Cold chain supply chain meat inventory optimization regulation and control method and system

Abstract

The invention provides a cold chain supply chain meat inventory optimization regulation and control method and system, and the method comprises the steps: collecting multi-dimensional data to construct an impact factor data set, fusing a long and short-term memory network model, a gradient boosting tree algorithm model and an attention mechanism, and precisely predicting meat demands. Comparing the inventory, the threshold value and the predicted value, and generating an instruction according to a preset rule; and continuously optimizing the model through online learning, error evaluation and rolling prediction. According to the invention, the problem of overstock or stockout caused by inaccurate demand prediction and lack of a scientific regulation and control mechanism in the existing cold chain supply chain meat inventory management is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of cold chain logistics and supply chain management, and in particular to a method and system for optimizing and controlling meat inventory in a cold chain supply chain. Background Art

[0002] As people's living standards improve, demand for meat products continues to grow. The cold chain plays a key role in the entire process of meat products, from production to end-user sales. However, meat inventory management in the current cold chain faces many challenges.

[0003] Under traditional management models, inventory forecasting often relies on simple historical data statistical methods, such as moving averages and exponential smoothing, or on managers' empirical judgment. These methods have significant flaws. First, meat product demand is influenced by a complex combination of factors, such as holiday consumption peaks, market price fluctuations, shifting consumer preferences, and logistics disruptions caused by the epidemic. Simple statistical methods fail to comprehensively account for these multi-dimensional dynamic factors, resulting in significant deviations between forecast results and actual demand. Second, manual experience lacks scientific quantitative analysis and systematicity, making it impossible to make timely and accurate inventory decisions in a rapidly changing market environment.

[0004] Improper inventory management can lead to a range of serious consequences. Overstocking can significantly increase cold chain storage costs, leading to product losses such as spoiled or expired meat, resulting in wasted resources. Inventory shortages can lead to missed sales opportunities and impact company revenue. Therefore, accurately forecasting and optimizing meat inventory within the cold chain supply chain has become a pressing issue for the industry. Summary of the Invention

[0005] The present invention provides a method and system for optimizing and controlling meat inventory in a cold chain supply chain, which solves the problem of inventory backlog or shortage caused by inaccurate demand forecasting and lack of scientific control mechanisms in the existing meat inventory management of the cold chain supply chain.

[0006] According to one aspect of the present invention, a method for optimizing and controlling meat inventory in a cold chain supply chain is provided, comprising the following steps: Collect sales terminal data, meteorological data, holiday and promotion data, and supply chain collaboration data to build a multi-dimensional influencing factor dataset that includes time series characteristics, environmental variables, event identifiers, and supply status; The long-short-term memory network model is used to extract the time series characteristics of historical sales data, capture periodic patterns and long-term trends, and generate a static basic meat demand forecast value. A gradient boosting tree algorithm model is used to establish a nonlinear mapping relationship between the demand influencing factors and meat demand fluctuations to predict the meat demand fluctuations. The attention mechanism is used to calculate the importance of different influencing factors to meat demand in different time periods, identify the demand influencing factors that play a key role in demand changes, and dynamically assign weights to the demand influencing factors based on their importance to obtain a dynamically revised meat demand forecast value. Comparing the current inventory quantity and the safety stock threshold with the revised meat demand forecast value, and generating replenishment, transfer and / or promotion instructions based on preset decision rules; Use the multi-dimensional influencing factor data collected daily for online learning, and update the prediction results of the long short-term memory network model and the gradient boosting tree algorithm model after online learning; compare the updated prediction results with the actual sales data, and calculate the prediction error rate; if the prediction error rate exceeds the set threshold, start the rolling prediction mechanism, expand the time window of the training data, retrain the model, and adjust the distribution of the demand influencing factor weights, and save the updated model for continuous optimization.

[0007] Furthermore, the safety stock threshold calculation formula is: safety stock = (estimated maximum consumption - average consumption) × procurement lead time.

[0008] Furthermore, the preset decision rules include: If the current inventory quantity is less than the revised meat demand forecast value and less than the safety stock threshold, a replenishment instruction is generated; If the revised meat demand forecast value is less than the safety stock threshold value and less than the current stock quantity, a transfer instruction is generated; If the safety stock threshold is less than the revised meat demand forecast value and less than the current inventory quantity, a promotion instruction is generated and the promotion plan is pushed to the marketing terminal; wherein, the promotion plan is pushed to the marketing terminal through a built-in promotion plan template library and a dynamic matching plan according to the promotion instruction.

[0009] According to another aspect of the present invention, a cold chain meat inventory optimization and control system is provided, comprising: The first module collects sales terminal data, meteorological data, holiday and promotion data, and supply chain collaboration data to construct a multi-dimensional influencing factor dataset that includes time series characteristics, environmental variables, event identifiers, and supply status; a second module, the second module being communicatively connected to the first module and configured to transmit the multi-dimensional influencing factor dataset in the first module to the second module; generating a basic meat demand forecast based on historical sales data, establishing a nonlinear mapping relationship between the multi-dimensional influencing factors and demand fluctuations, and generating a revised meat demand forecast through dynamic weight allocation; a third module, the third module being communicatively connected to the second module and configured to transmit the revised meat demand forecast value in the second module to the third module; compare the current inventory quantity and the safety stock threshold with the revised meat demand forecast value, and generate replenishment, transfer, and / or promotion instructions based on preset decision rules; The fourth module is communicated with the first module and the second module respectively; online learning is performed based on the multi-dimensional influencing factor data in the first module, and the prediction results of the long short-term memory network model and the gradient boosting tree algorithm model after online learning are updated; the prediction error rate is evaluated based on actual sales data, and when the error rate exceeds a preset threshold, model retraining is triggered, and the influencing factor weight distribution is adjusted, and the saved and updated model retraining results are transmitted to the second module for continuous optimization.

[0010] Furthermore, the first module includes: Data interface unit, used to access sales terminal systems, meteorological APIs, and supply chain management systems to obtain real-time data; Data preprocessing unit, used to clean and extract features from the collected data; A data transmission unit is used to transmit the processed multi-dimensional impact factor data set to the second module and the fourth module respectively.

[0011] Furthermore, the second module includes: Time series forecasting unit, used to extract cyclical patterns and long-term trends from historical sales data; A fluctuation prediction unit, configured to establish a nonlinear mapping relationship between the multi-dimensional influencing factors and demand fluctuations; Attention weight unit, used to dynamically calculate the importance weight of each influencing factor; A data interaction unit is used to receive the multi-dimensional influencing factor data set of the first module through a communication connection with the first module; and is used to transmit the revised meat demand forecast value to the third module through a communication connection with the third module.

[0012] Furthermore, the third module includes: a threshold comparison unit, configured to receive the revised meat demand forecast value from the second module and compare it with the current inventory quantity and the safety stock threshold; An instruction generation unit, configured to generate replenishment, transfer and / or promotion instructions based on preset decision rules; The instruction transmission unit is used to transmit the instruction to the corresponding execution component.

[0013] Furthermore, the fourth module includes: An online learning unit, connected to the first module for acquiring real-time data; an error evaluation unit, configured to compare the revised meat demand forecast value with actual sales data and calculate a forecast error rate; a retraining triggering unit, configured to request the first module to expand the training data window via a communication connection with the first module when the prediction error rate exceeds a preset threshold; A weight updating unit is used to adjust the weight distribution of the influencing factors and transmit the optimized model retraining results to the second module through a communication connection with the second module.

[0014] Furthermore, it also includes: A fifth module is communicatively connected to the first, second, and third modules, respectively, to obtain data from the first, second, and third modules via the communication connections, and to present demand forecast trends, inventory status changes, and decision-making instruction execution results via a terminal.

[0015] The present invention provides a method and system for optimizing and controlling meat inventory in a cold chain supply chain, which solves the problem of inventory backlog or stockout caused by inaccurate demand forecasts and the lack of a scientific control mechanism in the existing cold chain supply chain meat inventory management. By collecting multi-dimensional influencing factor data to construct a data set, the long-short-term memory network model, the gradient boosting tree algorithm model and the attention mechanism are used to accurately predict meat demand; the predicted value is compared with the inventory quantity and the safety stock threshold, and replenishment, transfer and promotion instructions are generated based on preset decision rules to achieve dynamic control of inventory; the model is continuously optimized using online learning, error evaluation and rolling forecast mechanisms, thereby improving the accuracy of meat demand forecasts, the scientific nature of inventory management and the overall operational efficiency of the supply chain, effectively reducing the risk of inventory backlog or stockouts, and improving the intelligence level of meat inventory management in the cold chain supply chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings: FIG1 is a flow chart of method steps according to an embodiment of the present invention; FIG2 is a schematic diagram of system modules according to an embodiment of the present invention. DETAILED DESCRIPTION

[0017] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of this application can be combined with each other. In order to enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," and the like in the specification and claims of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate to facilitate the embodiments of the present invention described herein. In addition, the terms "including," "comprising," and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, product, or apparatus comprising a series of steps or elements is not necessarily limited to the steps or elements explicitly listed, but may include other steps or elements that are not explicitly listed or that are inherent to the process, method, product, or apparatus.

[0019] The Long Short-Term Memory Network (LSTM) model is a special recurrent neural network (RNN) designed specifically to address the vanishing and exploding gradient problems of traditional RNNs when processing long sequences. It is good at capturing long-term dependencies in time series.

[0020] The Gradient Boosting Decision Tree (GBDT) algorithm is an iterative algorithm based on ensemble learning. It builds a strong learner by combining multiple weak learners (decision trees). It uses the principle of gradient descent to gradually optimize the model's ability to fit complex nonlinear relationships, thereby improving the model's predictive accuracy.

[0021] The Attention Mechanism is a machine learning technique that simulates human attention. It calculates the importance weights of different input parts, enabling the model to selectively focus on key information.

[0022] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. Example 1

[0023] 1 , the cold chain supply chain meat inventory optimization and control method according to an embodiment of the present invention includes: S102, accessing the sales terminal system through the data interface unit to obtain the meat sales records of each terminal in the past 12 months, including sales time, category, quantity, price, etc.

[0024] Connect to the weather API to obtain real-time weather data such as temperature, humidity, and precipitation probability for the city where each terminal is located.

[0025] Extract holiday calendars and promotional activity plans, such as discount levels and event times, from the WMS (Warehouse Management System), as well as supply chain collaboration data, such as supplier inventory and logistics delivery timeliness.

[0026] The collected data is cleaned to remove outliers, such as records with zero meat sales, and time series features are extracted, such as daily sales, week-on-week growth rate, environmental variables, event identifiers, and inventory levels.

[0027] S104: Input historical sales data into the long short-term memory network model to extract sales cycle patterns and seasonal trends within a fixed period, such as a 30% increase in weekend sales and a 20% decrease in summer beef sales, to generate a basic meat demand forecast.

[0028] The gradient boosting tree algorithm model was used to analyze the nonlinear mapping relationship between influencing factors such as meteorological data and promotional activities and fluctuations in meat demand. For example, for every 5°C increase in temperature, pork sales decreased by 5%; promotional activities could increase sales by 25%.

[0029] Calculate the importance weight of each influencing factor. For example, in summer, the weight of "temperature" is 0.6, and the weight of promotional activities is 0.3; in winter, the weight of "holidays" is increased to 0.7. The basic forecast value and the demand fluctuation forecast value are combined according to the weights to obtain the revised meat demand forecast value.

[0030] S106: Compare the revised meat demand forecast with the current inventory quantity and the safety stock threshold (calculated as follows: safety stock = (estimated maximum consumption - average consumption) × procurement lead time). If the following conditions are met, a corresponding instruction is generated: If the current inventory quantity is less than the revised meat demand forecast value and less than the safety stock threshold, a replenishment instruction is generated; the replenishment instruction is pushed to the purchasing department or supplier system, the purchasing department confirms the order and executes the purchasing process, and the inventory data is updated through the first module after the goods arrive.

[0031] If the revised meat demand forecast value is less than the safety stock threshold and less than the current inventory quantity, an allocation instruction is generated; the allocation instruction is sent to the warehousing and logistics department or the transportation system to trigger the inventory transfer operation between warehouses, and the excess inventory is transferred to warehouses or sales terminals with higher demand, so as to balance the inventory of each node and reduce the backlog cost; the warehouse is transferred out to perform picking, packing and outbound operations, and the inventory data is updated; after the transfer-in party completes the acceptance, the inventory changes are synchronized through the first module.

[0032] If the safety stock threshold is less than the revised meat demand forecast value and less than the current inventory quantity, a promotion instruction is generated and the promotion plan is pushed to the marketing terminal; wherein, the promotion plan is pushed to the marketing terminal through the built-in promotion plan template library and the promotion plan is dynamically matched according to the promotion instruction; promotion activities are triggered to accelerate inventory turnover and avoid deterioration and loss or capital occupation due to inventory backlog; the marketing department executes the promotion activities, tracks sales data in real time and feeds back to the system through the first module to update the inventory and demand forecasting model.

[0033] S108: After the end of each day's business, the newly collected data is input into the model to update the parameter / influence factor data; the predicted value is compared with the actual sales value, and the prediction error rate is calculated, for example, the error rate this week is 8%; if the error rate exceeds the threshold, for example, it is set to 5%, the rolling prediction mechanism is activated, the training data window is extended from 12 months to 18 months, the model is retrained and the influence factor weights are adjusted. Example 2

[0034] The cold chain supply chain meat inventory optimization and control system of the embodiment of the present invention is deployed in a cloud server cluster, adopts a microservice architecture, and each module communicates through a message queue, supporting horizontal expansion to cope with high concurrency scenarios.

[0035] refer to Figure 2 The cold chain supply chain meat inventory optimization and control system of the embodiment of the present invention includes: In the first module, the data interface unit deploys the API gateway and connects to the sales terminal system, meteorological service provider, WMS, and supply chain management platform through the interface; the data preprocessing unit performs data cleaning and uses the feature engineering toolkit to extract time series features; the data transmission unit stores the processed data in the distributed file system and pushes it to the downstream module through the message queue.

[0036] In the second module, the time series prediction unit implements the LSTM model and is deployed on a GPU server to accelerate computing; the fluctuation prediction unit constructs GBDT to support batch training and online prediction; the attention weight unit implements the attention mechanism to dynamically calculate the weights of each influencing factor; the data interaction unit provides a service interface to receive data from the first module and output the revised meat demand forecast value to the third module.

[0037] The third module includes a threshold comparison unit that obtains real-time inventory data from the cache server and compares it with the revised meat demand forecast value; an instruction generation unit that implements preset decision rules based on the rule engine and generates control instructions; and an instruction transmission unit that sends the control instructions to the warehouse management system (WMS) and marketing system through a message queue.

[0038] In the fourth module, the online learning unit performs incremental training on daily new data; the error evaluation unit regularly calculates the model prediction error rate and stores it in the time series database; the retraining trigger unit is used to call the automated training script to expand the data set and retrain the model when the error rate exceeds the set threshold; the weight update unit synchronizes the optimized model parameters / influence factor data to the prediction model of the second module.

[0039] The fifth module obtains data from each module through communication connections; uses drawing tools to draw demand forecast trend charts, inventory water level change curves, and instruction execution effect analysis tables; provides multi-dimensional data query functions, and supports filtering and display by conditions such as region, time, and category.

[0040] Compared to existing technologies, the solution provided by the above-mentioned embodiments of this application constructs a comprehensive dataset of influencing factors by collecting multi-dimensional data from sales terminals, weather, holiday promotions, and supply chain collaboration. It then integrates a long-short-term memory network model, a gradient boosting tree algorithm model, and an attention mechanism to establish a dynamic forecasting system. This solves the problem of traditional solutions relying on a single data set and insufficient forecasting accuracy, thereby improving the accuracy of meat demand forecasts. Furthermore, by establishing intelligent decision-making rules based on the comparison of inventory quantity, safety stock threshold, and demand forecast, combined with the function of automatically matching promotional plans, it solves the problem of rigid traditional inventory decision-making that easily leads to backlogs or stockouts, and improves the scientific nature of inventory management.

[0041] By leveraging online learning and dynamic optimization mechanisms, the system updates models and adjusts weights based on real-time data, achieving adaptive evolution. This avoids the drawbacks of traditional solutions that are unable to adapt to market dynamics, thereby improving the overall operational efficiency of the supply chain. The coordinated application of these technical means effectively reduces the risk of inventory backlogs or stockouts, and enhances the intelligent level of meat inventory management in the cold chain supply chain.

[0042] Compared with the prior art, the above-mentioned embodiment 1 of the present application collects multi-dimensional data such as sales terminals, weather, holiday promotions and supply chain collaboration. After cleaning and feature extraction, it uses the long short-term memory network model and the gradient boosting tree algorithm model, combined with dynamic weight distribution, to accurately capture the cyclical laws of meat demand, seasonal trends and the nonlinear effects of various influencing factors, solving the problem of traditional solutions relying on single data and low prediction accuracy, thereby improving the accuracy of meat demand forecasts. At the same time, based on the scientific safety stock threshold calculation formula and preset decision rules, the system automatically generates replenishment, allocation, and promotion instructions, and links the procurement, warehousing, and marketing departments to execute them, changing the rigidity of traditional inventory decisions that easily lead to backlogs or out-of-stock situations, improving the scientific nature of inventory management, and effectively reducing the risk of inventory backlogs or out-of-stocks. In addition, the daily data online learning and rolling forecasting mechanism enable the system to dynamically optimize the model according to actual sales conditions, thereby improving the overall operational efficiency of the supply chain.

[0043] In Example 2, a cold chain supply chain meat inventory optimization and control system with a microservice architecture deployed in the cloud has clear division of labor among modules and efficient communication through message queues, supporting horizontal expansion to cope with high-concurrency scenarios. The data interface unit enables multi-platform data access, while the data preprocessing unit and transmission unit ensure data quality and flow; the functional units of the second module accelerate model calculation and prediction; the third module enables real-time comparison of inventory data and accurate issuance of instructions; the fourth module ensures continuous optimization of the model; and the fifth module visualizes data and decision results. This modular and intelligent system architecture comprehensively improves the intelligence level of cold chain supply chain meat inventory management and further ensures the efficiency and accuracy of inventory control.

[0044] It should be noted that for the sake of simplicity, the aforementioned embodiments are described as a series of actions. However, those skilled in the art should be aware that this application is not limited to the order of the actions described, as certain steps may be performed in other orders or simultaneously according to this application.

[0045] Through the description of the above embodiments, those skilled in the art will clearly understand that the methods according to the above embodiments can be implemented using software plus the necessary general-purpose hardware platform. Of course, they can also be implemented using hardware devices, but in many cases the former is a more preferred implementation method. Based on this understanding, the technical solution of this application, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk) and includes a number of instructions for enabling a terminal device (which can be a computer) to execute the methods described in each embodiment of this application.

[0046] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for optimizing and controlling meat inventory in a cold chain supply chain, characterized in that: The following steps are involved: Collect sales terminal data, meteorological data, holiday and promotion data, and supply chain collaboration data to build a multi-dimensional influencing factor dataset that includes time series characteristics, environmental variables, event identifiers, and supply status; The long short-term memory network model is used to extract the time series characteristics of historical sales data, capture cyclical patterns and long-term trends, and generate static meat basic demand forecasts; A nonlinear mapping relationship between the demand influencing factors and meat demand fluctuations is established using a gradient boosting tree algorithm model to predict meat demand fluctuations. An attention mechanism is used to calculate the importance of different influencing factors to meat demand in different time periods, identify the demand influencing factors that play a key role in demand changes, and dynamically assign weights to the demand influencing factors based on their importance to obtain a dynamically revised meat demand forecast value. Comparing the current inventory quantity and the safety stock threshold with the revised meat demand forecast value, and generating replenishment, transfer and / or promotion instructions based on preset decision rules; Use the multi-dimensional influencing factor data collected daily for online learning, and update the prediction results of the long short-term memory network model and the gradient boosting tree algorithm model after online learning; compare the updated prediction results with the actual sales data, and calculate the prediction error rate; if the prediction error rate exceeds the set threshold, start the rolling prediction mechanism, expand the time window of the training data, retrain the model, and adjust the distribution of the demand influencing factor weights, and save the updated model for continuous optimization.

2. A method for optimizing and controlling meat inventory in a cold chain supply chain, characterized in that: The calculation formula for the safety stock threshold is: safety stock = (estimated maximum consumption - average consumption) × procurement lead time.

3. A method for optimizing and controlling meat inventory in a cold chain supply chain, characterized in that: The preset decision rules include: If the current inventory quantity is less than the revised meat demand forecast value and less than the safety stock threshold, a replenishment instruction is generated; If the revised meat demand forecast value is less than the safety stock threshold value and less than the current stock quantity, a transfer instruction is generated; If the safety stock threshold is less than the revised meat demand forecast value and less than the current inventory quantity, a promotion instruction is generated and the promotion plan is pushed to the marketing terminal; wherein, the promotion plan is pushed to the marketing terminal through a built-in promotion plan template library and a dynamic matching plan according to the promotion instruction.

4. A cold chain supply chain meat inventory optimization and control system, characterized in that: include: The first module collects sales terminal data, meteorological data, holiday and promotion data, and supply chain collaboration data to construct a multi-dimensional influencing factor dataset that includes time series characteristics, environmental variables, event identifiers, and supply status; A second module, the second module being communicatively connected to the first module and configured to transmit the multi-dimensional impact factor dataset in the first module to the second module; Generate a basic meat demand forecast based on historical sales data, establish a nonlinear mapping relationship between the multi-dimensional influencing factors and demand fluctuations, and generate a revised meat demand forecast through dynamic weight allocation; a third module, the third module being communicatively connected to the second module and configured to transmit the revised meat demand forecast value in the second module to the third module; compare the current inventory quantity and the safety stock threshold with the revised meat demand forecast value, and generate replenishment, transfer, and / or promotion instructions based on preset decision rules; A fourth module, wherein the fourth module is respectively in communication with the first module and the second module; performing online learning based on the multi-dimensional influencing factor data in the first module, and updating the prediction results of the long short-term memory network model and the gradient boosting tree algorithm model after online learning; The prediction error rate is evaluated based on actual sales data. When the error rate exceeds a preset threshold, model retraining is triggered, and the weight distribution of the influencing factors is adjusted. The saved and updated model retraining results are transmitted to the second module for continuous optimization.

5. The system according to claim 4, characterized in that The first module includes: Data interface unit, used to access sales terminal systems, meteorological APIs, and supply chain management systems to obtain real-time data; Data preprocessing unit, used to clean and extract features from the collected data; A data transmission unit is used to transmit the processed multi-dimensional impact factor data set to the second module and the fourth module respectively.

6. The system according to claim 4, characterized in that The second module includes: Time series forecasting unit, used to extract cyclical patterns and long-term trends from historical sales data; A fluctuation prediction unit, configured to establish a nonlinear mapping relationship between the multi-dimensional influencing factors and demand fluctuations; Attention weight unit, used to dynamically calculate the importance weight of each influencing factor; A data interaction unit is used to receive the multi-dimensional influencing factor data set of the first module through a communication connection with the first module; and is used to transmit the revised meat demand forecast value to the third module through a communication connection with the third module.

7. The system according to claim 4, wherein: The third module includes: a threshold comparison unit, configured to receive the revised meat demand forecast value from the second module and compare it with the current inventory quantity and the safety stock threshold; An instruction generation unit, configured to generate replenishment, transfer and / or promotion instructions based on preset decision rules; The instruction transmission unit is used to transmit the instruction to the corresponding execution component.

8. The system according to claim 4, wherein: The fourth module includes: An online learning unit, connected to the first module for acquiring real-time data; an error evaluation unit, configured to compare the revised meat demand forecast value with actual sales data and calculate a forecast error rate; a retraining triggering unit, configured to request the first module to expand the training data window via a communication connection with the first module when the prediction error rate exceeds a preset threshold; A weight updating unit is used to adjust the weight distribution of the influencing factors and transmit the optimized model retraining results to the second module through a communication connection with the second module.

9. The system according to claim 4, wherein: Also includes: The fifth module is communicated with the first module, the second module and the third module respectively, obtains the data of the first module, the second module and the third module through the communication connection, and presents the demand forecast trend, inventory status change and decision instruction execution results through the terminal.

Images