Textile raw material collecting and selling matching method and system based on semantic data

Abstract

The invention relates to the technical field of textile raw material collection and sale matching, and provides a semantic data-based textile raw material collection and sale matching method and system. The method comprises the following specific implementation steps: collecting a textile raw material purchase record, a sales record, inventory data and market data within a preset time, determining an association edge weight between different node attributes by using a semantic rule, and further constructing a textile raw material collection and sales knowledge graph; thirdly, constructing a textile raw material purchasing prediction model, and predicting a future purchasing plan by inputting multi-dimensional textile raw material time sequence data and a purchasing adjacency matrix; and finally, sales records and inventory data are monitored in real time, purchase prediction deviation is calculated, whether a prompt needs to be sent to purchase management personnel is judged, and the acquisition and sales knowledge graph and the purchase prediction model are updated in time, so that the accuracy of textile raw material acquisition and sales matching and the efficiency of supply chain management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of textile raw material procurement and marketing matching, and in particular to a textile raw material procurement and marketing matching method and system based on semantic data. Background Art

[0002] In the textile industry, supply chain management, as a key link to ensure smooth production and timely response to market demand, covers the entire process from raw material procurement to finished product delivery. With globalization and the diversification of market demand, the supply chain of the textile industry faces increasingly complex challenges, including fluctuations in raw material prices, inventory management difficulties, and the need to respond quickly to market changes, which leads to excess or shortage of inventory during the production process. In order to improve the efficiency of the supply chain and reduce costs, it is necessary to optimize the collaborative work of each link to ensure the efficient use of resources. Among them, procurement and sales matching, as a key link connecting procurement and sales, can improve the operational efficiency of textile enterprises by accurately analyzing market demand and the supply capacity of textile enterprises, optimizing procurement strategies and sales plans.

[0003] At present, the main methods for matching procurement and sales of textile raw materials are: ① procurement demand forecast based on sales data; ② procurement plan driven by market demand; ③ procurement plan regulation based on inventory management. The above methods design and adjust the procurement plan through a single influencing element, ignoring the interactive relationship between market demand, inventory data, sales data and procurement data, resulting in errors in procurement decisions, leading to an imbalance in the supply and demand of textile raw materials, and further affecting the benefits of textile enterprises.

[0004] Therefore, a textile raw material procurement and sales matching method and system based on semantic data is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a textile raw material procurement and marketing matching method and system based on semantic data. By collecting textile raw material procurement records, sales records, inventory data and market data within a predetermined time, and using semantic rules to determine the weight of the associated edges between different node attributes, a textile raw material procurement and marketing knowledge graph is constructed; then, a textile raw material procurement prediction model is constructed, and future procurement plans are predicted by inputting multi-dimensional textile raw material time series data and procurement adjacency matrix; finally, sales records and inventory data are monitored in real time, procurement prediction deviations are calculated, and it is determined whether it is necessary to issue prompts to procurement management personnel, and the procurement and marketing knowledge graph and procurement prediction model are updated in time, thereby improving the accuracy of textile raw material procurement and marketing matching and the efficiency of supply chain management.

[0006] A textile raw material procurement and sales matching method based on semantic data, comprising: Obtain textile raw material purchase records within a predetermined period of time, and collect textile raw material sales records, textile raw material inventory data and textile market data within the purchase cycle; Determine the attribute edge weights of the associated edges between different node attributes through the semantic rules between the textile raw material procurement records, the textile raw material sales records, the textile raw material inventory data and the textile market data; construct a textile raw material procurement and sales knowledge graph based on the node attributes and the associated edges; Constructing a textile raw material procurement forecasting model, inputting the multidimensional textile raw material time series data and the procurement adjacency matrix in the previous procurement cycle; extracting the multidimensional textile raw material features of the multidimensional textile raw material time series data through the time series feature extraction layer; performing feature exchange on the procurement adjacency matrix and the multidimensional textile raw material features according to the multidimensional feature interaction layer, and predicting future procurement plans in combination with the procurement forecasting layer; The textile raw material sales records and the textile raw material inventory data are monitored in real time, the purchase forecast deviation is calculated, and it is determined whether to issue a prompt to the purchase management personnel and update the textile raw material purchase and sales knowledge graph and the textile raw material purchase forecast model.

[0007] Preferably, the textile raw material purchase record, the textile raw material sales record, the textile raw material inventory data and the textile market data at the same purchase time point are obtained to construct the textile raw material purchase and sales knowledge graph; The textile raw material procurement and sales knowledge graph includes textile raw material data nodes and the associated edges; the textile raw material data nodes include procurement nodes, sales nodes, inventory nodes and market demand nodes, and each of the textile raw material data nodes includes several node attributes; the associated edges are used to connect the node attributes, indicating the data interaction between two node attributes; for two different node attributes, the attribute edge weight of the associated edge is set in combination with the attribute data sequence and the corresponding semantic rules between the textile raw material data nodes.

[0008] Preferably, the specific process of setting the attribute edge weight of the associated edge includes: Constructing the semantic rules between the textile raw material data nodes, including a first semantic rule between the procurement node and the sales node, a second semantic rule between the inventory node and the sales node, and a third semantic rule between the procurement node and the market demand node; According to the attribute data sequence of the node attributes, the time series difference weight is calculated; the attribute correlation weight between different node attributes is calculated through the cosine similarity function; according to the semantic rules between the nodes corresponding to different node attributes, the time series difference weight and the attribute correlation weight are combined to obtain the attribute edge weight of the associated edge.

[0009] Preferably, the textile raw material procurement prediction model includes the temporal feature extraction layer, the multi-dimensional feature interaction layer and the procurement prediction layer; the specific prediction process includes: Performing data cleaning on the textile raw material sales records, the textile raw material inventory data and the textile market data in the previous procurement cycle to obtain the multi-dimensional textile raw material time series data; obtaining the adjacency matrix of the textile raw material procurement and sales knowledge graph, and performing normalization processing to obtain the procurement adjacency matrix; Inputting the multi-dimensional textile raw material time series data into the time series feature extraction layer, and obtaining the multi-dimensional textile raw material features through a convolutional neural network, including the historical sales quantity, historical sales season, historical inventory quantity and market price change of the textile raw materials; Input the multidimensional textile raw material features and the procurement adjacency matrix into the multidimensional feature interaction layer, capture the data change law of the multidimensional textile raw material features through Bi-LSTM, and generate textile raw material time series features; combine the graph convolutional neural network and the weighted fusion mechanism to perform feature interaction on the textile raw material time series features and the procurement adjacency matrix to obtain multidimensional textile raw material interaction features; The procurement prediction layer predicts the multi-dimensional textile raw material interaction characteristics according to the fully connected network and the prediction head, and outputs the future procurement plan.

[0010] Preferably, the specific process of determining whether to issue a prompt to the purchasing manager and updating the textile raw material procurement and sales knowledge graph and the textile raw material procurement prediction model includes: The textile raw material sales records and the textile raw material inventory data within the procurement cycle are monitored in real time, and the procurement forecast deviation is calculated; if the procurement forecast deviation is greater than or equal to a predetermined deviation threshold, a prompt is issued to the procurement management personnel; through the procurement forecast deviation, the ideal textile raw material sales records and the ideal textile raw material inventory data are determined, and the textile raw material procurement and sales knowledge graph and the textile raw material procurement forecast model are updated.

[0011] A textile raw material procurement and sales matching system based on semantic data, comprising: The procurement and sales data collection module obtains the textile raw material procurement records within a predetermined time, and collects the textile raw material sales records, textile raw material inventory data and textile market data within the procurement cycle; A procurement and sales data association module determines the attribute edge weights of the associated edges between different node attributes through the semantic rules between the textile raw material procurement records, the textile raw material sales records, the textile raw material inventory data and the textile market data; and constructs a textile raw material procurement and sales knowledge graph based on the node attributes and the associated edges; The procurement plan forecasting module constructs a textile raw material procurement forecasting model, inputs the multidimensional textile raw material time series data and the procurement adjacency matrix in the previous procurement cycle; extracts the multidimensional textile raw material features of the multidimensional textile raw material time series data through the time series feature extraction layer; performs feature exchange on the procurement adjacency matrix and the multidimensional textile raw material features according to the multidimensional feature interaction layer, and predicts the future procurement plan in combination with the procurement forecasting layer; The prompt and model update module monitors the textile raw material sales records and the textile raw material inventory data in real time, calculates the purchase forecast deviation, determines whether to issue prompts to the purchase management personnel, and updates the textile raw material procurement and sales knowledge graph and the textile raw material procurement forecast model.

[0012] Preferably, the procurement and sales data association module obtains the textile raw material procurement records, the textile raw material sales records, the textile raw material inventory data and the textile market data at the same procurement time point to construct the textile raw material procurement and sales knowledge graph; The textile raw material procurement and sales knowledge graph includes textile raw material data nodes and the associated edges; the textile raw material data nodes include procurement nodes, sales nodes, inventory nodes and market demand nodes, and each of the textile raw material data nodes includes several node attributes; the associated edges are used to connect the node attributes, indicating the data interaction between two node attributes; for two different node attributes, the attribute edge weight of the associated edge is set in combination with the attribute data sequence and the corresponding semantic rules between the textile raw material data nodes.

[0013] Preferably, the specific process of setting the attribute edge weight of the associated edge includes: Constructing the semantic rules between the textile raw material data nodes, including a first semantic rule between the procurement node and the sales node, a second semantic rule between the inventory node and the sales node, and a third semantic rule between the procurement node and the market demand node; According to the attribute data sequence of the node attributes, the time series difference weight is calculated; the attribute correlation weight between different node attributes is calculated through the cosine similarity function; according to the semantic rules between the nodes corresponding to different node attributes, the time series difference weight and the attribute correlation weight are combined to obtain the attribute edge weight of the associated edge.

[0014] Preferably, the specific forecasting process of the procurement plan forecasting module includes: The multidimensional textile raw material time series data is obtained according to the textile raw material sales records, the textile raw material inventory data and the textile market data in the previous procurement cycle; the procurement adjacency matrix is ​​obtained through the adjacency matrix of the textile raw material procurement and sales knowledge graph; the multidimensional textile raw material time series data is input into the time series feature extraction layer to obtain the multidimensional textile raw material features; the multidimensional textile raw material features and the procurement adjacency matrix are input into the multidimensional feature interaction layer to obtain the multidimensional textile raw material interaction features; the procurement prediction layer predicts the multidimensional textile raw material interaction features according to the fully connected network and the prediction head, and outputs the future procurement plan.

[0015] Preferably, the specific implementation process of the prompt and model update module includes: The textile raw material sales records and the textile raw material inventory data within the procurement cycle are monitored in real time, and the procurement forecast deviation is calculated; if the procurement forecast deviation is greater than or equal to a predetermined deviation threshold, a prompt is issued to the procurement management personnel; through the procurement forecast deviation, the ideal textile raw material sales records and the ideal textile raw material inventory data are determined, and the textile raw material procurement and sales knowledge graph and the textile raw material procurement forecast model are updated.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The textile raw material procurement and sales knowledge graph proposed in the present invention can comprehensively and accurately reflect the association relationship between different node attributes by combining the semantic rules between textile raw material procurement records, sales records, inventory data and market data, thereby providing more accurate data support for procurement decisions. By optimizing the weights of node attributes and associated edges, the accuracy of textile raw material procurement and sales matching can be effectively improved, ensuring that procurement plans can closely match changes in market demand.

[0017] 2. The textile raw material procurement prediction model proposed in the present invention, combined with the time series feature extraction layer and the feature interaction layer, can deeply mine the historical sales and inventory data of textile raw materials and predict future procurement needs; through time series feature learning and feature interaction learning, procurement forecasts are more timely and accurate, effectively avoiding the risk of excess or shortage inventory and improving the efficiency of procurement and sales matching.

[0018] 3. The present invention can calculate the purchase forecast deviation by real-time monitoring of sales records and inventory data, and promptly issue a warning to the purchase management personnel when the deviation exceeds a predetermined threshold. This process ensures that the purchase and sales matching always maintains high accuracy in a dynamically changing market environment. At the same time, the textile raw material purchase and sales knowledge map and purchase forecast model are automatically updated according to feedback information, so that they are continuously optimized and the accuracy and response speed of purchase decisions are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of a textile raw material procurement and sales matching method based on semantic data provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of the textile raw material procurement and sales knowledge graph provided for the embodiment of the present invention; Figure 3 A system structure diagram of a textile raw material procurement and sales matching system based on semantic data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0021] In the past, the procurement plans of textile enterprises were formulated based on historical data and experience. However, due to sudden changes in market demand, lagging sales data and inaccurate inventory information, the procurement plans could not reflect the actual demand in a timely manner, resulting in inventory backlog or shortage, which in turn affected production efficiency and capital flow. Therefore, textile enterprises can use procurement and sales matching technology to optimize the coordination between textile raw material procurement and sales, improve supply chain efficiency, avoid inventory backlog or production line stagnation due to excessive or insufficient procurement of textile raw materials, and thus reduce operating costs.

[0022] In order to illustrate the effectiveness of the method and system for textile raw material procurement and marketing matching based on semantic data proposed by the present invention, the specific implementation methods of the textile raw material procurement and marketing matching process will be described in detail in combination with the accompanying drawings of this embodiment and the following two embodiments.

[0023] Embodiment 1 Textile company A is mainly engaged in the production and sales of textiles and the wholesale and retail of textile raw materials. It needs to determine the actual demand for different textile raw materials in order to formulate a reasonable procurement plan. In order to improve the accuracy of procurement plans and reduce resource waste, Textile Company A introduced a textile raw material procurement and sales matching method based on semantic data. Figure 1, which is a flowchart of a textile raw material procurement and marketing matching method based on semantic data provided by an embodiment of the present application. The specific steps of the textile raw material procurement and marketing matching method include: S1, obtaining textile raw material procurement records within a predetermined time, and collecting textile raw material sales records, textile raw material inventory data and textile market data within the procurement cycle; S2, constructing a textile raw material procurement and marketing knowledge graph, and determining the attribute edge weights of the associated edges between different node attributes through semantic rules; S3, constructing a textile raw material procurement prediction model to predict future procurement plans; S4, real-time monitoring of textile raw material sales records and textile raw material inventory data, determining whether to issue prompts to procurement management personnel and updating the textile raw material procurement and marketing knowledge graph and the textile raw material procurement prediction model.

[0024] Corresponding to the above step S1, the textile raw material procurement records of textile enterprise A within a predetermined time are obtained, and the textile raw material sales records, textile raw material inventory data and textile market data within the procurement cycle are collected; wherein the textile market data includes market area, textile type and textile unit price.

[0025] Specifically, referring to Tables 1 to 3, Table 1 lists some examples of textile raw material procurement records, Table 2 lists some examples of textile raw material sales records, and Table 3 lists some examples of textile raw material inventory data; wherein, textile raw material procurement records include purchase order number, purchase type, purchase quantity, purchase unit price, purchase date and transportation cost; textile raw material sales records include sales order number, sales type, sales quantity, sales area, sales method, sales date and sales amount; textile raw material inventory data includes textile type, inventory quantity, warehousing date and shelf life.

[0026] Table 1. Partial example of textile raw material procurement records of A textile enterprise

[0027] Table 2. Some examples of textile raw material sales records of A textile enterprise

[0028] Table 3. Some examples of textile raw material inventory data of textile enterprise A

[0029] Further, corresponding to the above step S2, the textile raw material purchase records, textile raw material sales records, textile raw material inventory data and textile market data at the same purchase time point are obtained to construct a textile raw material purchase and sales knowledge graph; refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of the knowledge graph for textile raw material procurement and sales; The knowledge graph of textile raw material procurement and sales includes textile raw material data nodes and associated edges; Figure 2In the figure, the solid ellipse represents the textile raw material data node, the dotted circle represents the node attribute of the textile raw material data node, and the arrow line represents the semantic rules between different textile raw material data nodes; the textile raw material data nodes include procurement nodes, sales nodes, inventory nodes and market demand nodes, and each textile raw material data node includes several node attributes; the association edge is used to connect the node attributes, indicating the data interaction between two node attributes; for two different node attributes, the attribute edge weight of the association edge is set in combination with the attribute data sequence and the semantic rules between the corresponding textile raw material data nodes.

[0030] Specifically, the node attributes of the textile raw material data nodes in the textile raw material procurement and sales knowledge graph are obtained by screening textile raw material procurement records, textile raw material sales records, textile raw material inventory data and textile market data; for example, the sales amount in the textile raw material sales record is screened out, that is, the corresponding sales node does not have the node attribute of sales amount.

[0031] Optionally, the semantic information in textile raw material procurement records, textile raw material sales records, textile raw material inventory data and textile market data is converted into specific data through unique hot encoding to facilitate the construction of textile raw material procurement and sales knowledge graph and the subsequent establishment of textile raw material procurement prediction model.

[0032] Optionally, in the textile raw material sales records and textile raw material inventory data, textile raw materials such as non-woven fabrics and curtain fabrics are processed from purchased textile raw materials (such as pure cotton yarn, polyester DTY and viscose staple fiber). In the process of constructing the textile raw material procurement and sales knowledge graph, it is necessary to convert the processed textile raw material data into the corresponding purchased textile raw material data based on the proportion of purchased textile raw materials used by the processed textile raw materials in the textile raw material sales records and textile raw material inventory data; for example, the curtain fabric is composed of 65% polyester DTY and 35% pure cotton yarn, then the sales quantity of the curtain fabric in Table 2 should be converted into the sales quantity of polyester DTY as 0.338 tons and the sales quantity of pure cotton yarn as 0.182 tons.

[0033] By constructing a knowledge graph of textile raw material procurement and sales, and conducting a comprehensive analysis of procurement, sales, inventory data and market demand data, the real-time supply and demand relationship of various textile raw materials can be accurately reflected. By establishing textile raw material data nodes and associated edges, and setting edge weights in combination with attribute data sequences and semantic rules, the associations between various links can be captured more effectively, providing data support for subsequent textile raw material procurement and sales forecasting models, thereby significantly improving the accuracy of textile fabric procurement and sales matching, optimizing procurement decisions, and reducing resource waste and operational risks in the supply chain.

[0034] Furthermore, the specific process of setting the attribute edge weight of the associated edge includes: Constructing semantic rules between textile raw material data nodes, including the first semantic rule between the procurement node and the sales node, the second semantic rule between the inventory node and the sales node, and the third semantic rule between the procurement node and the market demand node; Optionally, the first semantic rule is that if the sales quantity of the sales node increases, the purchase quantity corresponding to the purchase node increases; the second semantic rule is that if the inventory of the inventory node is greater than a predetermined inventory threshold, the sales volume can increase according to market demand; the third semantic rule is that if the unit price of the textile at the market demand node increases, the purchase quantity corresponding to the purchase node increases; the specific calculation formula is expressed as: ; in, Indicates the first Rule weights of semantic rules; Indicates the node attributes of the B node Through the Function of semantic rules The calculated value exceeds the predetermined threshold , the corresponding rule weight coefficient is 0.7; for example, in the first semantic rule, when the sales quantity growth rate of the sales node is Greater than a predetermined threshold When , the rule weight between the purchasing node and the sales node is 0.7; Indicates In semantic rules Does not exceed the predetermined threshold When , the corresponding rule weight coefficient is 0.2; The first Node attributes; According to the node attributes The attribute data sequence of , calculate the time series difference weight, which represents the numerical difference of node attributes Changes with time The ratio of the node attributes The rate of change of node attributes is calculated by cosine similarity function. and The attribute correlation weight between According to the semantic rules between nodes corresponding to different node attributes, combined with the time difference weight and attribute correlation weight, the attribute edge weight of the associated edge is obtained. The specific calculation formula is: ; in, Represents node attributes and The attribute edge weight between them; Represents node attributes The time difference weight of Represents node attributes and The attribute correlation weight between them; Represents node attributes and The rule weights between the corresponding textile raw material data nodes; , and Indicates the influencing factors with different weights, , and The values ​​are all in the interval Inner .

[0035] By combining attribute data sequences and semantic rules, setting the attribute edge weights of the associated edges can achieve more accurate textile raw material procurement and sales forecasts and market demand analysis. Among them, the constructed semantic rules help capture the intrinsic relationships between different data nodes, such as the association between sales nodes and procurement nodes, the dynamic adjustment of inventory nodes and sales nodes, and the price-driven effect between procurement nodes and market demand, which can effectively reflect the changes in node attributes in different scenarios. In addition, the calculation of time series difference weights and attribute correlation weights can quantify the speed of node attribute changes and the similarity between attributes, further improving the adaptability to dynamic market environments.

[0036] Furthermore, corresponding Figure 1 In step S3, a textile raw material procurement prediction model is constructed to predict future procurement plans. The textile raw material procurement prediction model includes a time series feature extraction layer, a multi-dimensional feature interaction layer, and a procurement prediction layer. The specific prediction process includes: Data cleaning is performed on the textile raw material sales records, textile raw material inventory data and textile market data in the previous procurement cycle to obtain multi-dimensional textile raw material time series data; the adjacency matrix of the textile raw material procurement and sales knowledge graph is obtained and normalized to obtain the procurement adjacency matrix; The multi-dimensional textile raw material time series data is input into the time series feature extraction layer, and the multi-dimensional textile raw material features are obtained through the convolutional neural network, including the historical sales quantity, historical sales season, historical inventory quantity and market price change of textile raw materials; Input the multidimensional textile raw material characteristics and the procurement adjacency matrix to the multidimensional feature interaction layer, capture the data change law of the multidimensional textile raw material characteristics through Bi-LSTM, and generate the textile raw material time series characteristics; combine the graph convolutional neural network and the weighted fusion mechanism to perform feature interaction on the textile raw material time series characteristics and the procurement adjacency matrix to obtain the multidimensional textile raw material interaction characteristics. The specific calculation formula is: ; in, Representing multi-dimensional textile raw material interaction characteristics; represents a Bi-LSTM network; Represents multi-dimensional textile raw material characteristics; Represents a graph convolutional neural network; represents the procurement adjacency matrix; Represents the weight coefficient, by taking the time series characteristics of textile raw materials Input to the channel attention mechanism; The procurement prediction layer predicts the interactive features of multi-dimensional textile raw materials based on the fully connected network and the prediction head, and outputs the future procurement plan, including the procurement type and quantity at the next procurement time point.

[0037] By combining the time series feature extraction layer, the multidimensional feature interaction layer and the procurement forecasting layer, the textile raw material procurement forecasting model can effectively capture the complex patterns in the textile raw material sales, inventory and market data, and improve the modeling capabilities of historical sales, inventory changes and market trends, thereby optimizing procurement plans, reducing inventory backlogs or supply shortages caused by forecast deviations, and improving the overall supply chain operational efficiency and responsiveness.

[0038] Furthermore, real-time monitoring of textile raw material sales records and textile raw material inventory data is performed to calculate procurement forecast deviations, determine whether to issue prompts to procurement management personnel, and update the textile raw material procurement and sales knowledge graph and procurement forecast model; Figure 1 The S4 step includes: Real-time monitoring of textile raw material sales records and textile raw material inventory data within the procurement cycle, through , calculate the purchase forecast deviation; For the The purchase forecast deviation of various textile raw materials; For the The purchase quantity of the textile raw materials; For the Sales volume of textile raw materials; If the purchase forecast deviation is greater than or equal to the predetermined deviation threshold, a prompt is issued to the purchase management personnel; the ideal sales record of textile raw materials and the ideal inventory data of textile raw materials are determined through the purchase forecast deviation, and the textile raw material procurement and sales knowledge graph and the textile raw material purchase forecast model are updated; among them, the ideal sales record of textile raw materials refers to the sales record without considering whether the inventory is insufficient; the ideal inventory data of textile raw materials refers to the inventory data determined according to the ideal sales record of textile raw materials without considering the inventory capacity and the actual inventory quantity of textile fabrics, that is, when the inventory is insufficient, the corresponding textile raw material quantity can be expressed in the form of a negative number.

[0039] By monitoring the sales records and inventory data of textile raw materials in real time and calculating the deviation of procurement forecasts, the deviation between procurement forecasts and actual conditions can be effectively identified. By continuously updating the knowledge graph of textile raw material procurement and sales and the procurement forecast model, not only the accuracy of the forecast is improved, but also the matching of procurement and sales plans can be optimized, further improving the accuracy of textile fabric procurement and sales, reducing the risk of inventory backlogs and shortages, and thus improving the efficiency and responsiveness of the overall supply chain.

[0040] The embodiment of the present application effectively improves the accuracy and response speed of procurement decisions through the proposed textile raw material procurement and sales knowledge map and procurement forecast model. By combining procurement records, sales data, inventory information and market demand, the knowledge map can fully reflect the correlation between the attributes of each node, ensuring that the procurement plan is highly consistent with changes in market demand. The deep learning ability of the procurement forecast model makes procurement forecasts more timely and accurate, reduces the risk of excess or shortage inventory, and thus improves procurement and sales matching efficiency. In addition, real-time monitoring of sales and inventory data, combined with forecast deviation calculation, can issue prompts in time when procurement forecasts deviate, ensuring that procurement managers can make adjustments quickly, and automatically optimize knowledge maps and forecast models based on feedback, so that procurement decisions are continuously optimized under changing market conditions, improving the intelligence level of supply chain management, and further improving procurement efficiency.

[0041] Embodiment 2 The embodiment of the present application applies a textile raw material procurement and sales matching system based on semantic data proposed by the present invention to the procurement department of textile enterprise A to improve procurement efficiency and the accuracy of procurement plans. Figure 3 A system structure diagram of a textile raw material procurement and marketing matching system based on semantic data provided in an embodiment of the present application, the textile raw material procurement and marketing matching system includes a procurement and marketing data acquisition module, a procurement and marketing data association module, a procurement plan prediction module and a prompt and model update module.

[0042] Specifically, the purchase and sales data collection module obtains the textile raw material purchase records within a predetermined time, and collects the textile raw material sales records, textile raw material inventory data and textile market data within the purchase cycle.

[0043] Furthermore, the procurement and sales data association module obtains textile raw material procurement records, textile raw material sales records, textile raw material inventory data and textile market data at the same procurement time point to construct a textile raw material procurement and sales knowledge graph; The knowledge graph of textile raw material procurement and sales includes textile raw material data nodes and associated edges; textile raw material data nodes include procurement nodes, sales nodes, inventory nodes and market demand nodes, and each textile raw material data node includes several node attributes; associated edges are used to connect node attributes and represent the data interaction between two node attributes; for two different node attributes, the attribute edge weight of the associated edge is set in combination with the attribute data sequence and the semantic rules between the corresponding textile raw material data nodes.

[0044] Furthermore, the specific process of setting the attribute edge weight of the associated edge includes: Constructing semantic rules between textile raw material data nodes, including the first semantic rule between the procurement node and the sales node, the second semantic rule between the inventory node and the sales node, and the third semantic rule between the procurement node and the market demand node; According to the attribute data sequence of the node attributes, the time series difference weight is calculated; the attribute correlation weight between different node attributes is calculated through the cosine similarity function; according to the semantic rules between the nodes corresponding to different node attributes, the time series difference weight and the attribute correlation weight are combined to obtain the attribute edge weight of the associated edge.

[0045] Furthermore, the procurement plan prediction module constructs a textile raw material procurement prediction model, inputs the multi-dimensional textile raw material time series data and the procurement adjacency matrix in the above-mentioned procurement cycle, and predicts the future procurement plan; the specific prediction process includes: Data cleaning is performed on the textile raw material sales records, textile raw material inventory data and textile market data in the previous procurement cycle to obtain multi-dimensional textile raw material time series data; the adjacency matrix of the textile raw material procurement and sales knowledge graph is obtained and normalized to obtain the procurement adjacency matrix; The multi-dimensional textile raw material time series data is input into the time series feature extraction layer, and the multi-dimensional textile raw material features are obtained through the convolutional neural network, including the historical sales quantity, historical sales season, historical inventory quantity and market price change of textile raw materials; Input the multidimensional textile raw material characteristics and the procurement adjacency matrix to the multidimensional feature interaction layer, capture the data change law of the multidimensional textile raw material characteristics through Bi-LSTM, and generate the textile raw material time series characteristics; combine the graph convolutional neural network and the weighted fusion mechanism to perform feature interaction on the textile raw material time series characteristics and the procurement adjacency matrix to obtain the multidimensional textile raw material interaction characteristics. The specific calculation formula is: ; in, Representing multi-dimensional textile raw material interaction characteristics; represents a Bi-LSTM network; Represents multi-dimensional textile raw material characteristics; Represents a graph convolutional neural network; represents the procurement adjacency matrix; Represents the weight coefficient, by taking the time series characteristics of textile raw materials Input to the channel attention mechanism; The procurement prediction layer predicts the interactive features of multi-dimensional textile raw materials based on the fully connected network and the prediction head, and outputs the future procurement plan, including the procurement type and quantity at the next procurement time point.

[0046] Furthermore, the specific implementation process of the prompt and model update module includes: Real-time monitoring of textile raw material sales records and textile raw material inventory data within the procurement cycle, through , calculate the purchase forecast deviation; For the The purchase forecast deviation of various textile raw materials; For the The purchase quantity of the textile raw materials; For the See Table 4 for some examples of purchasing forecast bias.

[0047] Table 4. Some examples of procurement forecast deviations for different textile raw materials of Textile Enterprise A

[0048] If the purchase forecast deviation is greater than or equal to the predetermined deviation threshold, a prompt is issued to the purchase management personnel; the ideal sales record of textile raw materials and the ideal inventory data of textile raw materials are determined through the purchase forecast deviation, and the textile raw material procurement and sales knowledge graph and the textile raw material purchase forecast model are updated; among them, the ideal sales record of textile raw materials refers to the sales record without considering whether the inventory is insufficient; the ideal inventory data of textile raw materials refers to the inventory data determined according to the ideal sales record of textile raw materials without considering the inventory capacity and the actual inventory quantity of textile fabrics, that is, when the inventory is insufficient, the corresponding textile raw material quantity can be expressed in the form of a negative number.

[0049] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A textile raw material procurement and sales matching method based on semantic data, characterized in that: include: Obtain textile raw material purchase records within a predetermined period of time, and collect textile raw material sales records, textile raw material inventory data and textile market data within the purchase cycle; Determine the attribute edge weights of the associated edges between different node attributes through the semantic rules between the textile raw material procurement records, the textile raw material sales records, the textile raw material inventory data and the textile market data; construct a textile raw material procurement and sales knowledge graph based on the node attributes and the associated edges; Constructing a textile raw material procurement prediction model, inputting the multidimensional textile raw material time series data and the procurement adjacency matrix in the previous procurement cycle; extracting the multidimensional textile raw material features of the multidimensional textile raw material time series data through the time series feature extraction layer; According to the multi-dimensional feature interaction layer, the procurement adjacency matrix and the multi-dimensional textile raw material features are exchanged, and the future procurement plan is predicted in combination with the procurement prediction layer; The textile raw material sales records and the textile raw material inventory data are monitored in real time, the purchase forecast deviation is calculated, and it is determined whether to issue a prompt to the purchase management personnel and update the textile raw material purchase and sales knowledge graph and the textile raw material purchase forecast model.

2. The textile raw material procurement and sales matching method based on semantic data according to claim 1 is characterized in that: Acquire the textile raw material purchase record, the textile raw material sales record, the textile raw material inventory data and the textile market data at the same purchase time point, and construct the textile raw material purchase and sales knowledge graph; The textile raw material procurement and sales knowledge graph includes textile raw material data nodes and the associated edges; the textile raw material data nodes include procurement nodes, sales nodes, inventory nodes and market demand nodes, and each of the textile raw material data nodes includes several node attributes; the associated edges are used to connect the node attributes, indicating the data interaction between two node attributes; for two different node attributes, the attribute edge weight of the associated edge is set in combination with the attribute data sequence and the corresponding semantic rules between the textile raw material data nodes.

3. The textile raw material procurement and sales matching method based on semantic data according to claim 2 is characterized in that: The specific process of setting the attribute edge weight of the associated edge includes: Constructing the semantic rules between the textile raw material data nodes, including a first semantic rule between the procurement node and the sales node, a second semantic rule between the inventory node and the sales node, and a third semantic rule between the procurement node and the market demand node; According to the attribute data sequence of the node attributes, the time series difference weight is calculated; the attribute correlation weight between different node attributes is calculated through the cosine similarity function; according to the semantic rules between the nodes corresponding to different node attributes, the time series difference weight and the attribute correlation weight are combined to obtain the attribute edge weight of the associated edge.

4. The textile raw material procurement and sales matching method based on semantic data according to claim 1 is characterized in that: The textile raw material procurement prediction model includes the temporal feature extraction layer, the multi-dimensional feature interaction layer and the procurement prediction layer; the specific prediction process includes: Performing data cleaning on the textile raw material sales records, the textile raw material inventory data and the textile market data in the previous procurement cycle to obtain the multi-dimensional textile raw material time series data; obtaining the adjacency matrix of the textile raw material procurement and sales knowledge graph, and performing normalization processing to obtain the procurement adjacency matrix; Inputting the multi-dimensional textile raw material time series data into the time series feature extraction layer, and obtaining the multi-dimensional textile raw material features through a convolutional neural network, including the historical sales quantity, historical sales season, historical inventory quantity and market price change of the textile raw materials; Input the multidimensional textile raw material features and the procurement adjacency matrix into the multidimensional feature interaction layer, capture the data change law of the multidimensional textile raw material features through Bi-LSTM, and generate textile raw material time series features; combine the graph convolutional neural network and the weighted fusion mechanism to perform feature interaction on the textile raw material time series features and the procurement adjacency matrix to obtain multidimensional textile raw material interaction features; The procurement prediction layer predicts the multi-dimensional textile raw material interaction characteristics according to the fully connected network and the prediction head, and outputs the future procurement plan.

5. The textile raw material procurement and sales matching method based on semantic data according to claim 1 is characterized in that: The specific process of determining whether to issue a prompt to the purchasing manager and updating the textile raw material procurement and sales knowledge graph and the textile raw material procurement prediction model includes: The textile raw material sales records and the textile raw material inventory data within the procurement cycle are monitored in real time, and the procurement forecast deviation is calculated; if the procurement forecast deviation is greater than or equal to a predetermined deviation threshold, a prompt is issued to the procurement management personnel; through the procurement forecast deviation, the ideal textile raw material sales records and the ideal textile raw material inventory data are determined, and the textile raw material procurement and sales knowledge graph and the textile raw material procurement forecast model are updated.

6. A textile raw material procurement and sales matching system based on semantic data, characterized in that: include: The procurement and sales data collection module obtains the textile raw material procurement records within a predetermined time, and collects the textile raw material sales records, textile raw material inventory data and textile market data within the procurement cycle; A procurement and sales data association module determines the attribute edge weights of the associated edges between different node attributes through the semantic rules between the textile raw material procurement records, the textile raw material sales records, the textile raw material inventory data and the textile market data; and constructs a textile raw material procurement and sales knowledge graph based on the node attributes and the associated edges; The procurement plan forecasting module constructs a textile raw material procurement forecasting model, inputs the multidimensional textile raw material time series data and the procurement adjacency matrix in the previous procurement cycle; extracts the multidimensional textile raw material features of the multidimensional textile raw material time series data through the time series feature extraction layer; According to the multi-dimensional feature interaction layer, the procurement adjacency matrix and the multi-dimensional textile raw material features are exchanged, and the future procurement plan is predicted in combination with the procurement prediction layer; The prompt and model update module monitors the textile raw material sales records and the textile raw material inventory data in real time, calculates the purchase forecast deviation, determines whether to issue prompts to the purchase management personnel, and updates the textile raw material procurement and sales knowledge graph and the textile raw material procurement forecast model.

7. A textile raw material procurement and sales matching system based on semantic data according to claim 6, characterized in that: The procurement and sales data association module obtains the textile raw material procurement records, the textile raw material sales records, the textile raw material inventory data and the textile market data at the same procurement time point, and constructs the textile raw material procurement and sales knowledge graph; The textile raw material procurement and sales knowledge graph includes textile raw material data nodes and the associated edges; the textile raw material data nodes include procurement nodes, sales nodes, inventory nodes and market demand nodes, and each of the textile raw material data nodes includes several node attributes; the associated edges are used to connect the node attributes, indicating the data interaction between two node attributes; for two different node attributes, the attribute edge weight of the associated edge is set in combination with the attribute data sequence and the corresponding semantic rules between the textile raw material data nodes.

8. The textile raw material procurement and sales matching system based on semantic data according to claim 7 is characterized in that: The specific process of setting the attribute edge weight of the associated edge includes: Constructing the semantic rules between the textile raw material data nodes, including a first semantic rule between the procurement node and the sales node, a second semantic rule between the inventory node and the sales node, and a third semantic rule between the procurement node and the market demand node; According to the attribute data sequence of the node attributes, the time series difference weight is calculated; the attribute correlation weight between different node attributes is calculated through the cosine similarity function; according to the semantic rules between the nodes corresponding to different node attributes, the time series difference weight and the attribute correlation weight are combined to obtain the attribute edge weight of the associated edge.

9. The textile raw material procurement and sales matching system based on semantic data according to claim 6 is characterized in that: The specific forecasting process of the procurement plan forecasting module includes: The multidimensional textile raw material time series data is obtained according to the textile raw material sales records, the textile raw material inventory data and the textile market data in the previous procurement cycle; the procurement adjacency matrix is ​​obtained through the adjacency matrix of the textile raw material procurement and sales knowledge graph; the multidimensional textile raw material time series data is input into the time series feature extraction layer to obtain the multidimensional textile raw material features; the multidimensional textile raw material features and the procurement adjacency matrix are input into the multidimensional feature interaction layer to obtain the multidimensional textile raw material interaction features; the procurement prediction layer predicts the multidimensional textile raw material interaction features according to the fully connected network and the prediction head, and outputs the future procurement plan.

10. The textile raw material procurement and sales matching system based on semantic data according to claim 6, characterized in that: The specific implementation process of the prompt and model update module includes: The textile raw material sales records and the textile raw material inventory data within the procurement cycle are monitored in real time, and the procurement forecast deviation is calculated; if the procurement forecast deviation is greater than or equal to a predetermined deviation threshold, a prompt is issued to the procurement management personnel; through the procurement forecast deviation, the ideal textile raw material sales records and the ideal textile raw material inventory data are determined, and the textile raw material procurement and sales knowledge graph and the textile raw material procurement forecast model are updated.

Images