CN116976948A - Method and system for generating dynamic feedback flow diagram of full value chain of manufacturing enterprise - Google Patents

Abstract

The invention discloses a method and a system for generating a dynamic feedback flow graph of a full value chain of a manufacturing enterprise, wherein the method comprises the steps of obtaining a value added data source from a database of an enterprise intelligent system; extracting data information from the value-added data source, and preprocessing the data; defining the data information after data preprocessing to obtain abstract description of the data information; the data information of abstract description is configured in a dynamic feedback link of a full value chain of a manufacturing enterprise, matched with corresponding indexes, subjected to data association and predicted through a neural network; and performing value chain mapping on the value added data source to generate a dynamic feedback flow diagram of the full value chain of the manufacturing enterprise, and storing the dynamic feedback flow diagram in a database. The invention generates the dynamic feedback flow diagram of the full value chain of the manufacturing enterprise, can reasonably plan the supply/production time and period, improves the horizontal efficiency of the supply chain, reduces the raw material redundancy, realizes the stock optimization, reduces the production cost, and also provides a new solution idea for the business process link optimization of the enterprise.

Description

A method and system for generating a dynamic feedback flow diagram for the entire value chain of a manufacturing enterprise

Technical field

The invention belongs to the technical field of intelligent management of the entire value chain of manufacturing enterprises, and specifically relates to a method and system for generating a dynamic feedback flow diagram for the entire value chain of manufacturing enterprises based on neural network prediction and inventory level control.

Background technique

With the rapid development of Internet technology and the widespread application of smart equipment, a new supply chain model marked by "digitization + intelligence" is rapidly emerging. Lean supply chain treats all activities in the supply chain as manufacturing enterprise production activities. It is an organic part of the process, through planning unified information and information sharing, coordinating and integrating all activities in the process in planning, transportation, production, storage, distribution and other links; in the era of networked sales, customer demand is the starting point, and the supply chain network is the consumer Support, this makes the supply chain under the network model more timely and responsive than the traditional supply chain.

Compared with the traditional supply chain structure, the network supply chain structure has many network members as a whole. It is necessary to improve the service awareness of network members and enhance the coordination relationship between resources; sales companies can rely on them in the supply chain production environment. Big data predicts customer behavior, analyzes customer needs based on the prediction results and makes corresponding production deployments. Suppliers provide timely supplies. In each value chain process, the factors that can affect supply, production and sales are analyzed. Collection can be used to predict demand orders; in the entire supply chain process, information can be exchanged in all links, and personalized services can be provided throughout the circulation process from parts production, product production to sales, while ensuring supply. The entire process of production and sales activities runs safely and smoothly.

Contents of the invention

The purpose of this invention is to provide a method and system for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise, which predicts and inventory levels through neural networks for a series of activities generated by enterprise supply, production, marketing, product recycling, service links, etc. Adjustment to weaken the impact of bullwhip effect on inventory level fluctuations of enterprises in various links can not only reasonably plan supply and production time and cycle, improve supply chain level efficiency, reduce raw material redundancy, but also achieve inventory optimization and reduce production costs. Provide new solutions for the optimization of enterprise business process links.

In order to achieve the above objects, the present invention provides the following technical solutions:

The invention provides a method for generating a dynamic feedback flow diagram of the full value chain of a manufacturing enterprise, which includes the following steps:

S1: Collect value-added data from the database of the enterprise's intelligent system, form a value-added data source, and store the data;

S2: Extract data information from value-added data sources and perform data preprocessing on the extracted data information;

S3: Define the data information after data preprocessing and obtain an abstract description of the data information;

S4: Configure the abstractly described data information in the dynamic feedback link of the manufacturing enterprise's full value chain, match it with the corresponding indicators in the dynamic feedback link of the manufacturing enterprise's full value chain, perform data association, and then make predictions through neural networks;

S5: Perform value chain mapping on value-added data sources to generate a dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise, and store it in the database.

Preferably, the enterprise intelligent system includes: customer management system, marketing service management system, enterprise resource management system, warehousing logistics management system, enterprise financial management system, production process control system, production process execution management system, and product recycling management system. , product after-sales service system, product testing management system.

Preferably, collecting the value-added data includes: collecting various text data information, customer behavior information, email information, customer web page data information, communication data information, and production process parameters from various management systems of value-added data through data collection equipment. information and product testing information.

Preferably, the data preprocessing includes: dividing the data in the value-added data source into data types, comparing the divided data in each type with the data in the value-added data source, finding missing values, and Delete abnormal values.

Preferably, in step S3, the data information after data preprocessing is defined to obtain an abstract description of the data information, including:

The data information after data preprocessing is defined as: basic data related to the supply production system, basic data related to production from the production system, basic data related to marketing from the marketing service system, and recycling services related Basic data from the recycling service system;

Establish a unified and identifiable standardized model for data information in multiple systems, save supply indicators, production indicators, marketing indicators and recycling service indicators in the form of key values, use "keys" to distinguish data dimensions, and "values" to save them Get the value to realize the abstract description of the data information.

Preferably, in step S4, the corresponding indicators in the dynamic feedback link of the entire value chain of the manufacturing enterprise include: supply indicators related to supply, production indicators related to production, marketing indicators related to marketing, and recycling related indicators. Recycling service indicators;

In step S4, the abstractly described data information is configured in the dynamic feedback link of the full value chain of the manufacturing enterprise. The method of matching the corresponding indicators in the dynamic feedback link of the full value chain of the manufacturing enterprise includes:

Establish a data management system in the database, index and store abstractly described data information in the database through corresponding indicators in the dynamic feedback link of the manufacturing enterprise's full value chain, and complete the acquisition of dynamic feedback data of the manufacturing enterprise's full value chain;

In the step S4, the data association includes: performing correlation analysis on the customer's basic operating behavior and user needs, and then correlating the customer behavior information in the value-added data source to build a correlation feature data set;

In the step S4, prediction is made through the neural network, including: predicting the purchasing behavior of potential customers through the neural network based on the associated feature data set. In the supply chain stage, based on the retailer's historical sales data and the manufacturer's shipment history data With the manufacturer's order demand historical data and the factors that affect sales, shipments or order demand data in the supply chain, we comprehensively consider the reverse forecast of order production and sales, and combine the inventory adjustment rate of the supply chain enterprise as a selection balancing variable. Determine the expected supply order volume in the supply link, the expected production volume in the production link, and the expected order quantity in the sales link.

Preferably, predicting the purchasing behavior of potential customers through neural networks includes the following steps:

(1) Customer operation data preprocessing: After performing data domain processing on the value-added data source, and then performing feature selection, the customer's basic operation process and different operation behaviors in the customer behavior information are associated with the customer needs in the customer behavior information. Analysis, then construct a feature data set, and use the correlation coefficient method to select the optimal feature subset;

(2) Define fitness: Use the mean square error of the predicted value of the LSTM neural network as the fitness value fit of the particle, where:

;

Among them, y is the real value, is the expected output value;

(3) Use the position information of particles as parameters of the LSTM neural network and construct an LSTM neural network;

(4) Train the LSTM neural network: update the individual extreme value and group extreme value of the particle according to the fitness value of each particle;

(5) Use nonlinear inertial weights to iteratively update the particle’s speed and position information based on the particle’s individual extreme value and group extreme value;

(6) When the speed and position information of the particles meet the conditions or reach the maximum number of iterations, enter the next step to obtain the optimized parameters, otherwise return to step (3);

(7) After obtaining the optimized parameters, increase the number of iterations, retrain the LSTM neural network, and obtain the trained LSTM neural network;

(8) Based on the optimal feature subset, predict customer purchasing behavior through the trained LSTM neural network.

Preferably, the selection method of the optimal feature subset is:

First, the Pearson coefficient is used as the selection index of the feature subset, and the parameter a is selected as the threshold in the feature subset selection process. Based on the value-added data, the feature pairs whose absolute value of correlation is less than the parameter a are found. Based on the selected feature pairs Compare the correlation of these features to the feature data set respectively, remove feature pairs with small correlations, and obtain a new feature subset;

The evaluation function is used to evaluate the generated feature subset to obtain the evaluation result. The index evaluated by the feature subset is compared with the evaluation result. If the index evaluated by the feature subset is greater than the evaluation result, the evaluation result and the optimal feature subset are updated. , otherwise, stop the algorithm and output the current optimal feature subset;

Optimize the LSTM neural network model parameters based on the optimal feature subset;

Predict customer demand for different products at the same time interval based on corresponding indicators in the dynamic feedback link of the entire value chain of manufacturing enterprises.

The present invention also provides a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise generated by the above method. The dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise includes: a dynamic feedback flow diagram visual editor and a configuration page, and a dynamic feedback flow diagram. Flow graph variable types and connection models, database systems and real-time interfaces.

The invention also provides a data sharing integrated management platform for dynamic feedback flow diagrams of the entire value chain of manufacturing enterprises, including:

Supply chain enterprise business collaboration service platform, supply chain enterprise business collaboration platform server, manufacturing enterprise A supplier group, supplier enterprise internal server, manufacturing enterprise group, manufacturing enterprise internal server, manufacturing enterprise B dealer group and recycler group;

The supply chain enterprise business collaboration service platform is communicatively connected to the supply chain enterprise business collaboration platform server;

The internal server of the supplier enterprise is located within the supplier group of manufacturing enterprise A;

The internal server of the manufacturing enterprise is located within the manufacturing enterprise group;

The manufacturing company A supplier group can access the recycler group, the manufacturing company group can access the manufacturer company B dealer group and the recycler group, the recycler group and the manufacturing company B dealer group can only access the supply chain The enterprise business collaboration service platform accesses business data, and all access requests have specific access rights.

The invention also provides an analysis method for the dynamic feedback flow diagram of the full value chain of a manufacturing enterprise. The analysis method is based on the dynamic feedback flow diagram of the full value chain of a manufacturing enterprise and includes the following steps: data collection and classification, data processing, and data query. and storage, data visualization and analysis;

The data collection refers to obtaining direct data from various data systems, uniformly receiving and integrating it into value-added data sources, and inputting different types of data information into a unified data storage space;

The data processing refers to defining and describing the data obtained during the collection process;

The data query and storage refers to storing the processed data in the data space and establishing an index table structure;

The data visualization and analysis refers to graphicalizing and displaying the data after the dynamic feedback results are transmitted to form a dynamic feedback value chain flow diagram, and then rationally selecting the data in the data space to form historical data in the supply chain process. Warehouse, and based on the direct data obtained in each link system and the historical data stored in the data, the data association is determined and displayed through process indicators, data docking and functional association, and then used for calculation of historical data to analyze the entire value chain process.

It can be seen from the above technical solutions that compared with the existing technology, the technical effects of the present invention are:

Based on the dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise, this invention establishes data correlation between the customer's operational behavior and customer needs, and predicts the purchase intention of potential customers by training the LSTM neural network. Based on the prediction results, the manufacturer will The production volume in one cycle and the supplier's order volume in the next cycle visualize the dynamic data changes in the business links of the full value chain, and integrate the inventory levels and market demand of suppliers, manufacturers, dealers and recyclers. Key indicators such as quantity are displayed to understand the operation status of each business activity in real time and maintain the management of supply chain inventory levels; the dynamic feedback flow chart of the entire value chain of the manufacturing enterprise of the present invention can flexibly select important indicators. It reflects the resource flow and value direction in each link of supply, production, marketing services, and recycling services, monitors the business data of each business link, and visually displays the dynamic feedback flow diagram of the entire value chain through the terminal. At this time, the dynamic feedback flow diagram can be used to analyze the value chain links at any time, and find the problem in a targeted manner, so that managers can adjust the strategies of supply, production, and marketing services.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only These are embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without exerting creative efforts.

The method for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise according to the present invention will be further described below in conjunction with the accompanying drawings;

Figure 1 is a schematic structural diagram of the dynamic feedback flow diagram data sharing integrated management platform for the entire value chain of a manufacturing enterprise according to the present invention;

Figure 2 is a flow chart of the supply chain analysis method of dynamic feedback flow chart data of the entire value chain of a manufacturing enterprise according to the present invention;

Figure 3 is a schematic diagram of the method for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise according to the present invention;

Figure 4 is a schematic diagram of the prediction process of dynamic feedback of the entire value chain of the manufacturing enterprise based on the trained LSTM neural network according to the present invention;

Figure 5 is an illustration of the cause and effect relationship of the dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise of the present invention;

Figure 6 is a schematic diagram of the usage interface of the shared integrated management platform of the dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise of the present invention.

Detailed ways

Specific implementations of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the invention but are not intended to limit the scope of the invention.

Figure 1 shows the dynamic feedback value chain data sharing and integrated management method of the manufacturing enterprise of the present invention. For the clustered alliance form of manufacturing enterprises, the enterprise can interact with the database of the platform through its own internal database and through the data exchange program, and then interact with the enterprise. The business collaboration service platform connects data.

Figure 2 is a flow chart of the present invention's dynamic feedback data supply chain analysis method for the entire value chain of a manufacturing enterprise, including data collection, data query and storage, and data visualization and analysis.

As shown in Figure 3, the embodiment of the present invention discloses a method for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise, which includes the following steps:

S1: Collect value-added data from the database of the enterprise's intelligent system, form a value-added data source, and store the data;

S2: Extract data information from value-added data sources and perform data preprocessing on the extracted data information;

S3: Define the data information after data preprocessing and obtain an abstract description of the data information;

S4: Configure the abstractly described data information in the dynamic feedback link of the manufacturing enterprise's full value chain, match it with the corresponding indicators in the dynamic feedback link of the manufacturing enterprise's full value chain, perform data association, and then make predictions through neural networks;

S5: Perform value chain mapping on value-added data sources to generate a dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise, and store it in the database.

Among them, in step S1, the enterprise intelligent system includes: customer management system, marketing service management system, enterprise resource management system, warehousing logistics management system, enterprise financial management system, production process control system, production process execution management system, product Recycling management system, product after-sales service system, product testing management system.

Collecting the value-added data includes: collecting various text data information, customer behavior information, email information, customer web page data information, communication data information, production process parameter information and products from various management systems of value-added data through data collection equipment. detection information.

Data preprocessing includes: dividing the data in the value-added data source into data types, comparing the data in each divided type with the data in the value-added data source, finding missing values, and deleting abnormal values.

In step S3, the data information after data preprocessing is defined to obtain an abstract description of the data information, including:

The data information after data preprocessing is defined as: basic data related to the supply production system, basic data related to production from the production system, basic data related to marketing from the marketing service system, and recycling services related Basic data from the recycling service system;

Establish a unified and identifiable standardized model for data information in multiple systems, save supply indicators, production indicators, marketing indicators and recycling service indicators in the form of key values, use "keys" to distinguish data dimensions, and "values" to save them Get the value to realize the abstract description of the data information.

In order to achieve the above technical solution, the data collection described in the present invention is an interface that collects data from outside the system and inputs it into the system. The data collection system integrates signals, sensors, actuators, data collection equipment and application software; in the network The collaborative manufacturing platform obtains direct data from various data systems and integrates it into a value-added data source storage medium that uses a large number of information; the data source initially comes from various aspects of the network operation front-end page, production plants, financial statements, and supply recovery. Available in various systems and modules:

The data obtained from these systems are uniformly received and integrated into the data sources of the system of the present invention. The data sources are used to drive the visual display of information data; standardized systems will provide built-in value-added data sources and event target sources, which can be directly adapted Use it to obtain data within a set time period. For non-standardized systems, support expansion needs to be carried out in the form of specifications.

The storage of data sources requires clear identification settings; identification includes supply capacity, sales attributes, material production cycle, equipment utilization, standard production cycle, equipment change time, personnel, shift working time, equipment name and number, customer base Information, distribution status, product market sales quantity, market sales fluctuation, customer purchasing power, product discount sensitivity, delivery time tolerance; supply capacity represents the output efficiency of raw material production and processing of the supplier company; sales attributes refer to the supply of goods to consumers Selected options and dimensions, such as product color, package and size settings;

Material production cycle refers to the time consumed in producing each material; equipment utilization refers to the ratio of equipment usage time to theoretical production time; standard production cycle refers to the theoretical product processing and production rate; equipment changeover time refers to The time it takes for the equipment to switch from producing one component to producing other components;

Personnel refers to the number of people who can actually participate in the work during a cycle of duty; shift working time refers to the production time of a factory cycle specified by the enterprise, that is, the interval between two adjacent shifts. According to the common practice on the production line, it is generally 8 hours to Within 12 hours, arrangements can be made according to actual conditions;

Equipment name and number refer to the identifying information of the equipment; basic customer information includes records of basic information such as the customer's address, name, age or gender; delivery status refers to delivery information such as the time and location of the product;

Product market sales quantity refers to the total quantity of product online marketing and offline marketing; market sales fluctuation refers to the impact of natural factors or other objective factors on product sales within a period of time; customer purchasing power refers to the purchase price that customers can accept. ;

Product discount sensitivity refers to the attractiveness of a company's product discounts to customers; delivery time tolerance refers to the acceptance of the timeliness of goods being delivered to customers;

The described data processing involves defining the collected data after obtaining it from the data source, cleaning the data and other operations to maintain the standardized format of the data;

Finally, the data sources are abstracted. Initially, the basic data related to supply and production system are specified, the basic data related to production are from the production system, the basic data related to marketing services are from the marketing service system, and the basic data related to recycling services are from Basic data of recycling service system.

By establishing data specifications and standards, we describe supply indicators, production indicators, marketing service indicators, recycling service indicators, etc., and configure the data sources in the dynamic feedback flow diagram of the full value chain of manufacturing enterprises; use these data sources to implement data-driven Real-time dynamic feedback flow graph analysis solution. The data query and storage refers to storing data in a data storage medium as historical data for the retention of dynamic feedback processes on a time scale;

The data query refers to establishing an index table and storing the processed data in the data medium. The data visualization and analysis refers to visualizing the initial data in the form of charts and other forms, and then graphically displaying the data after the dynamic feedback results are transmitted in the same window, and dynamically analyzing the changes in indicator parameters in each link in real time to form full value. Dynamic feedback flow diagram of chain links.

As shown in Figure 5 and Figure 6, in order to optimize the above technical solution, the basic data related to the supply production system in the present invention include: procurement (inquiry, contract, order, bidding), supplier management Scope, basic information on raw materials, inventory type, supply capacity, sales attributes, direct transportation costs, indirect transportation costs, quality data files, service levels; basic data related to production from the production system include: equipment selection, Start-up, operation, downtime and number of personnel, efficiency, facilities, equipment, tools, materials, consumables, direct operating costs, indirect operating costs, output of production equipment, process conditions and quality inspection data files.

Basic data related to marketing services from the marketing service system include market sales discounts, inventory, sales volume, market fluctuations, product price forecasts, distribution, product performance, basic customer information, shopping behavior, web product browsing, product evaluation sensitivity , customer purchasing power, product promotion sensitivity, delivery efficiency acceptance.

Basic data related to recycling services from the recycling service system include: market recycling level, product durability, product remanufacturing technology level, metal content ratio required for recycled products, inventory, product testing and repair, dismantling, products Recycling types, quantities, recycling costs, operating costs, and business qualifications of enterprises.

The abstraction refers to establishing a unified identifiable standardized model of data in multiple systems, saving supply indicators, production indicators, marketing services and recycling indicators in the form of key values, using "keys" to distinguish data dimensions, and "values". "Save its value.

Supply indicators include inventory levels, expected production quantities, product production orders, parts productivity, and supply capabilities.

Production indicators include production quantity, inventory quantity, delayed delivery, product sales rate forecast, product shipment rate, expected production rate, actual production capacity, and product qualification rate. Marketing service indicators include brand promotion, cooperation compatibility, assets and liabilities, marketing capabilities, product services, customer service, credibility and honor, and social responsibility.

The customer's product operation process includes: browsing products, collecting products, adding products to the shopping cart, submitting orders, and paying orders. By correlating the customer's basic operational behavior with customer needs, and then correlating the customer behavior data in the value-added data source, a correlated feature data set is constructed;

Finally, the purchasing behavior of potential customers is predicted through the LSTM neural network; in the supply chain stage, based on the retailer's historical sales data, the manufacturer's shipment history data and the manufacturer's order demand history data, it affects sales and shipments in the supply chain. Or other factors of order demand data, comprehensively consider reverse forecasting order production and sales, and combine the inventory adjustment rate of supply chain enterprises as a selection balancing variable to determine the expected supply order volume in the supply link and the expected production volume in the production link. The expected order quantity in the sales process;

When it is necessary to analyze the impact of market customer demand changes on time factors during the production process, the customer-related feature data set is called from the data warehouse; the mathematical model used to extract the most relevant data with supply, production, sales, product recycling and other indicators is used Analyze and predict the inventory levels of each link in the value chain to guide the smooth operation of the supply chain.

In step S4, the corresponding indicators in the dynamic feedback link of the entire value chain of the manufacturing enterprise include: supply indicators related to supply, production indicators related to production, marketing indicators related to marketing, and recycling service indicators related to recycling. ;

In step S4, the abstractly described data information is configured in the dynamic feedback link of the full value chain of the manufacturing enterprise. The method of matching the corresponding indicators in the dynamic feedback link of the full value chain of the manufacturing enterprise includes:

Establish a data management system in the database, index and store abstractly described data information in the database through corresponding indicators in the dynamic feedback link of the manufacturing enterprise's full value chain, and complete the acquisition of dynamic feedback data of the manufacturing enterprise's full value chain;

In the step S4, the data association includes: performing correlation analysis on the customer's basic operating behavior and user needs, and then correlating the customer behavior information in the value-added data source to build a correlation feature data set;

In the step S4, prediction is made through the neural network, including: predicting the purchasing behavior of potential customers through the neural network based on the associated feature data set. In the supply chain stage, based on the retailer's historical sales data and the manufacturer's shipment history data With the manufacturer's order demand historical data and the factors that affect sales, shipments or order demand data in the supply chain, we comprehensively consider the reverse forecast of order production and sales, and combine the inventory adjustment rate of the supply chain enterprise as a selection balancing variable. Determine the expected supply order volume in the supply link, the expected production volume in the production link, and the expected order quantity in the sales link.

In step 4, the matching method refers to establishing a data management system, using the database to establish data storage and indexing, and completing the acquisition of dynamic feedback data for the entire value chain of manufacturing enterprises. The customer's product operation process at the marketing service terminal includes: browsing products, collecting products, adding products to the shopping cart, submitting orders, and paying orders. By correlating the customer's basic operational behavior with user needs, and then correlating the customer behavior data in the value-added data source, a correlated feature data set is constructed. Finally, the purchasing behavior of potential customers is predicted through neural network.

As shown in Figure 2, to further optimize the above technical solution, the steps of the analysis process based on the trained LSTM neural network for customer operating behavior and dealer expected order quantity, manufacturer expected productivity, and supplier supply order prediction include:

Step 1: Data acquisition; related to prediction is mainly product attribute information and some customer behavior data; product attribute information includes: sales date, product brand, sales season, category, price, original price, product quantity, sales days, etc.; There are five main types of customer operations: browsing products, collecting products, adding products to shopping carts, submitting orders, and paying orders. Among them, collection operations are an important feature for demand forecasting and are generally highly correlated with the demand for final products; these operations will be recorded in digital persistence devices to form a complete customer operation behavior data set. , can be used to predict the demand for different products; customer operation data can record the number of clicks and views on the product, the number of collections, the number of times added to the shopping cart, the number of shares, and product sales.

Step 2: Feature selection; analyze the customer's basic operating process and the connection between different operating behaviors in the process and customer needs, that is, the main features related to the needs are the operations performed by the customer, and then analyze the customer behavior data through the data set, Obtain the transformation relationship between customers' different operating behaviors and customer needs;

For example, the conversion rate is the number of product sales/the number of executions of each operation. By analyzing the conversion rate, we can derive a simple linear relationship between demand and customer operation behavior, and infer whether these customer operation behaviors affect demand. Important features can complete this basic feature selection process.

Step 3: Data processing; there are many missing values and unreasonable data in the original data, so the data set needs to be cleaned, denoised and other operations. For data with missing values, zero padding or averaging of adjacent periods can be used to ensure the continuity of the data and avoid the situation where a large number of data samples become unusable. For unreasonable data, for example, malicious brushing leads to high sales but few customer clicks, shopping carts, and collections; or the data type does not match and the data value exceeds the normal expected range. Data elimination is used to deal with these phenomena.

Step 4: Construct associated features; the purpose of selecting the optimal feature subset is to eliminate redundant features that are irrelevant to the prediction target, reduce the dimension of the data set, and improve the efficiency of model training; based on the features of the original data set, use Correlation methods construct more correlation features.

For example, combining the original features of the original data set or using statistical methods to construct valuable new features: for example, the purchase operation can reflect the sales volume of the product, and the maximum value, minimum value and standard deviation of the product sales volume can be counted separately, which can be objective Reflect the sales situation and degree of dispersion; the sliding window method can be used. For each product, the relevant characteristics can be calculated within the time range of 1, 3, 5, 7, and 14 days before it, by changing the window length and sliding step size. New features can be obtained in different time frames. The optimal feature subset is then selected, but the data is normalized before finding the optimal subset. Because normalization can solve the difference in value range between different features, so that each feature is of the same order of magnitude, it is suitable for comprehensive comparative evaluation; but the biggest benefit of normalization is that it can make the process of finding the optimal solution significantly smoother. , it is easier to converge to obtain the optimal solution; after normalization processing, the correlation between different features and between different features and the target variable can be calculated and analyzed, and the correlation between different features can be analyzed. Eliminate redundant features. The method is to use the Pearson correlation coefficient method to prepare for subsequent model training.

Step 5: Select a demand forecast model based on LSTM, and use the improved particle swarm algorithm to optimize network parameters.

Step 6: Model prediction. The selected evaluation indicators are used to predict the demand for different products with the same time interval.

In the forecasting process of the value chain, the sales volume forecast at the retail end is combined with the historical sales data of the network platform and the click rate, browsing rate, transaction rate and other data of potential customers to predict, and finally obtain the sales volume forecast of the retailer; the enterprise based on Forecast the sales volume and set the upper limit of the expected inventory, and set the deadline. The difference between the expected inventory and the actual inventory is compared with the predicted quantity for replenishment, and the maximum quantity is determined as the order quantity for the next period, and is reported to the upper level of the supply chain. Make a pre-order.

The manufacturer's sales forecast is combined with the historical data of shipments. The expected inventory value is obtained by using the predicted data and setting the inventory duration period. The difference between the expected inventory and the actual inventory is compared with the expected inventory. The maximum value is the quantity of product expected to be produced;

The supplier's sales forecast is combined with the manufacturer's historical production and ordering data. The expected inventory value is obtained through the predicted data and the inventory duration period is set. The difference between the expected inventory and the actual inventory is compared with the expected inventory. Take the maximum value as the quantity of parts expected to be supplied by the supplier;

Recycling by recyclers includes parts recycling and product recycling. According to the sales rate of the sales market and the market changes of recycled products, the product recycling proportions in different cycles are set, and the recycled products are added to the company's inventory for further processing. Reproduction and utilization;

The marketing service part shifts from the service profit chain to the service value chain. Its main function is to promote customer service from the aspects of employee satisfaction, internal corporate quality, and customer satisfaction, develop potential customers, and convert potential customers into formal purchases of products. customers, thus driving product sales by retailers, product production by enterprises, and supply of parts by suppliers.

As shown in Figure 4, predicting the purchasing behavior of potential customers through neural networks includes the following steps:

(1) Customer operation data preprocessing: After performing data domain processing on the value-added data source, and then performing feature selection, the customer's basic operation process and different operation behaviors in the customer behavior information are associated with the customer needs in the customer behavior information. Analysis, then construct a feature data set, and use the correlation coefficient method to select the optimal feature subset;

(2) Define fitness: Use the mean square error of the predicted value of the LSTM neural network as the fitness value fit of the particle, where:

;

Among them, y is the real value, is the expected output value;

(3) Use the position information of particles as parameters of the LSTM neural network and construct an LSTM neural network;

(4) Train the LSTM neural network: update the individual extreme value and group extreme value of the particle according to the fitness value of each particle;

(5) Use nonlinear inertial weights to iteratively update the particle’s speed and position information based on the particle’s individual extreme value and group extreme value;

(6) When the speed and position information of the particles meet the conditions or reach the maximum number of iterations, enter the next step to obtain the optimized parameters, otherwise return to step (3);

(7) After obtaining the optimized parameters, increase the number of iterations, retrain the LSTM neural network, and obtain the trained LSTM neural network;

(8) Based on the optimal feature subset, predict customer purchasing behavior through the trained LSTM neural network.

Specifically, the selection method of the optimal feature subset is:

First, the Pearson coefficient is used as the selection index of the feature subset, and the parameter a is selected as the threshold in the feature subset selection process. Based on the value-added data, the feature pairs whose absolute value of correlation is less than the parameter a are found. Based on the selected feature pairs Compare the correlation of these features to the feature data set respectively, remove feature pairs with small correlations, and obtain a new feature subset;

The evaluation function is used to evaluate the generated feature subset to obtain the evaluation result. The index evaluated by the feature subset is compared with the evaluation result. If the index evaluated by the feature subset is greater than the evaluation result, the evaluation result and the optimal feature subset are updated. , otherwise, stop the algorithm and output the current optimal feature subset;

Optimize the LSTM neural network model parameters based on the optimal feature subset;

Predict customer demand for different products at the same time interval based on corresponding indicators in the dynamic feedback link of the entire value chain of manufacturing enterprises.

The LSTM neural network prediction link also includes: manufacturers and suppliers;

Predict the manufacturer's expected product quantity through the trained LSTM neural network: predict through the manufacturer's sales volume, combined with the historical data of the manufacturer's shipment volume, through the LSTM neural network, use the predicted data and set the inventory duration Period, obtain the expected inventory quantity, compare the difference between the expected inventory quantity and the actual inventory quantity, and the expected inventory quantity, and take the maximum value as the manufacturer's expected product quantity;

Predict the supplier's expected supply quantity through the trained LSTM neural network: the supplier's sales chain is combined with the manufacturer's historical production and order data to predict through the LSTM neural network, and the expected inventory is obtained through the predicted data and setting the inventory duration period Quantity, the difference between the expected inventory quantity and the actual inventory quantity is compared with the expected inventory quantity, and the maximum value is taken as the quantity expected to be supplied by the supplier.

The present invention also provides a method for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise through the above method. The dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise includes: a dynamic feedback flow diagram visual editor and a configuration page. Feedback flow graph variable types and connection models, database systems and real-time interfaces.

The visual editor and configuration page of the manufacturing enterprise's full value chain dynamic feedback flow diagram are used by professional analysts to customize the element indicators, page layout, connection mode, icon style, and visual parameter display of the full value chain dynamic feedback. The dynamic feedback flow diagram data of the whole value chain is combined with the data of the dynamic feedback output of the actual supply, production, marketing services and recycling service links, associates the data in the actual links with event indicators, and links the corresponding processes with the dynamic feedback of the whole value chain. Flow graph functions are bound. The full value chain dynamic feedback real-time interface is used to display the dynamic changes of indicators on the real-time interface, and supports the operation and correlation analysis of data.

The present invention also provides a method for analyzing the dynamic feedback flow diagram of the full value chain of a manufacturing enterprise generated by the above method. The method is based on the dynamic feedback flow diagram of the full value chain of a manufacturing enterprise and includes the following steps: data collection and classification, Data processing, data query and storage, data visualization and analysis.

The data collection refers to obtaining direct data from various data systems, uniformly receiving and integrating it into value-added data sources, and inputting different types of data information into a unified data storage space;

The data processing refers to the definition, description and standardization of the data obtained during the collection process;

The data query and storage refers to storing the processed data in the data space and establishing an index table structure;

The data visualization and analysis refers to graphically displaying the data after the dynamic feedback results are transmitted to form a dynamic feedback value chain flow diagram. Afterwards, the data in the data space will be rationally selected to form a historical data warehouse in the supply chain process. Based on the direct data obtained in each link system and the historical data stored in the data, it can be determined through process indicators and data docking and functional association. Data association is displayed and then used for calculation of historical data to analyze the entire value chain process.

Specifically, the data acquisition is an interface that collects data from outside the system and inputs it into the system. The data acquisition system integrates signals, sensors, actuators, data acquisition equipment and application software. In the network collaborative manufacturing platform, direct data is obtained from various data systems and integrated into value-added data source storage media that uses numerous information. Data sources are initially obtained from various systems and modules in network operation front-end pages, production plants, financial reporting, and supply recovery. For example: customer management system, marketing service management system, enterprise resource management system, warehousing logistics management system, enterprise financial management system, production process control system, product recycling management system, product after-sales service system, product testing management system. The data obtained from these systems are uniformly received and integrated into the data source of the system of the present invention, and the data source is used to drive the visual display of information data. Standardized systems will provide built-in value-added data sources and event target sources, which can be directly adapted to obtain data within a set time period. For non-standardized systems, support expansion needs to be carried out in the form of specifications. The storage of data sources requires clear identity settings. The identification includes supply capacity, sales attributes, material production cycle, equipment utilization, standard production cycle, equipment change time, personnel, shift working time, equipment name and number, basic customer information, distribution status, product market sales quantity, market Sales fluctuation, customer purchasing power, product discount sensitivity, and delivery time tolerance; supply capacity represents the output efficiency of raw material production and processing of the supplier company; sales attributes refer to the options and dimensions of goods for consumers to choose, such as product color, Settings such as packages and sizes; material production cycle refers to the time consumed in producing each material; equipment utilization refers to the ratio of equipment usage time to theoretical production time; standard production cycle refers to the theoretical product processing and production rate; Equipment changeover time refers to the time it takes for the equipment to switch from producing one component to producing other components; personnel refers to the number of people who can actually participate in the work in a cycle; shift working time refers to the time the enterprise The specified production time of a factory cycle, that is, the interval between two adjacent shifts, is generally between 8 hours and 12 hours according to the common practice on the production line, and can be arranged according to the actual situation. The name and number of the equipment refer to the identifying information of the equipment; the basic customer information includes records of the customer’s address, name, age or gender; the delivery status refers to the delivery information such as the time and place of the product; the product market sales quantity refers to the product The total amount of online marketing and offline marketing; market sales fluctuation refers to the impact of natural factors or other objective factors on product sales within a period of time; customer purchasing power refers to the purchase price that customers can accept; product discount sensitivity refers to The extent to which a company's product discounts are attractive to customers; delivery time tolerance refers to the acceptance of the timeliness of goods being delivered to customers;

The data processing described includes defining the collected data after obtaining it from the data source, cleaning the data, and other operations to maintain the standardized format of the data. Finally, the data sources are abstracted. Initially, the basic data related to supply and production system are specified, the basic data related to production are from the production system, the basic data related to marketing services are from the marketing service system, and the basic data related to recycling services are from Basic data of recycling service system. By establishing data specifications and standards, supply indicators, production indicators, marketing service indicators, recycling service indicators, etc. are described, and the data sources are configured in the dynamic feedback flow diagram of the entire value chain of manufacturing enterprises. Utilize these data sources to implement data-driven real-time dynamic feedback flow graph analysis solutions. The data query and storage refers to storing the data in the data storage medium as historical data for the retention of the dynamic feedback process on the time scale; the data query refers to establishing an index table and storing the processed data in the data medium. middle. The data visualization and analysis refers to visualizing the initial data in the form of charts and other forms, and then graphically displaying the data after the dynamic feedback results are transmitted in the same window, and dynamically analyzing the changes in indicator parameters in each link in real time to form full value. Dynamic feedback flow diagram of chain links. The data in the data storage medium needs to be rationally selected to form a historical data warehouse in the entire value chain process. Based on the direct data obtained in each link system and the historical data stored in the data, process indicators can be connected to the data and functionally associated. Determine the data association and display it, which can then be used to calculate historical data and analyze the entire value chain process.

The present invention also provides a system for implementing the above analysis method. The system includes a data collection module, a data processing module, a data query and storage module, and a data visualization and analysis module; the data collection module is used to collect data from each system and Integrated into the data collection system, used to integrate and upload data from various collection systems; the data processing module is used to clean and standardize the collected data; the data query and storage module is used to process the processed data The data is stored in the database and an index table is established; the data visualization and analysis module is used to visualize the data of each link in the form of a dynamic feedback flow diagram of the entire value chain, and can use the data to analyze the process of the entire value chain. Calculation.

As shown in Figure 1, the supply chain enterprise business collaborative service platform in the present invention includes three parts, which adopt a data sharing integrated management method, namely the supply collaborative platform subsystem, the production collaborative platform subsystem, and the marketing service collaborative platform subsystem. Among them, the manufacturing/supply enterprise recycling group can access the inventory management, parts management, and distribution management in the supply collaboration platform subsystem and the production collaboration platform subsystem, and cooperate with the replenishment of remanufactured products/parts by manufacturing/supply enterprises; the above analysis The system modules of the method are embedded in it, making it easier to call data on the platform.

The invention also provides a manufacturing enterprise full value chain dynamic feedback flow chart data sharing integrated management platform, including a supply chain enterprise business collaboration service platform, a supply chain enterprise business collaboration platform server, a manufacturing enterprise A supplier group, and a supplier enterprise. Internal server, manufacturing enterprise group, manufacturing enterprise internal server, manufacturing enterprise B dealer group and recycler group;

The supply chain enterprise business collaboration service platform is communicatively connected to the supply chain enterprise business collaboration platform server;

The internal server of the supplier enterprise is located within the supplier group of manufacturing enterprise A;

The internal server of the manufacturing enterprise is located within the manufacturing enterprise group;

The manufacturing enterprise A supplier group can access the recycler group, the manufacturing enterprise group can access the manufacturer enterprise B dealer group and the recycler group, the recycler group and the manufacturing enterprise B dealer group can only access the supply chain enterprise business collaboration service The platform provides access to business data, and they can only have specific access rights. The supply chain enterprise business collaboration service platform includes three parts, namely the supply collaboration platform subsystem, the production collaboration platform subsystem, and the marketing service collaboration platform subsystem. Among them, the manufacturing/supply enterprise recycling group can access the inventory management, parts management, and distribution management in the supply collaboration platform subsystem and the production collaboration platform subsystem, and cooperate with the replenishment of remanufactured products/parts by manufacturing/supply enterprises. By embedding the system modules of the above analysis methods into it, data can be called more conveniently on the platform.

Data analysis in the dynamic feedback flow graph analysis method of the entire value chain of manufacturing enterprises includes analyzing and learning the associations in data extraction information, using data mining algorithms to determine frequent item sets, and rationally selecting and deleting historical data to form a system that can reflect supply Data warehouse of chain state history for data modeling. When it is necessary to analyze the impact of changes in market customer demand on time factors during the production process, the customer-related feature data set is called from the data warehouse. Mathematical models that extract the most relevant data for supply, production, sales, product recycling and other indicators are used to analyze and predict inventory levels in each link in the value chain, thereby guiding the smooth operation of the supply chain links.

The present invention also provides a display information editing unit, which is used to generate information display and edit the process through its interactive interface. The release of display information displays the dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise.

The dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise of the present invention takes into account upstream suppliers, midstream manufacturers, downstream dealers and recycling service providers, and can flexibly monitor the business data of each link in the entire value chain. , and displays the dynamic feedback flow diagram of the entire value chain of manufacturing enterprises through the visual terminal. It can be used to determine the dealer's order rate in the next cycle, the manufacturer's expected productivity, and the supplier's order booking rate, reducing planning uncertainty among enterprises by adjusting part productivity, parts delivery rate, product productivity, product delivery rate, product sales rate, inventory adjustment rate, etc. to stabilize the inventory levels of suppliers, manufacturers, and distributors. By analyzing customers' operating behaviors in advance, we can discover customer groups with potential spending power in the market and provide decision-making guarantee for planning uncertainties among enterprises. Determine the processes that do not generate value-added in the product marketing process and provide managers with maximum marketing value-added services.

Example

To realize dynamic feedback flow diagram analysis of the entire value chain of manufacturing enterprises, the specific steps include the following:

Step 1. Collect information about supply, production, marketing services, and recycling services, and establish a sound data source storage. Data from the supply system related to the supply chain include the current inventory of raw material procurement, main processing equipment and labels and quantities, product production cycles, daily production capacity, types of main supplied products, corporate delivery forms, purchase discounts, supply Commercial transportation time and distance, re-purchased material specifications, common specification information include weight, material, size, etc., determine supplier number, purchase unit price, etc.; data from the production system related to the manufacturing process includes the work of collecting production equipment Time parameters, production process conditions and quality data, direct operating costs and indirect production costs, collect production data closely related to time parameters, including the pre-demand and follow-up relationship of production equipment, the number of effectively working equipment and the reliability of the workbench parameters, the proportion of production output, the proportion of qualified products, etc.; the data from the marketing system related to the marketing service link are closely related to the connection between the marketing effect and the selected marketing method. Including product price prediction, product performance promotion, delivery effectiveness, sales volume, competitive products, basic customer information, customer relationship network, customer purchasing power, promotion sensitivity, delivery time tolerance; data from the recycler system related to the recycling process Including enterprise recycling capabilities, recycling channels, remanufacturing rate of recycled products, recycling inventory quantity, repurchase prices of manufacturers and suppliers, market recyclability fluctuations, etc.;

Step 2. For the data collection and transfer, the common ones include hardware collection input and interface connection between software data systems;

Step 3. Combine the customer management system, marketing service system, enterprise resource management system, warehousing and logistics management system, enterprise financial management system, production process control system, production process execution management system, product recycling management system, product after-sales service system, and product testing Dynamic data of different consecutive periods are obtained in the management system and matched with each other in actual links as the basic data in the data source;

Step 4. Select the customer's operational behavior data, clean and standardize the data, associate features and select the optimal feature subset through the correlation coefficient method, reduce the complexity of the data, and use the data as training for the LSTM neural network set and test set. The optimal parameters are obtained through training to predict customer purchasing behavior.

Step 5: Establish a causal relationship diagram of dynamic feedback based on the key links of dynamic feedback of the entire value chain of the manufacturing enterprise that you want to focus on, and establish a dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise based on the mapping relationship, and combine the aforementioned steps 1 to The regular data in 4 are loaded into the flow chart of the dynamic feedback of the entire value chain of the manufacturing enterprise, and a data management system is established. As the business process progresses, the data obtained from each system will be processed and connected by selecting key indicators. Real-time display in the form of dynamic feedback flow graph.

The design through this method is a displayable method after data collection. Behind its internal logic and implementation method is the integration and interoperability of advanced modern Internet technology. On the basis of having a lot of basic data, the data is Through selection, screening, standardization and other processes, relevant indicators are selected to establish correlations with each other to form an overall closed-loop dynamic feedback logic flow diagram, and visual analysis is performed on the value links.

For example, regarding the recorded number of customer behavior operations: the number of customer behavior operations is affected by the quality of the product, the appearance of the product, the function of the product, etc. These effects are also affected by the enthusiasm of the company's employees, and the company's salary setting motivates employees. , the proficiency of employee skills, the importance employees attach to product quality, and a series of factors; in this way, a series of behind-the-scenes elements can be concentrated into one indicator, and the impact on other important indicators when this indicator changes can be deeply explored; At the same time, a causal relationship diagram is established to clarify the relationship between important indicators, form a dynamic feedback flow diagram, and realize the process analysis of dynamic feedback of the entire value chain.

In summary, the present invention discloses a method for generating dynamic feedback flow diagram analysis of the entire value chain of a manufacturing enterprise. It includes collecting value-added data from enterprise intelligent systems; processing the obtained value-added data sources, storing the processed value-added data sources and associating them with the corresponding business links; selecting appropriate indicators to perform mathematical calculations on the data sources Modeling operations; perform process mapping on value-added data sources to generate dynamic feedback value chain flow diagrams; perform output analysis on the generated dynamic feedback value chain flow diagrams of manufacturing enterprises, and store the results in a historical database. This invention is mainly aimed at a series of activities generated by enterprise supply/production/marketing/product recycling/service links, etc., to reduce the impact of the bullwhip effect on the fluctuation of enterprise inventory levels in each link through neural network prediction and inventory level adjustment. It can not only reasonably plan the supply/production time and cycle, improve the efficiency of the supply chain level, reduce the redundancy of raw materials, but also achieve inventory optimization and reduce production costs, and also provide new solutions for the optimization of enterprise business process links.

The method described in the present invention is designed in a displayable manner after data collection. Behind its internal logic and implementation method is the integration and interoperability of advanced modern Internet technology. On the basis of having a lot of basic data, the data Carry out selection, screening, standardization and other processes, select related indicators to establish correlations with each other, form an overall closed-loop dynamic feedback logic flow diagram, and perform visual analysis on the value links.

For example, regarding the recorded number of customer behavior operations: the number of customer behavior operations is affected by the quality of the product, the appearance of the product, the function of the product, etc. These effects are also affected by the enthusiasm of the company's employees, and the company's salary setting motivates employees. , the proficiency of employee skills, the importance employees attach to product quality, and a series of factors. In this way, a series of underlying elements can be concentrated on one indicator, and the impact on other important indicators when this indicator changes can be deeply explored. At the same time, a causal relationship diagram is established to clarify the relationship between important indicators, form a dynamic feedback flow diagram, and realize the process analysis of dynamic feedback of the entire value chain.

The above description of the disclosed embodiments enables those skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be practiced in other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A method for generating a dynamic feedback flow diagram for the entire value chain of a manufacturing enterprise, which is characterized by including the following steps: S1: Collect value-added data from the database of the enterprise's intelligent system, form a value-added data source, and store the data; S2: Extract data information from value-added data sources and perform data preprocessing on the extracted data information; S3: Define the data information after data preprocessing and obtain an abstract description of the data information; S4: Configure the abstractly described data information in the dynamic feedback link of the manufacturing enterprise's full value chain, match it with the corresponding indicators in the dynamic feedback link of the manufacturing enterprise's full value chain, perform data association, and make predictions through neural networks; S5: Perform value chain mapping on value-added data sources to generate a dynamic feedback flow diagram of the entire value chain of the manufacturing enterprise, and store it in the database. 2. The method for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise according to claim 1, characterized in that in step S1, the enterprise intelligent system includes: a customer management system, a marketing service management system, and an enterprise resource management system. System, warehousing logistics management system, corporate financial management system, production process control system, production process execution management system, product recycling management system, product after-sales service system, product testing management system; In step S1, collecting value-added data includes: collecting various text data information, customer behavior information, email information, customer web page data information, communication data information, and production process from various management systems of value-added data through data collection equipment. Parameter information and product testing information. 3. The manufacturing enterprise full value chain dynamic feedback flow diagram generation method according to claim 1, characterized in that the data preprocessing includes: dividing the data in the value-added data source into data types, and dividing the data into Compare data in each type with data in value-added data sources to find missing values and delete abnormal values. 4. The manufacturing enterprise full value chain dynamic feedback flow diagram generation method according to claim 1, characterized in that, in the step S3, the data information after data preprocessing is defined to obtain an abstract description of the data information, include: The data information after data preprocessing is defined as: basic data related to the supply production system, basic data related to production from the production system, basic data related to marketing from the marketing service system, and recycling services related Basic data from the recycling service system; Establish a unified and identifiable standardized model for data information in multiple systems, save supply indicators, production indicators, marketing indicators and recycling service indicators in the form of key values, use "keys" to distinguish data dimensions, and "values" to save them Get the value to realize the abstract description of the data information. 5. The method for generating a dynamic feedback flow diagram of the full value chain of a manufacturing enterprise according to claim 1, characterized in that in step S4, the corresponding indicators in the dynamic feedback link of the full value chain of the manufacturing enterprise include: and supply. Relevant supply indicators, production-related production indicators, marketing-related marketing indicators and recycling-related recycling service indicators; In step S4, the abstractly described data information is configured in the dynamic feedback link of the full value chain of the manufacturing enterprise. The method of matching the corresponding indicators in the dynamic feedback link of the full value chain of the manufacturing enterprise includes: Establish a data management system in the database, index and store abstractly described data information in the database through corresponding indicators in the dynamic feedback link of the manufacturing enterprise's full value chain, and complete the acquisition of dynamic feedback data of the manufacturing enterprise's full value chain; In the step S4, the data association includes: performing correlation analysis on the customer's basic operating behavior and user needs, and then correlating the customer behavior information in the value-added data source to build a correlation feature data set; In the step S4, prediction is made through the neural network, including: predicting the purchasing behavior of potential customers through the neural network based on the associated feature data set. In the supply chain stage, based on the retailer's historical sales data and the manufacturer's shipment history data With the manufacturer's order demand historical data and the factors that affect sales, shipments or order demand data in the supply chain, we comprehensively consider the reverse forecast of order production and sales, and combine the inventory adjustment rate of the supply chain enterprise as a selection balancing variable. Determine the expected supply order volume in the supply link, the expected production volume in the production link, and the expected order quantity in the sales link. 6. The method for generating a dynamic feedback flow diagram of the entire value chain of a manufacturing enterprise according to claim 5, characterized in that predicting the purchasing behavior of potential customers through a neural network includes the following steps: (1) Customer operation data preprocessing: After performing data domain processing on the value-added data source, and then performing feature selection, the customer's basic operation process and different operation behaviors in the customer behavior information are associated with the customer needs in the customer behavior information. Analysis, then construct a feature data set, and use the correlation coefficient method to select the optimal feature subset; (2) Define fitness: Use the mean square error of the predicted value of the LSTM neural network as the fitness value fit of the particle, where: ; Among them, y is the real value, is the expected output value; (3) Use the position information of particles as parameters of the LSTM neural network and construct an LSTM neural network; (4) Train the LSTM neural network: update the individual extreme value and group extreme value of the particle according to the fitness value of each particle; (5) Use nonlinear inertial weights to iteratively update the particle’s speed and position information based on the particle’s individual extreme value and group extreme value; (6) When the speed and position information of the particles meet the conditions or reach the maximum number of iterations, enter the next step to obtain the optimized parameters, otherwise return to step (3); (7) After obtaining the optimized parameters, increase the number of iterations, retrain the LSTM neural network, and obtain the trained LSTM neural network; (8) Based on the optimal feature subset, predict customer purchasing behavior through the trained LSTM neural network. 7. The manufacturing enterprise full value chain dynamic feedback flow diagram generation method according to claim 6, characterized in that the selection method of the optimal feature subset is: First, the Pearson coefficient is used as the selection index of the feature subset, and the parameter a is selected as the threshold in the feature subset selection process. Based on the value-added data, the feature pairs whose absolute value of correlation is less than the parameter a are found. Based on the selected feature pairs Compare the correlation of these features to the feature data set respectively, remove feature pairs with small correlations, and obtain a new feature subset; The evaluation function is used to evaluate the generated feature subset to obtain the evaluation result. The index evaluated by the feature subset is compared with the evaluation result. If the index evaluated by the feature subset is greater than the evaluation result, the evaluation result and the optimal feature subset are updated. , otherwise, stop the algorithm and output the current optimal feature subset; Optimize the LSTM neural network model parameters based on the optimal feature subset; Predict customer demand for different products at the same time interval based on corresponding indicators in the dynamic feedback link of the entire value chain of manufacturing enterprises. 8. A manufacturing enterprise full value chain dynamic feedback flow diagram generated by the method according to any one of claims 1 to 7, characterized in that the manufacturing enterprise full value chain dynamic feedback flow diagram includes: Dynamic feedback flow graph visual editor and configuration page, dynamic feedback flow graph variable types and connection models, database system and real-time interface. 9. A data sharing integrated management platform for dynamic feedback flow diagrams of the entire value chain of manufacturing enterprises, which is characterized by: Supply chain enterprise business collaboration service platform, supply chain enterprise business collaboration platform server, manufacturing enterprise A supplier group, supplier enterprise internal server, manufacturing enterprise group, manufacturing enterprise internal server, manufacturing enterprise B dealer group and recycler group; The supply chain enterprise business collaboration service platform is communicatively connected to the supply chain enterprise business collaboration platform server; The internal server of the supplier enterprise is located within the supplier group of manufacturing enterprise A; The internal server of the manufacturing enterprise is located within the manufacturing enterprise group; The manufacturing company A supplier group can access the recycler group, the manufacturing company group can access the manufacturer company B dealer group and the recycler group, the recycler group and the manufacturing company B dealer group can only access the supply chain The enterprise business collaboration service platform accesses business data, and all access requests have specific access rights. 10. An analysis method for the dynamic feedback flow diagram of the full value chain of a manufacturing enterprise, characterized in that the analysis method is based on the dynamic feedback flow diagram of the full value chain of a manufacturing enterprise and includes the following steps: data collection and classification, data processing, Data query and storage, data visualization and analysis; The data collection refers to obtaining direct data from various data systems, uniformly receiving and integrating it into value-added data sources, and inputting different types of data information into a unified data storage space; The data processing refers to defining and describing the data obtained during the collection process; The data query and storage refers to storing the processed data in the data space and establishing an index table structure; The data visualization and analysis refers to graphicalizing and displaying the data after the dynamic feedback results are transmitted to form a dynamic feedback value chain flow diagram, and then rationally selecting the data in the data space to form historical data in the supply chain process. Warehouse, and based on the direct data obtained in each link system and the historical data stored in the data, the data association is determined and displayed through process indicators, data docking and functional association, and then used for calculation of historical data to analyze the entire value chain process.