Supply Chain Scheduling Optimization Method and System Based on Multi-Source Data

Abstract

The present invention discloses a supply chain production scheduling optimization method and system based on multi-source data. The method extracts supplier procurement data through a data crawler, analyzes the commonality of raw materials in combination with a knowledge graph, and generates a raw material commonality list; based on delivery records and product attributes, a linear regression is used to predict the quality stability value, and the process similarity is calculated through a text similarity algorithm to construct a process similarity matrix; in combination with historical sales data, a multivariate linear regression and time series analysis are used to construct a market demand prediction model. The complex relationships among raw material commonality, process similarity, and market demand are characterized through a graph neural network to generate a product association network diagram, and the Pareto optimization is used to find the optimal balance point between delivery time and resource utilization rate to form a production scheduling plan; finally, an integer programming algorithm is used to comprehensively evaluate the performance score, quality stability value, raw material commonality list, and production scheduling plan to optimize the supplier portfolio. The present invention improves the production scheduling efficiency and accuracy and meets complex business requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of supply chain control, and in particular, to a supply chain production scheduling optimization method and system based on multi-source data. Background Art

[0002] In the context of globalization and digital transformation, the optimization of supply chain production scheduling is crucial for an enterprise's operational efficiency and market response ability. However, the integration and application of multi-source data face numerous technical challenges and are difficult to meet current requirements. First, the identification of raw material commonality is limited by the inconsistent data formats of suppliers, resulting in difficult data cleaning and matching, which hinders the refined management of resources. Second, the calculation of production process similarity relies on historical data and process flows, but the slight process differences between products are easily overlooked, affecting the accuracy of similarity assessment and limiting the optimal allocation of production resources. The analysis of market demand correlation requires the integration of information such as historical sales data, market forecasts, and seasonal fluctuations, but these data are scattered in different systems, making data synchronization and consistency maintenance a difficult problem and reducing the accuracy and timeliness of market forecasts. In addition, the construction of a product association network needs to comprehensively consider various factors, but there are complex non-linear relationships between these factors, and traditional linear models are difficult to accurately describe, resulting in insufficient network accuracy and affecting the collaborative efficiency of the supply chain. To meet the requirements, enterprises need to introduce advanced data integration technologies, intelligent modeling methods, and multi-objective optimization algorithms to achieve the intelligence and high efficiency of the supply chain.

[0003] Existing supply chain production scheduling optimization technologies often adopt linear programming methods based on rule engines. First, the system obtains raw material data from suppliers, but due to inconsistent data formats, manual intervention is required for data cleaning and format conversion. Then, based on historical production data and process flows, the system uses a simple Euclidean distance calculation method to evaluate production process similarity. For market demand analysis, the system obtains historical sales data from the ERP system, extracts market forecast information from the CRM system, and manually imports seasonal fluctuation data, and generates the final demand forecast through the weighted average method. Finally, when constructing a product association network, the system mainly considers the raw material composition and production process of products, uses traditional linear regression models to describe the relationships between various factors, and determines whether there is a correlation through a fixed threshold.

[0004] In the existing supply chain scheduling optimization technology, in terms of multi-source data integration, due to the inconsistent data formats provided by different suppliers, the data cleaning and matching processes are time-consuming and error-prone, hindering the refined management of resources. Secondly, the identification of production process differences usually relies on historical data and process flow descriptions, but the small process differences between products are often ignored, resulting in inaccurate process similarity evaluation and affecting the optimal allocation of production resources. In terms of market demand forecasting, existing methods are difficult to effectively integrate historical sales data, market forecasts, and seasonal fluctuation information, resulting in insufficient forecasting accuracy and timeliness and being unable to accurately reflect the complexity of market changes. In addition, the construction of a product association network needs to comprehensively consider various factors, but there are complex non-linear relationships between these factors, which are difficult to accurately describe by traditional linear models, resulting in insufficient network accuracy and affecting the collaborative efficiency of the supply chain. In terms of multi-objective optimization balance, the scheduling plan usually needs to balance multiple objectives such as delivery time and resource utilization rate. Existing methods are difficult to effectively handle the contradictions between these objectives, resulting in the optimization results often unable to achieve the global optimum. Finally, the evaluation of supplier performance often lacks comprehensiveness and scientificity, relying only on single indicators or subjective judgments and being unable to accurately reflect the comprehensive capabilities of suppliers, thus affecting the rationality of supplier selection and the stability of the supply chain. In summary, the existing supply chain scheduling optimization technology has significant deficiencies in multi-source data integration, production process difference identification, market demand forecasting, non-linear relationship modeling, multi-objective optimization balance, and supplier performance evaluation, resulting in the scheduling efficiency and accuracy being difficult to meet the complex business requirements. Summary of the Invention

[0005] The present invention provides a supply chain scheduling optimization method and system based on multi-source data to improve the scheduling efficiency and accuracy and meet complex business requirements.

[0006] In a first aspect, to solve the above technical problems, the present invention provides a supply chain scheduling optimization method based on multi-source data, including:

[0007] Obtain the procurement data, delivery records, detailed attributes, process flows, and historical sales data of suppliers, where the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data, and risk event records;

[0008] Extract raw material information according to the procurement data, and analyze it in combination with data traceability and knowledge graphs to obtain a raw material sharing list;

[0009] Predict the quality stability value using a linear regression algorithm according to the delivery records and the detailed attributes;

[0010] Conduct a comprehensive performance evaluation of suppliers according to the detailed attributes and the raw material sharing list to obtain a performance score;

[0011] According to the described process flow, a text similarity algorithm is used to calculate the process similarity of different products and generate a process similarity matrix;

[0012] Based on the process similarity matrix and historical sales data, a market demand prediction model is constructed and market demand is predicted to obtain the market demand prediction value;

[0013] According to the raw material sharing list, the process similarity matrix and the market demand prediction value, a product association network diagram is generated using a graph neural network;

[0014] According to the product association network diagram, the Pareto optimization method is used to find a balance between delivery time and resource utilization rate, and the production scheduling combination corresponding to the balance point is used as the optimized production scheduling plan;

[0015] According to the performance score, the quality stability value, the raw material sharing list and the optimized production scheduling plan, an integer programming algorithm is used to optimize the supplier combination to obtain the supplier combination plan.

[0016] In an alternative implementation, the extracting raw material information from the procurement data and performing analysis in combination with data traceability and knowledge graph to obtain the raw material sharing list includes:

[0017] The raw material information in the procurement data is extracted through data crawling technology, and the raw material information is normalized to obtain standardized data;

[0018] According to the entity relationship between the standardized data and the preset knowledge graph, a raw material traceability path is established;

[0019] The raw material traceability path is stored using graph database technology to form a raw material relationship network;

[0020] According to the raw material relationship network and the preset raw material sharing rules, raw material combinations with sharing relationships are identified in the knowledge graph, and a standardized raw material sharing list is generated according to the raw material combinations.

[0021] In an alternative implementation, the predicting the quality stability value using a regression algorithm according to the delivery record and the detailed attributes includes:

[0022] The order quantity, delivery quantity, quality value, batch value and season value are extracted from the delivery record and an initial data set is formed;

[0023] Data cleaning and feature extraction are performed based on the initial data set, the historical quality value, the process value and the material property value to obtain a feature data set;

[0024] Using a linear regression algorithm, taking the feature data set as input, training a linear regression model to obtain regression values;

[0025] Calculating a stability index based on the regression values, the order quantity, and the season value to obtain a quality stability value.

[0026] In an alternative embodiment, the comprehensive performance evaluation of the suppliers based on the detailed attributes and the raw material sharing list to obtain a performance score includes:

[0027] Calculating the delivery rate and the qualification rate of the suppliers based on the historical delivery data and the quality inspection records;

[0028] Quantifying the production capacity number and the risk degree of the suppliers based on the production capacity data and the risk event records;

[0029] Analyzing the sharing degree and the consistency degree among the suppliers based on the raw material sharing list;

[0030] Using the analytic hierarchy process to establish an evaluation model, assigning weights to the delivery rate, the qualification rate, the production capacity number, the risk degree, the sharing degree, the consistency degree, the preset timeliness and flexibility indicators to obtain weight data;

[0031] Calculating the performance score based on the delivery rate, the qualification rate, the production capacity number, the risk degree, the sharing degree, the consistency degree, and the weight data.

[0032] In an alternative embodiment, the constructing a market demand prediction model based on the process similarity matrix and the historical sales data and performing market demand prediction to obtain a market demand prediction value includes:

[0033] Based on the process similarity matrix and the historical sales data, using a multiple linear regression method to construct an initial market demand prediction model;

[0034] Using a time series analysis method to analyze the seasonal fluctuation characteristics of the historical sales data;

[0035] Incorporating the seasonal fluctuation characteristics into the initial market demand prediction model to adjust the model parameters, obtaining a market demand prediction model, and calculating the market demand prediction value based on the market demand prediction model.

[0036] In an alternative embodiment, the generating a product association network diagram using a graph neural network based on the raw material sharing list, the process similarity matrix, and the market demand prediction value includes:

[0037] Calculating a construction value based on the raw material sharing list, the process similarity matrix, and the market demand prediction value;

[0038] Use a graph neural network to characterize the nonlinear relationship of the constructed values, obtain the characterized values, and generate a product association network diagram through a visualization tool according to the characterized values.

[0039] In an alternative embodiment, the optimizing the supplier portfolio by using an integer programming algorithm according to the performance score, the quality stability value, the raw material sharing list, and the optimized production scheduling plan to obtain a supplier portfolio plan includes:

[0040] Determine the critical material requirements according to the performance score, the historical quality value, and the optimized production scheduling plan, and evaluate the supply guarantee ability of the suppliers to form a supply ability matrix;

[0041] According to the supply ability matrix, use an integer programming algorithm, with the performance score and the optimized production scheduling plan as constraint conditions, and the material sharing and quality stability as the objective function, to calculate the initial supplier portfolio;

[0042] Determine whether there are suppliers with insufficient supply guarantee ability or unqualified historical quality in the initial supplier portfolio. If so, screen eligible alternative suppliers in the supplier database according to the criteria of high score and stable quality, and update the supply ability matrix;

[0043] Use an integer programming algorithm to perform secondary optimization on the updated supply ability matrix, calculate the final supplier portfolio, and obtain the supplier portfolio plan.

[0044] In a second aspect, the present invention provides a supply chain production scheduling optimization system based on multi-source data, including:

[0045] A data acquisition module for acquiring the procurement data, delivery records, detailed attributes, process flows, and historical sales data of suppliers, where the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data, and risk event records;

[0046] A raw material sharing analysis module for extracting raw material information according to the procurement data and performing analysis in combination with data traceability and knowledge graphs to obtain a raw material sharing list;

[0047] A quality stability analysis module for predicting the quality stability value by using a linear regression algorithm according to the delivery record and the detailed attributes;

[0048] A performance evaluation module for comprehensively evaluating the performance of suppliers according to the detailed attributes and the raw material sharing list to obtain a performance score;

[0049] A process similarity analysis module for calculating the process similarity of different products by using a text similarity algorithm according to the process flows and generating a process similarity matrix;

[0050] A market demand forecasting module, configured to construct a market demand forecasting model based on the process similarity matrix and historical sales data, and conduct market demand forecasting to obtain a market demand forecasting value;

[0051] An association network generation module, configured to generate a product association network diagram by using a graph neural network according to the raw material sharing list, the process similarity matrix, and the market demand forecasting value;

[0052] A production scheduling plan optimization module, configured to find a balance between delivery time and resource utilization rate by using the Pareto optimization method according to the product association network diagram, and use the production scheduling combination corresponding to the balance point as an optimized production scheduling plan;

[0053] A supplier combination module, configured to optimize the supplier combination by using an integer programming algorithm according to the performance score, the quality stability value, the raw material sharing list, and the optimized production scheduling plan to obtain a supplier combination plan.

[0054] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the above-mentioned supply chain production scheduling optimization method based on multi-source data is implemented.

[0055] In a fourth aspect, the present invention further provides a computer-readable storage medium. The computer-readable storage medium includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the above-mentioned supply chain production scheduling optimization method based on multi-source data.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] (1) Through data traceability and knowledge graph technology, the standardized integration of multi-source supplier data and the accurate identification of shared raw materials are realized, solving the problems of data cleaning and matching caused by inconsistent data formats;

[0058] (2) By using a text similarity algorithm in combination with process descriptions, the production process similarity between products is accurately calculated, effectively capturing subtle differences and improving the accuracy of process similarity evaluation;

[0059] (3) Based on a graph neural network and the Pareto optimization method, an intelligent balance between market demand forecasting and production scheduling combination is realized, solving the conflict problem in multi-objective optimization;

[0060] (4) Optimize the supplier portfolio through integer programming algorithm, comprehensively considering quality stability, delivery ability and performance score, ensuring the scientificity and reliability of supplier selection.

[0061] In summary, based on data crawler and knowledge graph technologies, the present invention realizes the standardized integration of multi-source supplier data and the accurate identification of raw material sharing; secondly, linear regression algorithm and text similarity algorithm are adopted to predict the quality stability value and calculate the process similarity respectively, ensuring the accurate capture of production process differences; at the same time, combined with historical sales data and process similarity matrix, a market demand prediction model is constructed to improve the prediction accuracy and timeliness; on this basis, complex non-linear relationships are characterized by graph neural network, a product association network diagram is generated, and the Pareto optimization method is used to find the optimal balance point between delivery time and resource utilization rate to form an optimized production scheduling plan; finally, the supplier portfolio is optimized based on integer programming algorithm, comprehensively considering performance score, quality stability value and raw material sharing list, ensuring the scientificity and reliability of supplier selection. Through the organic combination of the above steps, the present method effectively solves the deficiencies of traditional methods in aspects such as data integration, process evaluation, demand prediction and supplier selection, and significantly improves the collaborative efficiency of the supply chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic flow chart of a supply chain production scheduling optimization method based on multi-source data provided by the first embodiment of the present invention;

[0063] Figure 2 is a schematic structural diagram of a supply chain production scheduling optimization system based on multi-source data provided by the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0065] Referring to Figure 1 , the first embodiment of the present invention provides a supply chain production scheduling optimization method based on multi-source data, including the following steps:

[0066] S11, obtain the procurement data, delivery records, detailed attributes, process flows and historical sales data of suppliers, wherein the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data and risk event records;

[0067] S12. Extract raw material information based on the procurement data, and analyze it in combination with data traceability and knowledge graphs to obtain a raw material sharing list;

[0068] S13. Use the linear regression algorithm to predict the quality stability value based on the delivery record and the detailed attributes;

[0069] S14. Conduct a comprehensive performance evaluation of the suppliers based on the detailed attributes and the raw material sharing list to obtain a performance score;

[0070] S15. Calculate the process similarity of different products using the text similarity algorithm according to the process flow, and generate a process similarity matrix;

[0071] S16. Construct a market demand prediction model based on the process similarity matrix and historical sales data and conduct market demand prediction to obtain a market demand prediction value;

[0072] S17. Generate a product association network diagram using a graph neural network based on the raw material sharing list, the process similarity matrix, and the market demand prediction value;

[0073] S18. Use the Pareto optimization method to find a balance between delivery time and resource utilization rate based on the product association network diagram, and use the scheduling combination corresponding to the balance point as the optimized scheduling plan;

[0074] S19. Optimize the supplier combination using the integer programming algorithm based on the performance score, the quality stability value, the raw material sharing list, and the optimized scheduling plan to obtain a supplier combination plan.

[0075] In step S11, obtain the procurement data, delivery records, detailed attributes, process flow, and historical sales data of the suppliers. Among them, the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data, and risk event records.

[0076] In this embodiment, the procurement data is sourced from the order information of suppliers, covering the types, quantities, and price information of raw materials. These data are extracted from the suppliers' e-procurement systems through data crawling technology to ensure the timeliness and accuracy of the data. The delivery records record the delivery time, quantity, and quality of the suppliers, reflecting the actual delivery capabilities of the suppliers, and are obtained from the logistics management system. The detailed attributes further refine the performance of suppliers in terms of quality, process, materials, and risk management, and can comprehensively reflect the operation level of the suppliers. Among them, the historical quality values are statistically obtained based on past quality inspection results, the process values are evaluated according to the complexity and technical level of the process flow, the material characteristic values involve the physical and chemical characteristics of the raw materials, the historical delivery data and quality inspection records are obtained from the logistics and quality control systems respectively, the production capacity data is sourced from the production management system, and the risk event records come from the logs of the risk management system.

[0077] The process flow details the production process steps of the products, usually provided by the production department, and is used to evaluate the process differences among different suppliers. The historical sales data is sourced from the enterprise's sales management system and records the sales situations of different products.

[0078] In step S12, raw material information is extracted according to the procurement data, and analyzed in combination with data traceability and knowledge graph to obtain a raw material sharing list.

[0079] In a specific implementation manner, the extracting raw material information according to the procurement data, and analyzing in combination with data traceability and knowledge graph to obtain a raw material sharing list includes:

[0080] Extract the raw material information in the procurement data through data crawling technology, and perform normalization processing on the raw material information to obtain standardized data;

[0081] Establish a raw material traceability path according to the entity relationship between the standardized data and the preset knowledge graph;

[0082] Adopt graph database technology to store the raw material traceability path to form a raw material relationship network;

[0083] Identify the raw material combinations with sharing relationships in the knowledge graph according to the raw material relationship network and the preset raw material sharing rules, and generate a standardized raw material sharing list according to the raw material combinations.

[0084] Specifically, first, raw material information, including keywords such as raw material name, specification, and supplier number, is extracted from the procurement data of suppliers through data crawling technology. Since the extracted raw data may have problems such as inconsistent formats or incomplete information due to diverse sources, it is necessary to normalize the raw material information. Normalization processing includes removing duplicate data, filling in missing values, and unifying naming specifications to ensure data consistency and usability. The normalized data is converted into standardized data for subsequent processing.

[0085] Next, the preset knowledge graph entity relationships are matched with the standardized data to generate the raw material traceability path. The knowledge graph represents entities and their relationships through a graph structure, where entities include nodes such as raw materials, suppliers, and production processes, and relationships include supply relationships, processing relationships, and dependency relationships. The construction of the raw material traceability path is achieved through graph traversal algorithms, specifically using depth-first search (DFS) or breadth-first search (BFS) algorithms. Starting from the raw material node, relevant nodes and edges in the knowledge graph are traversed to form a complete traceability link. For example, for raw material A, its traceability path may be A → supplier X → production process Y → raw material B, indicating the association relationship between A and B in the supply chain.

[0086] After the raw material traceability path is generated, it is stored and represented using graph database technology to form a raw material relationship network. The graph database stores data in the form of nodes and edges, supporting efficient graph traversal and query operations. For example, raw materials A, B, and C are represented as nodes in the graph database, and their relationships such as sharing the same supplier or production process are represented as edges. The graph database can visually display the complex relationships between raw materials for subsequent analysis.

[0087] According to the raw material relationship network and the preset raw material sharing rules, raw material combinations with sharing relationships are identified in the knowledge graph. Raw material sharing rules usually include conditions such as raw material attribute similarity, supplier consistency, and production process sharing. For example, when two raw materials are provided by the same supplier and have similar physical and chemical properties, they are considered to have a sharing relationship. In specific implementation, graph query languages (such as Cypher) are used to retrieve nodes and edges that meet the conditions in the graph database to generate raw material combinations with sharing relationships.

[0088] The process of generating the raw material sharing list involves standardizing the raw material combinations, including removing duplicate combinations, marking priorities, and assigning unique identifiers. The finally generated raw material sharing list is stored in structured data form, containing raw material numbers, sharing relationships, and priority information.

[0089] This step realizes the integration of multi-source procurement data and the joint analysis of the knowledge graph, and efficiently identifies raw material combinations with sharing relationships through graph database and graph algorithm technologies.

[0090] In step S13, according to the delivery record and the detailed attributes, a linear regression algorithm is used to predict the quality stability value.

[0091] In a specific implementation manner, the predicting the quality stability value by using a regression algorithm according to the delivery record and the detailed attributes includes:

[0092] Extract the order quantity, delivery quantity, quality value, batch value, and season value from the delivery record, and form an initial data set;

[0093] Perform data cleaning and feature extraction according to the initial data set, the historical quality value, the process value, and the material characteristic value to obtain a feature data set;

[0094] Use the linear regression algorithm, take the feature data set as the input, train the linear regression model, and obtain the regression value;

[0095] Calculate the stability index according to the regression value, the order quantity, and the season value to obtain the quality stability value.

[0096] Specifically, first, extract the order quantity, delivery quantity, quality value, batch value, and season value from the delivery record to form an initial data set. The order quantity represents the number of orders received by the supplier, the delivery quantity represents the number of goods actually delivered, the quality value is based on the quality inspection results of the delivered goods, the batch value is used to identify different production batches, and the season value reflects the time characteristics of order delivery (such as quarters or months). These data are obtained through database queries or data crawling technologies to ensure data integrity and accuracy.

[0097] Next, perform data cleaning and feature extraction on the initial data set to generate a feature data set. Data cleaning includes removing missing values, handling outliers, and unifying data formats. Feature extraction selects features that have a significant impact on quality stability prediction from the initial data set, including historical quality values, process values, and material characteristic values. The historical quality value is statistically obtained based on the quality inspection results in past delivery records, the process value represents the complexity and technical level of the production process, and the material characteristic value involves the physical and chemical properties of raw materials. The extracted features are normalized and transformed into standardized feature vectors for easy model training.

[0098] After the feature data set is generated, a linear regression algorithm is used for model training. The form of the linear regression model is:

[0099]

[0100] where is the predicted value, 、 、… are feature variables, , , … are regression coefficients, which are solved by minimizing the error between the predicted value and the actual value. The error function uses the mean squared error :

[0101]

[0102] where is the number of samples, is the actual value, is the predicted value. Through optimization algorithms such as the gradient descent method, the regression coefficients are iteratively updated until the error converges to obtain the optimal model parameters.

[0103] After training is completed, the linear regression model is used to predict the feature dataset to obtain the regression value. The regression value represents the expected quality performance of the supplier under specific conditions. To further quantify the quality stability, the stability index is calculated by combining the order volume and the season value. The calculation formula of the stability index is:

[0104]

[0105] where is the regression value, represents the influence coefficient of the preset order volume on stability, represents the influence coefficient of the preset season value on stability, is the normalization factor to ensure that the stability index is within a reasonable range. Finally, the stability index is output as the quality stability value to evaluate the quality performance of the supplier.

[0106] Through the analysis of historical data and process characteristics, a regression model that can predict the quality stability of suppliers is constructed. During the model training process, the linear regression algorithm learns the linear relationship between the feature variables and the quality performance by minimizing the error function to ensure the accuracy of the prediction results.

[0107] In step S14, a comprehensive performance evaluation of the supplier is performed according to the detailed attributes and the raw material sharing list to obtain a performance score.

[0108] In a specific implementation manner, the comprehensive performance evaluation of the supplier according to the detailed attributes and the raw material sharing list to obtain a performance score includes:

[0109] Calculating the delivery rate and the pass rate of the supplier according to the historical delivery data and the quality inspection records;

[0110] Quantifying the production capacity number and the risk degree of the supplier according to the production capacity data and the risk event records;

[0111] Analyze the sharing degree and consistency among suppliers according to the raw material sharing list;

[0112] Use the analytic hierarchy process to establish an evaluation model, assign weights to the delivery rate, the qualification rate, the production capacity, the risk degree, the sharing degree, the consistency, the preset timeliness and flexibility indicators, and obtain weight data;

[0113] Calculate the performance score according to the delivery rate, the qualification rate, the production capacity, the risk degree, the sharing degree, the consistency and the weight data;

[0114] Among them, the delivery rate, the qualification rate, the production capacity, the risk degree, the sharing degree and the consistency are calculated by the following formulas:

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] Among them, the performance score is calculated by the following formula:

[0122]

[0123] Among them, represents the delivery rate, represents the qualification rate, represents the production capacity, represents the risk degree, represents the sharing degree, represents the consistency, represents the total delivery quantity and represents the total order quantity. The total delivery quantity and the total order quantity are extracted from the enterprise's logistics management system, and respectively reflect the quantity of goods actually delivered by the supplier and the total number of received orders; represents the quantity of qualified products, which is obtained based on the quality inspection records and is used to evaluate the quality level of the products delivered by the supplier, represents the total production capacity and represents the time period. The total production capacity and the time period are obtained from the production management system, and respectively represent the maximum production capacity and the production cycle of the supplier within a specific time, represents the number of risk events and Represents the total number of events. The number of risk events and the total number of events are sourced from the logs of the risk management system and are used to quantify the risk management level of the supplier. Represents the quantity of common raw materials and Represents the total quantity of raw materials, calculated from procurement data and the raw material sharing list, indicating the quantity of raw materials shared among suppliers and their total quantity respectively. Represents the quantity of raw materials with stable supply, statistically obtained based on historical delivery data and quality inspection records, reflecting the stability of the supplier in raw material supply. 、 、 、 、 、 、 and Represents the weight data, Represents the timeliness index, Represents the flexibility index, Represents the performance score.

[0124] Specifically, first, calculate the delivery rate and the pass rate based on historical delivery data and quality inspection records. The delivery rate reflects the supplier's ability to deliver on time and in full, and the calculation formula is:

[0125]

[0126] Where Represents the actual delivery quantity, Represents the total number of orders. The pass rate reflects the quality level of the products delivered by the supplier, and the calculation formula is:

[0127]

[0128] Where Represents the quantity of qualified products, Represents the actual delivery quantity.

[0129] Next, quantify the production capacity number and the risk degree based on production capacity data and risk event records. The production capacity number represents the production capacity of the supplier, and the calculation formula is:

[0130]

[0131] Where Represents the total production capacity, Represents the time period. The risk degree reflects the risk management level of the supplier, and the calculation formula is:

[0132]

[0133] Where Represents the number of risk events, Represents the total number of events.

[0134] Analyze the sharing degree and consistency among suppliers based on the raw material sharing list. The sharing degree represents the proportion of raw materials shared among suppliers, and the calculation formula is:

[0135]

[0136] Where Represents the quantity of shared raw materials, Represents the total quantity of raw materials. The consistency represents the proportion of raw materials stably supplied by suppliers, and the calculation formula is:

[0137]

[0138] Where Represents the quantity of raw materials stably supplied, Represents the total quantity of raw materials.

[0139] Use the analytic hierarchy process to establish an evaluation model and assign weights to the delivery rate, qualification rate, production capacity, risk degree, sharing degree, consistency, and preset timeliness and flexibility indicators. The analytic hierarchy process calculates the weight values of each indicator by constructing a judgment matrix. The specific steps include: constructing a hierarchical structure, clarifying the goal layer, criterion layer, and indicator layer; designing a judgment matrix to determine the importance between indicators through pairwise comparison; calculating the weight vector by solving the maximum eigenvalue and its corresponding eigenvector of the judgment matrix using the eigenvalue method; and finally conducting a consistency test to ensure the rationality of weight allocation.

[0140] Finally, the weight data , , , , , , and Represent the weights of the delivery rate, qualification rate, production capacity, risk degree, sharing degree, consistency, timeliness, and flexibility respectively.

[0141] Calculate the performance score based on the delivery rate, qualification rate, production capacity, risk degree, sharing degree, consistency, and weight data. The formula is:

[0142]

[0143] Where Represents the timeliness indicator, Represents the flexibility indicator, Represents the performance score.

[0144] Through multi-dimensional data analysis and comprehensive evaluation using the Analytic Hierarchy Process, a quantified performance score is generated, providing a scientific basis for supplier selection and management. The application of the Analytic Hierarchy Process ensures the scientificity and rationality of weight allocation, while the comprehensive scoring formula intuitively reflects the comprehensive performance of suppliers. This evaluation method not only improves the accuracy of supplier management but also provides important support for supply chain optimization and risk control.

[0145] In step S15, according to the process flow, a text similarity algorithm is used to calculate the process similarity of different products and generate a process similarity matrix.

[0146] In a specific implementation manner, the calculating the process similarity of different products and generating a process similarity matrix according to the process flow includes:

[0147] The process similarity is calculated through the following formula:

[0148]

[0149] where and represent the process similarity of product relative to product , and represent the text vectors of the process flows of the two products, and represent and the weights of the th feature in the text vectors, is the total number of products.

[0150] Specifically, first, text preprocessing is performed on the process flow, including word segmentation, stop word removal, and stemming. Word segmentation decomposes the text into independent word units, stop word removal filters out insignificant words (such as "of", "is", etc.), and stemming reduces words to their basic forms (e.g., both "processing" and "processed" are reduced to "process"). The preprocessed text is transformed into a standardized sequence of words for subsequent processing.

[0151] Next, the preprocessed text is transformed into a text vector. The bag-of-words model or TF-IDF (term frequency-inverse document frequency) method is used to calculate the weight value of each word. The bag-of-words model represents text features by counting the frequency of words in a document, while the TF-IDF method comprehensively considers the frequency of words in a document and their rarity in the entire corpus.

[0152] where represents the word in the document The word frequency in represents the word Inverse document frequency, and the calculation formula is:

[0153]

[0154] where represents the total number of documents, represents the number of documents containing the word . The weight values obtained by the TF-IDF method form the text vector , where is the total number of words.

[0155] For the process flows of two products and , their process similarity is calculated by the cosine similarity formula:

[0156]

[0157] where represents the process similarity, and respectively represent and the weights of the th feature in the text vector. The value range of the cosine similarity is . The closer its value is to 1, the higher the text similarity; the closer it is to -1, the lower the text similarity.

[0158] Calculate the process similarity between all products to form a process similarity matrix , represents the process similarity between product and product . The process similarity matrix is a symmetric matrix, and the diagonal elements are 1, indicating that the similarity between a product and itself is the highest.

[0159] Through the text similarity algorithm, the process flow is transformed into a quantifiable similarity index, providing a scientific basis for production optimization and resource scheduling. The text vectorization process ensures the numerical expression of text features, and the cosine similarity calculation quantifies the process similarity between different products. This method not only improves the accuracy of process analysis but also provides important support for the formulation of production plans and the optimization of resource allocation.

[0160] In step S16, a market demand prediction model is constructed based on the process similarity matrix and historical sales data, and market demand prediction is performed to obtain a market demand prediction value.

[0161] In a specific implementation manner, constructing a market demand prediction model based on the process similarity matrix and historical sales data and performing market demand prediction to obtain a market demand prediction value includes:

[0162] Constructing an initial market demand prediction model by using the multiple linear regression method according to the process similarity matrix and historical sales data;

[0163] Analyzing the seasonal fluctuation characteristics of the historical sales data by using the time series analysis method;

[0164] Integrating the seasonal fluctuation characteristics into the initial market demand prediction model to adjust the model parameters, obtaining a market demand prediction model, and calculating the market demand prediction value according to the market demand prediction model.

[0165] Specifically, first, an initial market demand prediction model is preliminarily constructed by using the multiple linear regression method according to the process similarity matrix and historical sales data. The historical sales data includes product sales quantity, sales time, and sales region, and the process similarity matrix is used to quantify the process relevance between products. The form of the multiple linear regression model is:

[0166]

[0167] where represents the market demand prediction value, represents independent variables (such as process similarity, sales time, sales region), represents the regression coefficient, represents the random error. By minimizing the difference between the actual value and the predicted value, the regression coefficient is solved. The error function uses the mean square error;

[0168] Through optimization algorithms such as the gradient descent method, the regression coefficient is iteratively updated until the error converges, obtaining a preliminary prediction model.

[0169] Next, the time series analysis method is used to analyze the seasonal fluctuation characteristics of the historical sales data. The time series analysis method includes trend decomposition, seasonal adjustment, and periodic detection. By decomposing the historical sales data, the seasonal component , representing the periodic change of the market demand over time, is extracted. The extraction formula for the seasonal component is:

[0170]

[0171] where represents the time point of the sales data, represents the trend component, represents the periodic component. The seasonal component reflects the periodic fluctuation characteristics of the market demand and provides an important basis for model optimization.

[0172] Incorporate the seasonal fluctuation characteristics into the preliminarily constructed market demand forecasting model and adjust the model parameters. The optimized model form is as follows:

[0173]

[0174] where represents the regression coefficient of the seasonal component. By retraining the model and taking the seasonal component as the new independent variable, the optimized regression coefficient is solved. The optimized model can more accurately reflect the actual change law of market demand.

[0175] Finally, calculate the market demand forecast value according to the optimized market demand forecasting model . The model prediction formula is:

[0176]

[0177] where represents the market demand forecast value.

[0178] By combining the process similarity matrix and historical sales data, a market demand forecasting model that can quantify process relevance and seasonal fluctuation characteristics is constructed. The multiple linear regression method is used to initially establish the forecasting framework, and the time series analysis method is used to extract the seasonal component, and the forecasting accuracy is further improved through model optimization.

[0179] In step S17, according to the raw material sharing list, the process similarity matrix, and the market demand forecast value, use a graph neural network to generate a product association network diagram.

[0180] In a specific implementation manner, the using a graph neural network to generate a product association network diagram according to the raw material sharing list, the process similarity matrix, and the market demand forecast value includes:

[0181] Calculate the construction value according to the raw material sharing list, the process similarity matrix, and the market demand forecast value;

[0182] Use a graph neural network to depict the non-linear relationship of the construction value to obtain the depicted value, and generate a product association network diagram according to the depicted value through a visualization tool;

[0183] where the construction value is calculated by the following formula:

[0184]

[0185]

[0186] where the depicted value is calculated by the following formula:

[0187]

[0188] Among them, represents the predicted value of the product . represents the influence coefficient of the preset product . represents the process similarity of the product relative to the product . and represents the integration value of the product and the product . represents the construction value associated with the product and the product . , and represent the preset integration weights, represents the characterization value of the node at the th layer, and respectively represent the weight matrix and bias term of the th layer, represents the set of neighbor nodes of the node , and respectively represent the degrees of the nodes and , represents the characterization value of the node j at the th layer, represents the characterization value of the node at the th layer.

[0189] The weight matrix and the bias term are the preset trainable parameters of the graph neural network at the th layer, which are used to perform linear transformation and non - linear activation on node features respectively, and they are continuously optimized through the training process of the model to capture the complex relationships between nodes; represents the set of neighbor nodes of the node , reflecting the nodes directly connected to the node in the topological structure of the graph, and this node information is obtained through the graph data structure; and respectively represent the nodes and The degree, which is the number of edges connected to a node, is used to normalize the contributions of neighbor nodes during the feature aggregation process to ensure the balance of information transmission among nodes with different degrees; and respectively represent the feature characterization values of node j and node at the th layer. These values are obtained through layer-by-layer iterative calculations of the graph neural network and are used to characterize the state and feature performance of nodes at specific levels. The initial values of these characterization values are provided by the original features of the nodes and are continuously updated as the network depth increases, ultimately used to generate complex relationship representations between nodes.

[0190] Specifically, first, construct a construction value based on the raw material sharing list, process similarity matrix, and market demand prediction value. The construction value is used to quantify the association strength between products. For product , its integration value is calculated by the following formula:

[0191]

[0192] where represents the market demand prediction value of product , represents the influence coefficient of the preset product , represents the process similarity of product relative to product , and , and represent the preset integration weights. The integration weights are used to balance the contributions of market demand, influence coefficient, and process similarity to the construction value.

[0193] For product and product , its construction value is calculated by the following formula:

[0194]

[0195] Next, use a graph neural network to characterize the non-linear relationship of the construction value. The graph neural network (GNN) characterizes the non-linear relationship between nodes by aggregating the information of neighbor nodes layer by layer. For the characterization value of node at the th layer, it is calculated by the following formula:

[0196]

[0197] Through layer-by-layer iteration, the graph neural network can capture the complex relationships between nodes and generate more accurate characterization values.

[0198] Finally, based on the characterization values, a product association network graph is generated through a visualization tool. Nodes represent products, and edges represent the association strength between products. The visualization tool converts the information of nodes and edges into a graphical representation, intuitively showing the association relationships between products.

[0199] By integrating raw material sharing, process similarity, and market demand forecasting, an association network between products is constructed, and the graph neural network is used to characterize the non-linear relationships, generating a visual product association network graph. The application of the graph neural network ensures the accurate characterization of the relationships between nodes, while the visualization tool enhances the interpretability and practicality of the analysis results.

[0200] In step S18, according to the product association network graph, the Pareto optimization method is used to find a balance point between delivery time and resource utilization rate, and the production scheduling combination corresponding to the balance point is used as the optimized production scheduling plan.

[0201] Specifically, first, a multi-objective optimization model is constructed based on the product association network graph. The product association network graph contains nodes (products) and edges (association relationships between products), where the weight of the edge is determined by the association strength between products. Suppose there are types of products, and the delivery time and resource demand of each product are known. The optimization objectives are to minimize the total delivery time and maximize the total resource utilization rate . The delivery time objective function is defined as:

[0202]

[0203] where represents the production scheduling order of product , represents that product is selected for production scheduling, represents that product is not selected for production scheduling. The resource utilization rate objective function is defined as:

[0204]

[0205] where represents the resource demand of product , has a value range of , and the closer its value is to 1, the higher the resource utilization rate.

[0206] The form of the multi-objective optimization model is as follows:

[0207]

[0208]

[0209] Minimize the objective function, Maximize the objective function.

[0210] Next, the Pareto optimization method is used to solve the multi-objective optimization model. The goal of Pareto optimization is to find a set of Pareto optimal solutions, that is, when one objective cannot be further optimized, the other objective cannot be further optimized either. The Pareto optimal solution set is generated through the following steps:

[0211] S181. Initialize the solution set: Randomly generate a set of production scheduling combinations , where represents the th production scheduling combination, represents the number of initial solutions.

[0212] S182. Evaluate the objective function: For each production scheduling combination, calculate its objective function values and .

[0213] S183. Non-dominated sorting: Perform non-dominated sorting on the solution set , and divide the solution set into multiple non-dominated levels. The dominance relationship is defined as follows: If the production scheduling combination is not inferior to in terms of both delivery time and resource utilization, and is superior to in at least one objective, then dominates . The non-dominated levels are divided according to the solution dominance relationship, and the first level is the Pareto optimal solution.

[0214] S184. Update the solution set: Generate new solutions through operations such as crossover and mutation, and add them to the solution set , and perform non-dominated sorting again. Iteratively update the solution set until the iteration number is reached or the convergence condition is satisfied.

[0215] Finally, select a balance point from the Pareto optimal solution set as the optimized production scheduling plan. The balance point is selected using the linear weighted method, which transforms the multi-objective into a single-objective function:

[0216]

[0217] where and are weight coefficients, satisfying . Minimize A scheduling combination that balances delivery time and resource utilization can be obtained.

[0218] Through the Pareto optimization method, an optimized scheduling plan that balances delivery time and resource utilization is generated based on the product association network diagram. The multi-objective optimization model ensures the scientificity and rationality of the production plan. The Pareto optimal solution set provides multiple choices for decision-makers, and the selection method of the balance point further optimizes the allocation of production resources.

[0219] In step S19, according to the performance score, the quality stability value, the raw material sharing list, and the optimized scheduling plan, an integer programming algorithm is used to optimize the supplier combination to obtain a supplier combination plan.

[0220] In a specific implementation manner, the using the integer programming algorithm to optimize the supplier combination according to the performance score, the quality stability value, the raw material sharing list, and the optimized scheduling plan to obtain a supplier combination plan includes:

[0221] Determine the critical material requirements according to the performance score, the historical quality value, and the optimized scheduling plan, and evaluate the supply guarantee ability of suppliers to form a supply ability matrix;

[0222] According to the supply ability matrix, using the integer programming algorithm, with the performance score and the optimized scheduling plan as constraint conditions, and the material sharing and quality stability as objective functions, calculate the initial supplier combination;

[0223] Judge whether there are suppliers with insufficient supply guarantee ability or unqualified historical quality in the initial supplier combination. If so, according to the criteria of high score and stable quality, screen eligible alternative suppliers in the supplier library and update the supply ability matrix;

[0224] Use the integer programming algorithm to perform secondary optimization on the updated supply ability matrix, calculate the final supplier group, and obtain a supplier combination plan.

[0225] Specifically, first, determine the critical material requirements according to the performance score, the historical quality value, and the optimized scheduling plan, and evaluate the supply guarantee ability of suppliers to form a supply ability matrix. Suppose there are types of materials and suppliers. The performance score of supplier is , the quality stability value is , the supply guarantee ability is , and the demand for material in the optimized scheduling plan is . Supplier 's supply for material Supply capacity is determined by the following formula:

[0226]

[0227] Supply capacity matrix represents the supplier's supply capacity for materials.

[0228] Next, based on the supply capacity matrix, an integer programming algorithm is used to calculate the initial supplier combination. The optimization objectives are to minimize material commonality and maximize quality stability, and the objective function is defined as:

[0229]

[0230] where represents material and material 's commonality, obtained through the raw material common list, represents whether to select supplier , . The constraint conditions include:

[0231] 1. Supply guarantee capacity constraint:

[0232]

[0233] 2. Performance score constraint:

[0234]

[0235] where is the preset performance score threshold. By solving the integer programming model, the initial supplier combination is obtained.

[0236] Then, it is judged whether there are suppliers with insufficient supply guarantee capacity or unqualified historical quality in the initial supplier combination. For the unqualified suppliers , alternative suppliers are screened in the supplier library. The screening conditions for alternative suppliers are:

[0237] Supply guarantee capacity:

[0238]

[0239] Quality stability:

[0240]

[0241] where is the preset quality stability threshold. Through screening and replacement, the supply capacity matrix is updated:

[0242] .

[0243] Finally, the integer programming algorithm is used to perform secondary optimization on the updated supply capacity matrix to calculate the final supplier combination. The objective function and constraints are the same as those in the initial optimization, but the calculation is based on the updated supply capacity matrix Calculate. By solving the integer programming model, the final supplier combination is obtained .

[0244] Through the integer programming algorithm, under the constraints of multiple important factors such as performance scoring, quality stability, and material commonality, the supplier combination is optimized and selected. The introduction of the supply capacity matrix ensures the matching of the supplier's supply guarantee ability and production demand, and this method improves the efficiency and accuracy of supplier selection.

[0245] Refer to Figure 2 , the second embodiment of the present invention provides a supply chain scheduling optimization system based on multi-source data, including:

[0246] A data acquisition module for acquiring the procurement data, delivery records, detailed attributes, process flows, and historical sales data of suppliers, where the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data, and risk event records;

[0247] A raw material commonality analysis module for extracting raw material information according to the procurement data and performing analysis in combination with data traceability and knowledge graphs to obtain a raw material commonality list;

[0248] A quality stability analysis module for predicting the quality stability value using a linear regression algorithm according to the delivery records and the detailed attributes;

[0249] A performance evaluation module for comprehensively evaluating the performance of suppliers according to the detailed attributes and the raw material commonality list to obtain a performance score;

[0250] A process similarity analysis module for calculating the process similarity of different products using a text similarity algorithm according to the process flows and generating a process similarity matrix;

[0251] A market demand forecasting module for constructing a market demand forecasting model and performing market demand forecasting according to the process similarity matrix and historical sales data to obtain a market demand forecasting value;

[0252] An association network generation module for generating a product association network diagram using a graph neural network according to the raw material commonality list, the process similarity matrix, and the market demand forecasting value;

[0253] The production scheduling plan optimization module is used to find a balance between the delivery time and resource utilization rate according to the product association network diagram by using the Pareto optimization method, and take the production scheduling combination corresponding to the balance point as the optimized production scheduling plan;

[0254] The supplier combination module is used to optimize the supplier combination by using the integer programming algorithm according to the performance score, the quality stability value, the raw material sharing list and the optimized production scheduling plan, and obtain the supplier combination plan.

[0255] It should be noted that the supply chain production scheduling optimization device based on multi-source data provided in the embodiment of the present invention is used to execute all the process steps of the supply chain production scheduling optimization method based on multi-source data in the above embodiment. The working principles and beneficial effects of the two correspond one by one, so they will not be elaborated here.

[0256] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a supply chain production scheduling optimization program based on multi-source data. When the processor executes the computer program, it implements the steps in each of the above embodiments of the supply chain production scheduling optimization method based on multi-source data, such as Figure 1 the step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in each of the above device embodiments, such as the supply chain production scheduling optimization module based on multi-source data.

[0257] Exemplarily, the computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and this instruction segment is used to describe the execution process of the computer program in the electronic device.

[0258] The electronic device can be a computing device such as a desktop computer, a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device, and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine some components, or different components. For example, the electronic device may also include input / output devices, network access devices, a bus, etc.

[0259] The so-called processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects all parts of the electronic device using various interfaces and circuits.

[0260] The memory can be used to store the computer program and / or module. The processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0261] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0262] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative work.

[0263] The specific embodiments described above have further elaborated on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A supply chain production scheduling optimization method based on multi-source data, characterized in that, Including: Obtain the procurement data, delivery records, detailed attributes, process flows, and historical sales data of suppliers, where the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data, and risk event records; Extract raw material information from the procurement data and analyze it in combination with data traceability and knowledge graphs to obtain a raw material sharing list; Predict the quality stability value using a linear regression algorithm based on the delivery records and the detailed attributes; Conduct a comprehensive performance evaluation of the suppliers based on the detailed attributes and the raw material sharing list to obtain a performance score; Calculate the process similarity of different products using a text similarity algorithm based on the process flows and generate a process similarity matrix; Construct a market demand prediction model based on the process similarity matrix and historical sales data and conduct market demand prediction to obtain a market demand prediction value; Generate a product association network diagram using a graph neural network based on the raw material sharing list, the process similarity matrix, and the market demand prediction value; Use the Pareto optimization method to find a balance between delivery time and resource utilization based on the product association network diagram and take the production scheduling combination corresponding to the balance point as the optimized production scheduling plan; Optimize the supplier portfolio using an integer programming algorithm based on the performance score, the quality stability value, the raw material sharing list, and the optimized production scheduling plan to obtain a supplier portfolio plan; Among them, the constructing a market demand prediction model based on the process similarity matrix and historical sales data and conducting market demand prediction to obtain a market demand prediction value includes: Construct an initial market demand prediction model using a multiple linear regression method based on the process similarity matrix and historical sales data; Analyze the seasonal fluctuation characteristics of the historical sales data using a time series analysis method; Incorporate the seasonal fluctuation characteristics into the initial market demand prediction model for model parameter adjustment to obtain a market demand prediction model and calculate the market demand prediction value based on the market demand prediction model.

2. The supply chain production scheduling optimization method based on multi-source data according to claim 1, characterized in that The extracting raw material information from the procurement data and analyzing it in combination with data traceability and knowledge graphs to obtain a raw material sharing list includes: Extract raw material information from the procurement data through data crawling technology and perform normalization processing on the raw material information to obtain standardized data; Establish a raw material traceability path based on the entity relationship between the standardized data and a preset knowledge graph; Store the raw material traceability path using graph database technology to form a raw material relationship network; Identify raw material combinations with sharing relationships in the knowledge graph based on the raw material relationship network and preset raw material sharing rules and generate a standardized raw material sharing list based on the raw material combinations.

3. The supply chain production scheduling optimization method based on multi-source data according to claim 1, wherein, The predicting the quality stability value using a regression algorithm based on the delivery records and the detailed attributes includes: Extract the order quantity, delivery quantity, quality value, batch value, and season value from the delivery records and form an initial data set; Perform data cleaning and feature extraction based on the initial data set, the historical quality value, the process value, and the material characteristic value to obtain a feature data set; Using a linear regression algorithm, taking the feature dataset as input, training a linear regression model to obtain regression values; Calculating a stability index based on the regression values, the order quantity, and the season value to obtain a quality stability value.

4. The supply chain production scheduling optimization method based on multi-source data according to claim 1, wherein, The comprehensive performance evaluation of suppliers based on the detailed attributes and the raw material sharing list to obtain a performance score includes: Calculating the delivery rate and the qualified rate of suppliers based on the historical delivery data and the quality inspection records; Quantifying the production capacity and risk level of suppliers based on the production capacity data and the risk event records; Analyzing the sharing degree and consistency among suppliers based on the raw material sharing list; Using the analytic hierarchy process to establish an evaluation model, assigning weights to the delivery rate, the qualified rate, the production capacity, the risk level, the sharing degree, the consistency, the preset timeliness, and flexibility indicators to obtain weight data; Calculating the performance score based on the delivery rate, the qualified rate, the production capacity, the risk level, the sharing degree, the consistency, and the weight data.

5. The supply chain production scheduling optimization method based on multi-source data according to claim 1, characterized in that The generation of a product association network diagram using a graph neural network based on the raw material sharing list, the process similarity matrix, and the market demand forecast value includes: Calculating a construction value based on the raw material sharing list, the process similarity matrix, and the market demand forecast value; Using a graph neural network to depict the non-linear relationship of the construction value to obtain a depicted value, and generating a product association network diagram through a visualization tool based on the depicted value.

6. The supply chain production scheduling optimization method based on multi-source data according to claim 1, wherein The optimization of the supplier portfolio using an integer programming algorithm based on the performance score, the quality stability value, the raw material sharing list, and the optimized production scheduling plan includes: Determining the critical material requirements based on the performance score, the historical quality value, and the optimized production scheduling plan, and evaluating the supply guarantee ability of suppliers to form a supply ability matrix; Based on the supply ability matrix, using an integer programming algorithm, with the performance score and the optimized production scheduling plan as constraint conditions, and the material sharing and quality stability as the objective function, calculating the initial supplier portfolio; Judging whether there are suppliers with insufficient supply guarantee ability or historical quality in the initial supplier portfolio. If so, screening eligible alternative suppliers from the supplier library according to the criteria of high score and stable quality, and updating the supply ability matrix; Using an integer programming algorithm to perform secondary optimization on the updated supply ability matrix, calculating the final supplier group to obtain a supplier portfolio plan.

7. A supply chain production scheduling optimization system based on multi-source data, characterized in that, For implementing the supply chain production scheduling optimization method based on multi-source data as described in any one of claims 1 to 6, including: A data acquisition module for acquiring the procurement data, delivery records, detailed attributes, process flows, and historical sales data of suppliers, where the detailed attributes include historical quality values, process values, material characteristic values, historical delivery data, quality inspection records, production capacity data, and risk event records; A raw material sharing analysis module for extracting raw material information based on the procurement data and analyzing it in combination with data traceability and knowledge graphs to obtain a raw material sharing list; A quality stability analysis module, which is used to predict the quality stability value by using a linear regression algorithm according to the delivery record and the detail attributes; A performance evaluation module, which is used to comprehensively evaluate the performance of suppliers according to the detail attributes and the raw material sharing list to obtain a performance score; A process similarity analysis module, which is used to calculate the process similarity of different products by using a text similarity algorithm according to the process flow and generate a process similarity matrix; A market demand prediction module, which is used to construct a market demand prediction model and conduct market demand prediction according to the process similarity matrix and historical sales data to obtain a market demand prediction value; An association network generation module, which is used to generate a product association network diagram by using a graph neural network according to the raw material sharing list, the process similarity matrix and the market demand prediction value; A production scheduling plan optimization module, which is used to find a balance between the delivery time and the resource utilization rate by using the Pareto optimization method according to the product association network diagram and take the production scheduling combination corresponding to the balance point as the optimized production scheduling plan; A supplier combination module, which is used to optimize the supplier combination by using an integer programming algorithm according to the performance score, the quality stability value, the raw material sharing list and the optimized production scheduling plan to obtain a supplier combination plan.

8. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the supply chain production scheduling optimization method based on multi-source data as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the supply chain production scheduling optimization method based on multi-source data as described in any one of claims 1 to 6.

Images