

[0132] Where f is the embedding function, V is the node set, and R d is the target low-dimensional space.

[0133] Figure 4 It is a schematic diagram of the knowledge graph structure, such as Figure 4 As shown, the knowledge graph is finally constructed. Figure 4 These nodes are given different weights according to their importance in the supply chain and are distinguished by node size. Figure 4 The solid lines in the figure represent direct relationships in the supply chain (such as supply relationships, transportation routes, inventory management, etc.), which constitute the core framework of the supply chain; the dotted lines represent the associations formed by introducing cross-industry data sets (such as market demand forecasts, supply chain financial data, environmental impacts, etc.), showing the interaction between the supply chain and the broader economic environment. Through this complex and multi-level map, the key paths and nodes in the supply chain are intuitively reflected, helping companies identify the most important supply chain links, thereby providing strong support for optimizing supply chain management and improving forecasting accuracy.

[0134] 104. Use industrial models to identify key supply chain nodes and node relationships in the knowledge graph and convert them into numerical feature vectors for supply chain prediction.

[0135] Specifically, the first step is to extract key features as follows:

[0136] That is, using graph convolutional networks (GCN) as an industrial model to conduct in-depth analysis of the constructed knowledge graph and extract key features that are highly relevant to supply chain forecasting. The GCN model will help identify the most important information from key nodes in the supply chain (such as suppliers, products, logistics nodes, inventory, etc.) and the complex relationships between them, and convert this information into numerical feature vectors that can be used for supply chain forecasting. The specific steps are:

[0137] First, the graph convolutional network in the industrial model is used to identify the information of each node in the knowledge graph. In the supply chain knowledge graph, each node represents an important element in the supply chain, such as a supplier, product, logistics node, or inventory. The GCN model helps identify the importance of these nodes to supply chain forecasting by analyzing the initial features of each node (such as the supplier's supply capacity, the demand for the product, the transportation efficiency of the logistics node, etc.). The GCN model first inputs the initial feature vector of each node. For example, the feature vector of a supplier node may include the supplier's historical supply volume, delivery time, and product quality. Then, GCN integrates the initial features of these nodes with the information of their neighboring nodes (such as product nodes related to suppliers) through a convolution operation. Node analysis is shown in formula (12):

[0138]

[0139] In the formula, is the feature representation of the supplier node after integrating the neighbor information, N(i) is the set of neighbor nodes of the supplier node, such as its related product nodes, and W (0) It is the weight matrix of the model, and the optimal parameters are obtained through learning.

[0140] Through this process, the GCN model not only analyzes the basic information of the supplier, but also combines the information of other nodes directly related to the supplier (such as product demand, transportation capacity of logistics nodes, etc.), thereby generating a richer node representation.

[0141] Secondly, the neighbor information is used to aggregate the information of each node and its neighbor nodes to generate a new data representation. The process of neighbor information aggregation refers to integrating the information of each node in the supply chain with its related nodes (neighbor nodes) to generate a new data representation. For example, for a supplier node, its neighbor nodes may include all product nodes related to it. The GCN model combines the information of these product nodes with the supplier node’s own information to generate a new, more complex feature representation. For each supplier node, GCN collects information about all product nodes connected to it, such as product market demand fluctuations, production plans, etc. Then, GCN integrates this information with the supplier node’s own information by weighted summation or averaging to generate a new feature representation. The aggregation formula is (13):

[0142]

[0143] In the formula, It is the new feature representation of the supplier node generated after the second layer of convolution operation.

[0144] Through this aggregation operation, the feature vector of the supplier node not only contains its own attributes, but also reflects its relationship and interaction with all related product nodes.

[0145] Then, identify the important features as follows:

[0146] After completing multiple layers of convolution and information aggregation, the GCN model generates the final feature vectors for each node in the supply chain. These feature vectors contain the node's own information and the comprehensive information of all its neighboring nodes. Through model training and optimization, the GCN model can identify which features have the greatest impact on supply chain forecasts. The GCN model minimizes the prediction error through the back propagation algorithm, thereby automatically adjusting the weight parameter W. In the process, the feature vector of each node is continuously optimized, so that the features most relevant to supply chain forecasts are given higher weights, while irrelevant or minor features are weakened. The final goal is formula (14):

[0147] min W ∑ (i，j)∈edges (y ij -f W (h i ,h j )) 2 (14)

[0148] In the formula, y ij It is the behavior actually observed in the supply chain (such as the relationship between supplier supply and product demand). W (h i ,h j) is the model based on the feature vector h i and h j Predicted behavior.

[0149] Through continuous optimization, the GCN model eventually identifies the most important features for supply chain forecasting, namely the supply chain forecasting features. For example, it may be found that the correlation between supplier delivery time fluctuations and product demand fluctuations is the most critical to predicting supply chain performance. These optimized feature vectors will serve as the core input of the supply chain forecasting model.

[0150] Finally, the complex relationships in the knowledge graph are converted into numerical feature vectors as follows:

[0151] After identifying key features through the GCN model, the next step is to convert these complex relationships and information in the graph, namely the supply chain prediction features, into numerical feature vectors. The purpose is to express the complex relationships between nodes (such as the relationship between suppliers and products, and between products and logistics nodes) in a numerical form that can be processed by computers. These numerical values ​​will serve as input to the supply chain prediction model. The conversion process is:

[0152] (1) Node feature integration: After multiple layers of convolution operations, the GCN model generates a feature vector for each node that contains its own and neighbor information. For example, the feature vector of a supplier node contains not only its supply capacity, but also the market demand of the related product nodes and the transportation efficiency of the logistics node.

[0153] (2) Generate numerical feature vectors: The integrated information is finally expressed as a numerical feature vector, as shown in formula (15):

[0154]

[0155] In the formula, each value Represents the performance of a supplier node on specific attributes, such as supply capacity, response time, etc.

[0156] (3) The significance of eigenvectors: Each eigenvector Representing the comprehensive characteristics of node i in the supply chain, the numerical vector simplifies these complex relationships into a set of values ​​that can be used in machine learning models. These vectors provide a unified input format for the model, facilitating subsequent analysis and prediction.

[0157] The results are shown in Table 4:

[0158]

[0159]

[0160] Table 4

[0161] Table 4 is a specific representation of the numerical features. Each node (such as supplier A, product X, logistics node 1, inventory Y) has a corresponding feature value. These feature values ​​represent key attributes such as supply capacity, market demand fluctuations, transportation efficiency, and inventory management stability. These values ​​will be used as inputs to the supply chain forecasting model to more accurately predict various behaviors and trends in the supply chain.

[0162] The second step is to carry out scenario modeling and simulation, as follows:

[0163] After successfully extracting key features and converting them into numerical feature vectors, the next step is to build a supply chain scenario model based on these features and verify the applicability and stability of the model by simulating performance in different supply chain environments. The purpose is to ensure that the model can still make accurate predictions and reasonable decisions in complex and changing supply chain scenarios. The specific steps are as follows:

[0164] First, create a supply chain scenario model:

[0165] Based on the numerical feature vector, a supply chain scenario model is constructed. This model will combine the key nodes in the supply chain (such as suppliers, products, logistics nodes, inventory) and their relationships to simulate the dynamic performance of the supply chain under different conditions. The construction of the scenario model includes the following steps:

[0166] (1) Define input features: The feature vectors extracted from GCN are used as the input of the model. These feature vectors include the supplier’s supply capacity, the market demand fluctuation of the product, the transportation efficiency of the logistics node, etc.

[0167] (2) Constructing a prediction model: This application proposes to use a dynamic prediction model fused with an attention mechanism to predict the future performance of each node in the supply chain, such as Figure 5 The figure shows a schematic diagram of the hybrid model structure. This model combines a multi-head self-attention mechanism with a recurrent neural network (RNN) to capture the complex dependencies and temporal dynamics in the supply chain.

[0168] The input feature vector is processed by multiple attention heads, each of which is responsible for capturing different dependencies between features. Then, the outputs of all attention heads are combined to generate a rich representation. The output of the multi-head self-attention mechanism is represented as (16):

[0169] Z=Concat(head1,head2,…,head h )W o (16)

[0170] Where Z is the output representation of the multi-head self-attention mechanism, and each attention headi is generated by the input features after attention weighting, W o is the output transformation matrix.

[0171] The output of the attention mechanism is then input into the RNN, which processes the time series dynamics of these features and predicts future node states (such as demand, inventory levels, etc.). The RNN processing time series is represented as (17):

[0172] h t =RNN(Z,h t-1 ) (17)

[0173] In the formula, h t is the hidden state of the RNN at time step t. t-μ is the hidden state at the previous time step.

[0174] Finally, the supply chain forecast results are generated by linearly combining the output of the RNN or processing it with a multi-layer perceptron (MLP). The combined model output is expressed as (18):

[0175]

[0176] In the formula, is the final supply chain forecast value. t is the hidden state of the RNN at the last time step.

[0177] Second, simulate the performance under different supply chain environments:

[0178] Use supply chain scenario models to simulate forecast performance under different supply chain environments. Once the supply chain scenario model is established, you can start simulating performance under different supply chain environments. This includes simulating various scenarios:

[0179] Simulation scenario = {supply chain disruption, demand fluctuation, shipping delay}

[0180] Then for each simulated scenario, the model generates a set of predicted values And compare with the actual data to evaluate the applicability of the model. The data in the simulation process are shown in Table (5):

[0181] Situation Type Actual value Predicted value Prediction Error Supply chain disruptions 1000 980 20 Demand Fluctuations 1500 1525 -25 Shipping delays 800 790 10 Extreme events 1200 1150 50

[0182] Table 5

[0183] As can be seen from Table 5, the model is able to maintain a small prediction error under different scenarios, indicating that it has high stability in dealing with various supply chain challenges.

[0184] Finally, verify the applicability and stability of the model:

[0185] Based on the error of the prediction performance, determine the applicability and stability of the supply chain scenario model. Calculate the MSE error of the model in different scenarios to verify whether the model can maintain stable performance in complex and changing supply chain scenarios, as shown in formula (19):

[0186]

[0187] In the formula, n is the number of samples, is the predicted value of the model, y i is the actual observed value.

[0188] Figure 6 is a schematic diagram of MSE error, such as Figure 6 As shown, it can be concluded that the MSE in the supply chain disruption and demand fluctuation scenarios is low, 0.02 and 0.03 respectively, indicating that the model has a high prediction accuracy when dealing with these common supply chain scenarios. In the transportation delay scenario, the MSE is 0.025, showing that the model can still predict the performance of the supply chain well. Even in extreme events (such as extreme weather, sudden policy changes, etc.), the MSE of the model is 0.04, which has increased slightly but remains at a low level, proving the stability of the model in the face of extreme situations.

[0189] Through simulation tests, the high applicability and stability of the model in different scenarios were verified, indicating that it can provide reliable forecasting support for supply chain management, ensuring that the model can show high accuracy and stability in different supply chain scenarios, thereby providing reliable forecasting support for actual supply chain management.

[0190] Furthermore, since the supply chain environment is dynamically changing, market demand and various conditions in the supply chain may change over time. Therefore, based on the above embodiments, this embodiment also includes: introducing a dynamic adjustment and optimization mechanism, using an improved group strategy algorithm to continuously optimize the feature selection and parameter setting of the model, so as to maintain long-term accuracy and stability. Specifically:

[0191] First, collect the actual data of the supply chain and the predicted data in real time for comparison. During the operation of the model, the actual data in the supply chain (such as actual demand, actual inventory level, etc.) will be continuously collected and compared with the predicted results of the model. Including:

[0192] (1) Real-time data collection: Through each node of the supply chain, key data such as actual order volume, inventory changes, and transportation delays are collected in real time.

[0193] (2) Deviation calculation formula: For each time step t, calculate the predicted value With the actual value y tThe deviation Δy between t , as shown in formula (20):

[0194]

[0195] In the formula, y t is the actual value at time step t, is the predicted value at time step t, Δy t is the prediction deviation at time step t.

[0196] Secondly, based on the comparison results, deviation analysis is performed and optimization suggestions are generated. Once deviations are identified, the next step is to analyze the sources of these deviations and generate corresponding optimization suggestions. This includes:

[0197] (1) Deviation analysis: Analyze the performance of the model in a specific supply chain scenario based on the size and trend of the deviation. Focus on analyzing which features or parameters may have caused the deviation. For example, if the deviation is large when predicting the demand for a certain product, it may be because the model fails to fully consider the seasonal fluctuations in the market.

[0198] (2) Optimization suggestion generation formula: Based on the deviation, the feature weight adjustment suggestion Δw is calculated, as shown in formula (21):

[0199] Δw=η*Δy t *x t (twenty one)

[0200] Where η is the learning rate, which controls the step size of weight adjustment, and x t is the input feature vector, and Δw is the adjustment amount of the feature weight.

[0201] Through formula (21), the system can generate recommendations for adjusting feature weights based on the bias and feed these recommendations back to the optimization algorithm to adjust the model in the next step.

[0202] Finally, based on the optimization suggestions, the industrial model is dynamically adjusted and optimized through a real-time feedback loop. The real-time feedback loop and the real-time feedback mechanism ultimately form a closed-loop system to continuously monitor, analyze and optimize the model's prediction results. Including:

[0203] (1) Closed-loop process:

[0204] Monitoring: Continuously monitor the difference between the model's predictions and actual results.

[0205] Analysis: Analyze deviations in real time and generate optimization suggestions.

[0206] Adjust: Adjust feature weights and parameter configurations based on recommendations.

[0207] (2) Feedback closed loop formula (22):

[0208] New_Parameters=Old_Parameters+Δw (22)

[0209] Where New_Paramete is the parameter value of the model after receiving feedback and making adjustments. The new parameter is obtained by adding the old parameter and the adjustment amount Δw. Old_Parameters is the parameter value currently used by the model. Before feedback, the model uses these parameters to make predictions. Δw is the feature weight adjustment generated based on real-time feedback, which represents the correction made by the model based on actual data and prediction deviation. The specific calculation method is usually based on the prediction deviation Δy t and the corresponding eigenvector x t , and then multiply it by the learning rate η to control the amplitude of the adjustment.

[0210] Through a closed-loop system, the model is able to continuously adjust and optimize itself to adapt to changing conditions in the supply chain, ensuring the accuracy and timeliness of forecasts.

[0211] The supply chain forecasting method of the present invention has the following advantages:

[0212] Improve forecasting accuracy: By introducing knowledge graphs and industrial model algorithms, the complex relationships between nodes in the supply chain can be better captured, thereby significantly improving the accuracy of supply chain forecasts.

[0213] Strong dynamic adaptability: Combining dynamic adjustment and optimization mechanisms, the model can automatically adjust parameter configurations based on real-time feedback, has strong adaptive capabilities, and can cope with dynamic changes in the supply chain environment.

[0214] Optimize feature selection: Through improved feature combination and connection optimization methods, it is possible to flexibly select the features most relevant to supply chain forecasting, making up for the shortcomings in feature selection.

[0215] Integrate multi-industry data: By building a knowledge graph covering multi-industry data, data from different links in the supply chain can be effectively integrated, providing more comprehensive data support for supply chain forecasting.

[0216] Based on the same general inventive concept, the present invention also protects a supply chain prediction device based on a knowledge graph and an industrial model. The supply chain prediction device based on a knowledge graph and an industrial model provided by the present invention is described below. The supply chain prediction device based on a knowledge graph and an industrial model described below and the supply chain prediction method based on a knowledge graph and an industrial model described above can be referenced to each other.

[0217] Figure 7It is a structural diagram of a supply chain prediction device based on knowledge graph and industrial model provided in this embodiment.

[0218] like Figure 7 As shown, the supply chain prediction device based on knowledge graph and industrial model provided in this embodiment includes:

[0219] The collection module 701 is used to collect the features required for supply chain forecasting, including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data;

[0220] The analysis module 702 is used to perform data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain target features;

[0221] A construction module 703 is used to identify key entities and relationships in the target feature and construct a knowledge graph based on the key entities and relationships;

[0222] The prediction module 704 is used to use the industrial model to identify the key supply chain nodes and node relationships in the knowledge graph, and convert them into numerical feature vectors for supply chain prediction to perform supply chain prediction.

[0223] Figure 8 It is a schematic diagram of the structure of the electronic device provided in this embodiment.

[0224] like Figure 8 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the supply chain prediction method based on the knowledge graph and the industrial model, the method comprising: collecting the required features for supply chain prediction, the required features including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data; performing data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain the target features; identifying the key entities and relationships in the target features, and constructing the knowledge graph based on the key entities and relationships; using the industrial model to identify the key supply chain nodes and node relationships in the knowledge graph, and converting them into numerical feature vectors for supply chain prediction to perform supply chain prediction.

[0225] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0226] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the supply chain prediction method based on knowledge graphs and industrial models provided by the above methods, and the method includes: collecting features required for supply chain prediction, and the required features include historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data; performing data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, and the logistics data and warehousing data to obtain target features; identifying key entities and relationships in the target features, and constructing a knowledge graph based on the key entities and relationships; using industrial models to identify key supply chain nodes and node relationships in the knowledge graph, and converting them into numerical feature vectors for supply chain prediction to perform supply chain prediction.

[0227] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the supply chain forecasting method based on knowledge graphs and industrial models provided by the above-mentioned methods, the method comprising: collecting features required for supply chain forecasting, the required features including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data and warehousing data; performing data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain target features; identifying key entities and relationships in the target features, and constructing a knowledge graph based on the key entities and relationships; using an industrial model to identify key supply chain nodes and node relationships in the knowledge graph, and converting them into numerical feature vectors for supply chain forecasting to perform supply chain forecasting.

[0228] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0229] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0230] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 invention.

Claims

1. A supply chain prediction method based on knowledge graph and industrial model, characterized in that: include: Collecting features required for supply chain forecasting, including historical order data and production plan data, industrial product procurement data and centralized procurement information, logistics data, and warehousing data; Performing data cleaning, standardization and correlation analysis on the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain target features; Identify key entities and relationships in the target feature, and construct a knowledge graph based on the key entities and relationships; The industrial model is used to identify key supply chain nodes and node relationships in the knowledge graph, and is converted into numerical feature vectors for supply chain forecasting to perform supply chain forecasting.

2. The supply chain prediction method based on knowledge graph and industrial model according to claim 1 is characterized in that: Also includes: Collecting actual data and forecast data of the supply chain in real time for comparison; Based on the results of the comparison, deviation analysis is performed and optimization suggestions are generated; Based on the optimization suggestions, the industrial model is dynamically adjusted and optimized through a real-time feedback loop.

3. The supply chain prediction method based on knowledge graph and industrial model according to claim 1 is characterized in that: The historical order data and production plan data collected from the features required for supply chain forecasting include: Obtain historical order files and production plan files through the enterprise resource planning platform and manufacturing execution platform; Through data aggregation, the historical order files and the production plan files are weightedly aggregated according to time series and dimensions to obtain historical order data and production plan data.

4. The supply chain prediction method based on knowledge graph and industrial model according to claim 1 is characterized in that: The industrial product procurement data and centralized procurement information in the collection of features required for supply chain forecasting include: Obtain supplier data and material data through supply chain management system and procurement management system; Input the supplier data and the material data into the supplier and material association matrix, and output industrial product procurement data and centralized procurement information.

5. The supply chain prediction method based on knowledge graph and industrial model according to claim 1 is characterized in that: The logistics data and warehousing data collected from the features required for supply chain forecasting include: Obtain logistics information and warehousing information through logistics management system and warehousing management system; The logistics information and the warehousing information are input into the logistics and warehousing efficiency analysis model, and the efficiency index is output as logistics data and warehousing data.

6. The supply chain prediction method based on knowledge graph and industrial model according to claim 1 is characterized in that: The correlation analysis of the historical order data and production plan data, the industrial product procurement data and centralized procurement information, the logistics data and warehousing data to obtain target features includes: Represent the historical order data and production plan data, the industrial product procurement data and centralized procurement information, and the logistics data and warehousing data after data cleaning and standardization as corresponding features respectively; Determining the correlation between each of the features using a weighted Pearson correlation coefficient; Based on the correlation between the various features, the target feature is determined.

7. The supply chain prediction method based on knowledge graph and industrial model according to claim 1 is characterized in that: The identifying key entities and relationships in the target features and constructing a knowledge graph based on the key entities and relationships includes: Based on the target features, identifying entities in the supply chain, the entities including suppliers, products, and logistics nodes; Determine the relationship between the supplier, the product and the logistics node, and obtain the upstream and downstream relationship of the supply chain, the logistics path and transportation relationship, and the inventory management relationship; The suppliers, the products, the logistics nodes, the upstream and downstream relationships of the supply chain, the logistics routes and transportation relationships, and the inventory management relationships are integrated into a multi-level knowledge graph, and a knowledge graph is constructed through structured integration, mathematical modeling, and graph reasoning.

8. The supply chain prediction method based on knowledge graph and industrial model according to claim 7 is characterized in that: The supplier, the product, the logistics node, the upstream and downstream relationship of the supply chain, the logistics path and transportation relationship and the inventory management relationship are integrated into a multi-level knowledge graph, and a knowledge graph is constructed through structured integration, mathematical modeling and graph reasoning, including: The supplier, the product, the logistics node, the upstream and downstream relationship of the supply chain, the logistics path and transportation relationship and the inventory management relationship are formally represented by means of graph theory; Introducing cross-industry data sets, and modeling the probabilistic dependencies between the entities after the formal representation through Bayesian networks, to obtain a high-dimensional knowledge graph; Through graph embedding, the high-dimensional knowledge graph is embedded into a low-dimensional space, retaining the topological structure and semantic relationship of the nodes to obtain a knowledge graph.

9. The supply chain prediction method based on knowledge graph and industrial model according to any one of claims 1 to 8, characterized in that: The use of industrial models to identify key supply chain nodes and node relationships in the knowledge graph and convert them into numerical feature vectors for supply chain prediction includes: Using the graph convolutional network in the industrial model to identify the information of each node in the knowledge graph; Aggregating information of each node and its neighboring nodes using neighbor information to generate a new data representation; Based on the data representation, identifying supply chain prediction features using the graph convolutional network; The supply chain prediction features are converted into numerical feature vectors.

10. The supply chain prediction method based on knowledge graph and industrial model according to claim 9 is characterized in that: After converting the supply chain prediction features into numerical feature vectors, the method further includes: Based on the numerical feature vector, construct a supply chain scenario model; Using the supply chain scenario model, simulate forecasting performance under different supply chain environments; Based on the error of the forecast performance, the applicability and stability of the supply chain scenario model are determined.

Images