Multidimensional data analysis method and system based on manufacturing value chain joint large model

Abstract

The present invention provides a multidimensional data analysis method and system based on a joint large model of the manufacturing value chain, comprising: sending a collaborator's large model to a partner company; obtaining first model parameters obtained by the partner company using a supply and demand database to train the large model, and sending the results to the collaborator; obtaining second model parameters obtained by the collaborator by training a federated learning model using the average of the first model parameters, and sending the results to the partner company, for the partner company to use the second model parameters to continue training the large model trained using the supply and demand database; obtaining model indicators set by the collaborator; generating a joint large model of the manufacturing value chain based on the federated learning model that meets the model indicators; and completing a multidimensional data analysis task based on the joint large model of the manufacturing value chain. Generating the joint large model of the manufacturing value chain and completing the multidimensional data analysis task can meet the security, real-time, and personalization requirements of value chain management.

Description

Technical Field

[0001] The present invention relates to the field of computer data processing technology, and in particular to a multidimensional data analysis method and system based on a manufacturing value chain joint large model. Background Art

[0002] The value chain refers to a series of activities that transform inputs into outputs. It is a complex network composed of various participants such as raw material suppliers, manufacturers, distributors and end users. They are interconnected through various channels such as logistics, information flow and capital flow, and jointly complete the production, circulation and sales of products, creating value for products or services.

[0003] With the development of information technology, manufacturing value chain management faces more and more challenges and opportunities. Traditional centralized data processing and analysis methods can no longer meet the real-time, personalized, and security requirements of value chain management.

[0004] Federated learning is an emerging machine learning paradigm that effectively protects data privacy and security by sharing model parameters among multiple participants rather than raw data. It has broad application prospects in the manufacturing value chain, improving efficiency and transparency while protecting the privacy of businesses and individuals. Combining large models with federated learning technology can leverage global information for model training and optimization while protecting data privacy, thereby better addressing various challenges in manufacturing value chain management.

[0005] While existing technologies have made progress in addressing some issues in manufacturing value chain management, they still face shortcomings. For example, the data shared by all parties involved in the manufacturing value chain contains sensitive information such as commercial secrets and personal privacy, and traditional data sharing methods pose the risk of data leakage. Furthermore, traditional data sharing methods cannot meet the real-time, personalized, and security requirements of value chain management.

[0006] Therefore, a solution is urgently needed. Summary of the Invention

[0007] One of the purposes of the present invention is to provide a multidimensional data analysis method based on a large-scale joint model of the manufacturing value chain, which is applied to an execution server and is characterized by comprising:

[0008] Send collaborators’ large models to partner companies;

[0009] Obtaining first model parameters obtained by the partner company using the supply and demand database to train the large model, and sending the parameters to the collaborator;

[0010] Obtaining second model parameters obtained by the collaborator by training the federated learning model after averaging the first model parameters, and sending the second model parameters to the partner company for the partner company to use to continue training the large model that has been trained using the supply and demand database;

[0011] Get model metrics set by collaborators;

[0012] Generate a joint large model of the manufacturing value chain based on the federated learning model that achieves the model indicators;

[0013] Complete multi-dimensional data analysis tasks based on the joint large model of the manufacturing value chain.

[0014] Preferably, the training methods of the federated learning model include: homomorphic encryption, obfuscated circuits, differential privacy, and secure aggregation.

[0015] Preferably, the generation of a large joint model of the manufacturing value chain based on the federated learning model that achieves the model indicators includes:

[0016] For federated learning models that meet model indicators, parameter aggregation processing is performed to obtain a joint large model of the manufacturing value chain.

[0017] Preferably, the multi-dimensional data analysis task is completed based on the manufacturing value chain joint large model, including:

[0018] Based on the joint model of the manufacturing value chain, risk prediction is carried out according to the multi-dimensional data of the manufacturing value chain;

[0019] Based on the joint large model of the manufacturing value chain, risk management is carried out according to the multi-dimensional data of the manufacturing value chain;

[0020] Based on the joint big model of the manufacturing value chain, decision support is provided according to the multi-dimensional data of the manufacturing value chain.

[0021] Preferably, the risk prediction based on the manufacturing value chain joint large model and multi-dimensional data of the manufacturing value chain includes:

[0022] Based on the joint large-scale model of the manufacturing value chain, historical sales data, market development trend data, and cyclical change data in the multi-dimensional data of the manufacturing value chain are analyzed to determine customer demand data and market behavior data;

[0023] Based on the joint large model of the manufacturing value chain, the market development trend data, competitor strategy data, and macroeconomic environment data in the multi-dimensional data of the manufacturing value chain are analyzed to determine future market demand data and future product demand change data.

[0024] Preferably, the risk management based on the manufacturing value chain joint large model and multi-dimensional data of the manufacturing value chain includes:

[0025] Based on the joint large-scale model of the manufacturing value chain, the supplier status data and market status data in the multi-dimensional data of the manufacturing value chain are analyzed to determine the supplier problem data and market change data;

[0026] Based on the manufacturing value chain joint model, determine the risk impact degree data according to the supplier problem data and the market change data;

[0027] Based on the risk impact degree data, a risk warning is issued.

[0028] Preferably, the decision support based on the manufacturing value chain joint large model and multi-dimensional data of the manufacturing value chain includes:

[0029] Performing visualization processing on the supplier problem data and the market change data to obtain a visualization model;

[0030] sending the visualization model to a manager;

[0031] Based on the joint large model of the manufacturing value chain, determine risk solutions according to the supplier problem data and the market change data;

[0032] Based on the joint large-scale model of the manufacturing value chain, simulate the implementation of the risk solution and obtain simulation results;

[0033] The risk solution and the simulation result are sent to the manager.

[0034] The present invention also provides a multidimensional data analysis system based on a manufacturing value chain joint large model, comprising:

[0035] The execution server is used to perform the following steps:

[0036] Send collaborators’ large models to partner companies;

[0037] Obtaining first model parameters obtained by the partner company using the supply and demand database to train the large model, and sending the parameters to the collaborator;

[0038] Obtaining second model parameters obtained by the collaborator by training the federated learning model after averaging the first model parameters, and sending the second model parameters to the partner company for the partner company to use to continue training the large model that has been trained using the supply and demand database;

[0039] Get model metrics set by collaborators;

[0040] Generate a joint large model of the manufacturing value chain based on the federated learning model that achieves the model indicators;

[0041] Complete multi-dimensional data analysis tasks based on the joint large model of the manufacturing value chain.

[0042] The present invention has achieved the following beneficial effects:

[0043] In the present invention, the execution server only obtains the first model parameters obtained by the cooperative company using the supply and demand database to train the large model and sends them to the collaborator. The cooperative company does not need to provide additional data, which protects the data privacy of the cooperative company, avoids data leakage, and meets the security requirements in value chain management; in addition, the cooperative company trains the large model in real time to obtain the first model parameters, and the collaborator takes the average of the first model parameters and updates the federated learning model in real time. The collaborator can set the model indicators by himself. Finally, the federated learning model that meets the model indicators is used to generate a joint large model of the manufacturing value chain, complete the multi-dimensional data analysis task, and meet the real-time and personalized requirements in value chain management.

[0044] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0045] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 Flowchart of a multidimensional data analysis method based on a manufacturing value chain joint large model in an embodiment of the present invention;

[0048] Figure 2 Schematic diagram of the modeling process of the manufacturing value chain model in an embodiment of the present invention;

[0049] Figure 3 Schematic diagram of the architecture of the manufacturing value chain model in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the federated transfer learning architecture in an embodiment of the present invention;

[0051] Figure 5 This is a schematic diagram of the large-scale model architecture of the manufacturing value chain in an embodiment of the present invention;

[0052] Figure 6 Schematic diagram of a multidimensional data analysis system based on a large manufacturing value chain joint model in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0054] The embodiment of the present invention provides a multidimensional data analysis method based on the manufacturing value chain joint large model, which is applied to the execution server, such as Figure 1 As shown, including:

[0055] S1. Send the collaborator’s large model to the partner company;

[0056] S2. Obtaining first model parameters obtained by the partner company using the supply and demand database to train the large model, and sending the parameters to the collaborator;

[0057] S3. Obtaining second model parameters obtained by the collaborator by training the federated learning model after taking the average of the first model parameters, and sending the second model parameters to the partner company for the partner company to use to continue training the large model that has been trained using the supply and demand database;

[0058] S4. Obtain model indicators set by collaborators;

[0059] S5. Generate a joint large model of the manufacturing value chain based on the federated learning model that achieves the model indicators;

[0060] S6. Complete multi-dimensional data analysis tasks based on the joint large model of the manufacturing value chain.

[0061] The working principle and beneficial effects of the above technical solution are:

[0062] A large model refers to a machine learning model with large-scale parameters and complex structures. The large model sent by the collaborator to the partner company specifically refers to a large model that is pre-designed by the collaborator and has not been trained. After training, it can solve the specified task. Figure 2As shown, the collaborator provides the big model to the partner company. The partner company uses its own supply and demand database (which stores a large amount of manufacturing-related data) to train the local big model and returns the model parameters obtained from the training to the collaborator. The collaborator averages the parameters and updates the federated learning model (a neural network model). The updated model parameters (which indicate which data from the partner company the federated learning model determines is still required) are then delegated to the partner company, which then continues to train the local big model. The collaborator sets model metrics (such as convergence metrics) and uses the federated learning model that meets these metrics to generate a joint big model for the manufacturing value chain. Based on this joint big model, the multidimensional data analysis task is completed. Model parameters are configuration variables within the model, used to calculate the model output, and are an important component of the model. The first model parameter specifically refers to the model parameters obtained by the partner company after model training.

[0063] like Figure 3 As shown in the figure, in the federated learning model architecture, the collaborator first builds the federated learning model. The partner enterprise downloads the model and trains it using their local value chain database. The collaborator then encrypts the trained model parameters and returns them to the collaborator. The collaborator aggregates the encrypted data uploaded by the partner enterprise and updates the federated learning model itself. The updated model parameters are then passed to the partner enterprise for training. During training, the databases between the partners are not directly exchanged. Instead, model parameters are transferred (model parameters are internal configuration variables used to calculate the model output and are an important part of the model. During the federated learning process, the model structure between the collaborator and the partner enterprise is the same (both are large models pre-designed by the collaborator). The data transferred between the two is the model parameters. The partner enterprise sends the trained model parameters to the collaborator, who then averages the model parameters and sends the new model parameters to the partner enterprise.) Joint training is completed until the federated learning model meets the specified performance indicators. Model indicators are evaluation criteria for measuring the completion of the model task and are determined by the collaborator. For example, the accuracy rate (the proportion of correct model outputs among all model outputs) of the model in the test should exceed 80%.

[0064] Secondly, the training process of the federated learning model adopts the federated transfer learning model. Figure 4 As shown in Figure 2, in order to address the issues of low user overlap and low business duplication in the manufacturing value chain model training cooperation, the training process may need to adopt the federated transfer learning model. Assume that the sample of enterprise A is , the model parameters corresponding to the features are The sample of Company B is , the model parameters corresponding to the features are After a series of model transformations, the general representations are and The intermediate interpolation value calculated and encrypted by Enterprise A And send it to company B, while company B calculates and encrypts the intermediate interpolation value And send it to company B. Company A generates a random mask and calculate encryption losses With encrypted gradient Sent to company B, company B also generates a random mask And calculate the encrypted gradient Send to A. The enterprise will decrypt the result Send it to company B, and company B will decrypt the result and Send it to Enterprise A. Finally, Enterprise A and Enterprise B update their own model parameters respectively , , the two companies did not exchange actual data, but completed the update of model parameters.

[0065] The large model is the initially designed model and serves as the structural foundation of the federated learning model. The federated learning model is a model trained from the large model. The large model trained by a partner company using its database is a federated learning model. The model obtained by collaborators by averaging the parameters of the partner companies' models is also a federated learning model. The manufacturing value chain joint large model is derived from the aggregation of parameters from one or more federated learning models.

[0066] The execution server only obtains the first model parameters obtained by the cooperative company using the supply and demand database to train the large model and sends them to the collaborator. The cooperative company does not need to provide additional data, which protects the data privacy of the cooperative company, avoids data leakage, and meets the security requirements in value chain management. In addition, the cooperative company trains the large model in real time to obtain the first model parameters. The collaborator takes the average of the first model parameters and updates the federated learning model in real time. The collaborator can set the model indicators on his own. Finally, the federated learning model that meets the model indicators is used to generate a joint large model of the manufacturing value chain, complete the multi-dimensional data analysis task, and meet the real-time and personalized requirements in value chain management.

[0067] In one embodiment, the training method of the federated learning model includes: homomorphic encryption, obfuscated circuits, differential privacy, and secure aggregation.

[0068] Homomorphic encryption is a technology that uses decrypted ciphertext calculation results to compare with plaintext to see if they are the same, thus completing data calculations without leaking the source data.

[0069] A garbled circuit is a cryptographic protocol used to implement secure multi-party computation.

[0070] Differential privacy is a technology used to protect privacy from minor changes in the data source.

[0071] Secure aggregation is the process of multiple parameter parties collaborating to calculate the aggregate value without exposing their own private data.

[0072] These technologies are used in necessary processes in federated learning, such as parameter calculation and transmission.

[0073] In one embodiment, generating a manufacturing value chain joint model based on a federated learning model that achieves model indicators includes:

[0074] For federated learning models that meet model indicators, parameter aggregation processing is performed to obtain a joint large model of the manufacturing value chain.

[0075] The working principle and beneficial effects of the above technical solution are:

[0076] The manufacturing value chain model is composed of multiple federated learning models using model aggregation. Each federated learning model can perform a specific function, allowing the manufacturing value chain model to integrate multiple capabilities. Model aggregation can be achieved through averaging, stacking, or retraining. Averaging takes the average or weighted average of the parameters of multiple federated learning models and directly uses them as parameters for the manufacturing value chain model. Stacking connects multiple federated learning models to ensure that input is evenly distributed across them. Retraining uses the results of federated learning models to train the manufacturing value chain model, ensuring that its output closely matches the results of each federated learning model.

[0077] In one embodiment, the multi-dimensional data analysis task is completed based on the manufacturing value chain joint large model, including:

[0078] Based on the joint model of the manufacturing value chain, risk prediction is carried out according to the multi-dimensional data of the manufacturing value chain;

[0079] Based on the joint large model of the manufacturing value chain, risk management is carried out according to the multi-dimensional data of the manufacturing value chain;

[0080] Based on the joint big model of the manufacturing value chain, decision support is provided according to the multi-dimensional data of the manufacturing value chain.

[0081] The risk prediction is conducted based on the manufacturing value chain joint model and multi-dimensional data of the manufacturing value chain, including:

[0082] Based on the joint large-scale model of the manufacturing value chain, we analyze historical sales data, market development trend data, and cyclical change data in the multi-dimensional data of the manufacturing value chain to determine customer demand data and market behavior data. Historical sales data refers to all data related to the manufacturing industry that has been generated. Market development trend refers to the more obvious upward or downward trends in manufacturing-related data. Cyclic change data: data in certain industries in the manufacturing industry will have cyclical changes, such as seasonal changes. Customer demand data: for example, predicting a company's production data for the next year. Market behavior data: for example, predicting a company's main production direction for the next stage.

[0083] Based on the joint macro-model of the manufacturing value chain, we analyze market development trends, competitor strategies, and macroeconomic environment data within the multi-dimensional data of the manufacturing value chain to determine future market demand and changes in product demand. Market trends refer to significant upward or downward trends in manufacturing-related data. Competitor strategies refer to analyzing the data of other manufacturing companies to determine their production and development strategies. The macroeconomic environment refers to the manufacturing market environment and must be integrated with global economic policies.

[0084] The risk management is carried out based on the manufacturing value chain joint model and multi-dimensional data of the manufacturing value chain, including:

[0085] Based on the joint large model of the manufacturing value chain, the supplier status data and market status data in the multidimensional data of the manufacturing value chain are analyzed to determine the supplier problem data and market change data; supplier status data: data on which companies the supplier has provided and how many products in a period of time; market status data: data on the flow of various products in the market; supplier problem data: analysis of possible problems in supplier data; market change data: market fluctuations and hot spots.

[0086] Based on the joint large-scale model of the manufacturing value chain, risk impact data is determined based on the supplier problem data and the market change data; the risk impact data estimates the risk level based on the supplier and market data and indicates possible risks;

[0087] Conduct risk warning based on the risk impact data;

[0088] The manufacturing value chain joint model is based on the multi-dimensional data of the manufacturing value chain to provide decision support, including:

[0089] Performing visualization processing on the supplier problem data and the market change data to obtain a visualization model;

[0090] sending the visualization model to a manager;

[0091] Based on the joint model of the manufacturing value chain, determine risk solutions based on the supplier problem data and the market change data; risk solutions are plans to reduce or resolve risks, such as shifting production direction, reducing output this quarter, etc.;

[0092] Based on the joint large model of the manufacturing value chain, the risk solution is simulated to obtain simulation results; the risk solution is simulated, for example: predicting market data after reducing production.

[0093] The risk solution and the simulation result are sent to the manager.

[0094] The working principle and beneficial effects of the above technical solution are:

[0095] like Figure 5 Figure 2 shows the architecture of a federated big model for the manufacturing value chain. This model aggregates the parameters of multiple federated learning models, and the big model also provides parameter updates for the federated learning models. The federated big model leverages techniques such as prompt learning, multimodal representation, contextual inference, and instruction fine-tuning to achieve three functions: demand forecasting, risk management, and decision support. Demand forecasting includes three sub-functions: data analysis, market forecasting, and real-time updating. Data analysis analyzes historical sales data, market trends, and cyclical fluctuations to better understand customer needs and market behavior. Market forecasting predicts future market demand and identifies changes in product demand by considering factors such as market trends, competitor strategies, and the macroeconomic environment. Real-time updating involves real-time data collection and processing to promptly adjust and update the forecasting model to adapt to market dynamics. Risk management includes three sub-functions: risk identification, risk assessment, and risk early warning. Risk identification requires the big model to proactively identify potential risks in the value chain, including supplier issues and market fluctuations. Risk assessment involves quantitatively or qualitatively evaluating identified risks to determine their impact on the value chain. Risk early warning means issuing alerts before risks occur, enabling timely action to address them. Decision support includes three sub-functions: data visualization, scenario simulation, and decision rules. Data visualization transforms complex data into easy-to-understand graphical representations, helping managers quickly understand the current situation. Scenario simulation allows managers to simulate different decision scenarios and demonstrate their impact on the value chain. Decision rules refer to a series of rules and strategies developed within value chain management, including inventory management strategies, production scheduling rules, order processing procedures, and more, to ensure efficient operation of the value chain.

[0096] The embodiment of the present invention provides a multi-dimensional data analysis system based on the manufacturing value chain joint large model, such as Figure 6 As shown, including:

[0097] Execution server 1 is used to perform the following steps:

[0098] Send collaborators’ large models to partner companies;

[0099] Obtaining first model parameters obtained by the partner company using the supply and demand database to train the large model, and sending the parameters to the collaborator;

[0100] Obtaining second model parameters obtained by the collaborator by training the federated learning model after averaging the first model parameters, and sending the second model parameters to the partner company for the partner company to use to continue training the large model that has been trained using the supply and demand database;

[0101] Get model metrics set by collaborators;

[0102] Generate a joint large model of the manufacturing value chain based on the federated learning model that achieves the model indicators;

[0103] Complete multi-dimensional data analysis tasks based on the joint large model of the manufacturing value chain.

[0104] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A multidimensional data analysis method based on a large-scale joint model of the manufacturing value chain, applied to an execution server, characterized by: include: Send collaborators’ large models to partner companies; Obtaining first model parameters obtained by the partner company using the supply and demand database to train the large model, and sending the parameters to the collaborator; The supply and demand situation database is used to store manufacturing related data; Obtaining second model parameters obtained by the collaborator by training the federated learning model with the average of the first model parameters, and sending the second model parameters to the partner company for the partner company to use to continue training the large model that has been trained using the supply and demand database; Get model metrics set by collaborators; Generate a joint large model of the manufacturing value chain based on the federated learning model that meets the model indicators; Based on the joint large model of the manufacturing value chain, complete the multi-dimensional data analysis task; The federated learning models include multiple models, each of which has a corresponding function; generating a manufacturing value chain joint large model based on the federated learning models that meet the model indicators includes: performing parameter aggregation processing on multiple federated learning models that meet the model indicators to obtain the manufacturing value chain joint large model, including: taking an average or weighted average of the parameters of multiple federated learning models that meet the model indicators to obtain the manufacturing value chain joint large model; or connecting multiple federated learning models that meet the model indicators so that the input passes through each federated learning model in sequence to obtain the manufacturing value chain joint large model; or using the results of multiple federated learning models that meet the model indicators to train the manufacturing value chain joint large model respectively to obtain a trained manufacturing value chain joint large model; The multidimensional data analysis task based on the manufacturing value chain joint big model is completed, including: risk prediction based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain; risk management based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain; decision support based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain; The risk prediction based on the manufacturing value chain joint big model and the multi-dimensional data of the manufacturing value chain includes: analyzing the historical sales data, market development trend data, and cyclical change data in the multi-dimensional data of the manufacturing value chain based on the manufacturing value chain joint big model to determine customer demand data and market behavior data; analyzing the market development trend data, competitor strategy data, and macroeconomic environment data in the multi-dimensional data of the manufacturing value chain based on the manufacturing value chain joint big model to determine future market demand data and future product demand change data; The risk management based on the manufacturing value chain joint big model and the multi-dimensional data of the manufacturing value chain includes: analyzing the supplier status data and market status data in the multi-dimensional data of the manufacturing value chain based on the manufacturing value chain joint big model to determine supplier problem data and market change data; determining risk impact degree data based on the supplier problem data and the market change data based on the manufacturing value chain joint big model; and issuing risk warnings based on the risk impact degree data; The decision support is performed based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain, including: visualizing the supplier problem data and the market change data to obtain a visualization model; sending the visualization model to the manager; determining a risk solution based on the supplier problem data and the market change data based on the manufacturing value chain joint big model; simulating the implementation of the risk solution based on the manufacturing value chain joint big model to obtain simulation results; and sending the risk solution and the simulation results to the manager.

2. In the multidimensional data analysis method based on the manufacturing value chain joint large model according to claim 1, the training method of the federated learning model includes: Homomorphic encryption, obfuscated circuits, differential privacy, and secure aggregation.

3. A multidimensional data analysis system based on the joint large model of the manufacturing value chain, characterized by: include: The execution server is used to perform the following steps: Send collaborators’ large models to partner companies; Obtaining first model parameters obtained by the partner company using a supply and demand database to train the large model, and sending the parameters to the collaborator; the supply and demand database is used to store manufacturing-related data; Obtaining second model parameters obtained by the collaborator by training the federated learning model with the average of the first model parameters, and sending the second model parameters to the partner company for the partner company to use to continue training the large model that has been trained using the supply and demand database; Get model metrics set by collaborators; Generate a joint large model of the manufacturing value chain based on the federated learning model that meets the model indicators; Based on the joint large model of the manufacturing value chain, complete the multi-dimensional data analysis task; The federated learning models include multiple models, each of which has a corresponding function; generating a manufacturing value chain joint large model based on the federated learning models that meet the model indicators includes: performing parameter aggregation processing on multiple federated learning models that meet the model indicators to obtain the manufacturing value chain joint large model, including: taking an average or weighted average of the parameters of multiple federated learning models that meet the model indicators to obtain the manufacturing value chain joint large model; or connecting multiple federated learning models that meet the model indicators so that the input passes through each federated learning model in sequence to obtain the manufacturing value chain joint large model; or using the results of multiple federated learning models that meet the model indicators to train the manufacturing value chain joint large model respectively to obtain a trained manufacturing value chain joint large model; The multidimensional data analysis task based on the manufacturing value chain joint big model is completed, including: risk prediction based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain; risk management based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain; decision support based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain; The risk prediction based on the manufacturing value chain joint big model and the multi-dimensional data of the manufacturing value chain includes: analyzing the historical sales data, market development trend data, and cyclical change data in the multi-dimensional data of the manufacturing value chain based on the manufacturing value chain joint big model to determine customer demand data and market behavior data; analyzing the market development trend data, competitor strategy data, and macroeconomic environment data in the multi-dimensional data of the manufacturing value chain based on the manufacturing value chain joint big model to determine future market demand data and future product demand change data; The risk management based on the manufacturing value chain joint big model and the multi-dimensional data of the manufacturing value chain includes: analyzing the supplier status data and market status data in the multi-dimensional data of the manufacturing value chain based on the manufacturing value chain joint big model to determine supplier problem data and market change data; determining risk impact degree data based on the supplier problem data and the market change data based on the manufacturing value chain joint big model; and issuing risk warnings based on the risk impact degree data; The decision support is performed based on the manufacturing value chain joint big model and the multidimensional data of the manufacturing value chain, including: visualizing the supplier problem data and the market change data to obtain a visualization model; sending the visualization model to the manager; determining a risk solution based on the supplier problem data and the market change data based on the manufacturing value chain joint big model; simulating the implementation of the risk solution based on the manufacturing value chain joint big model to obtain simulation results; and sending the risk solution and the simulation results to the manager.

Images