Intelligent supply chain system and server platform

Abstract

The application provides a smart supply chain system and a server platform. The smart supply chain system manages the supply chain of various commodities, and is characterized in that it comprises: a commodity classification device that classifies various commodities based on historical data; a sales prediction device that predicts sales of each commodity based on historical data and the classification results of various commodities by the commodity classification device; and an intelligent restocking device that generates a restocking decision by applying an automatic restocking model based on the prediction results of the sales prediction device. Thus, the three important blocks of commodity classification, sales prediction and intelligent restocking can be effectively integrated, the service becomes accurate, the supply chain becomes transparent, flexible and agile, and each function is more coordinated.

Description

Technical Field

[0001] The present invention relates to a smart supply chain system and a server platform, and in particular to a layout method of a smart supply chain. Background Art

[0002] In recent years, there has been an urgent need to effectively manage the supply chain between manufacturers, suppliers, and sales outlets. Specifically, in supply chain management, it is first necessary to scientifically classify and locate goods, and then adopt corresponding sales and inventory strategies for different categories of goods. Secondly, more accurate sales forecasts are needed as a driver of product quantity. Demand planning based on this can support planning in all aspects. Every 1% increase in forecast accuracy can lead to several times the reduction in operating costs. Finally, it is necessary to apply the automatic replenishment model to specific scenarios to optimize the inventory structure so that the inventory continues to maintain a healthy level. In the case of warehousing, all SKUs (stock keeping units) that meet the replenishment conditions are located, and the replenishment task model is run to derive replenishment quantity recommendations, and finally generate replenishment decisions for production.

[0003] However, the existing technology does not propose a supply chain management solution that meets the above needs, and there is an urgent need for a smart supply chain layout that can meet the above needs. Summary of the Invention

[0004] The present invention addresses the above-mentioned technical problems in the prior art, and its purpose is to provide an intelligent supply chain system and server platform that can effectively integrate the three important sections of product classification, sales forecasting and intelligent replenishment, so as to make services more precise, the supply chain more transparent, flexible and agile, and various functions more coordinated.

[0005] One embodiment of the present invention provides an intelligent supply chain system for managing the supply chains of multiple commodities, and is characterized by comprising: a commodity classification device for classifying multiple commodities based on historical data; a sales forecasting device for making sales forecasts for each commodity based on historical data and the classification results of the multiple commodities by the commodity classification device; and an intelligent replenishment device for applying an automatic replenishment model to generate replenishment decisions based on the forecast results of the sales forecasting device.

[0006] As a result, the three important areas of product classification, sales forecasting and intelligent replenishment can be effectively integrated, making services more precise, the supply chain more transparent, flexible and agile, and various functions more coordinated.

[0007] In the above-mentioned smart supply chain system, the commodity classification device may also classify a plurality of commodities based on the ABC classification results of the sales data as historical data and the most recent sales time interval as two dimensions to obtain preliminary classification results, and calculate the commodity characteristic analysis indicators of each commodity based on the historical data, and further use the calculated commodity characteristic analysis indicators to perform multi-dimensional segmentation on the preliminary classification results as the final classification results. The commodity characteristic analysis indicators include at least one of sales frequency, coefficient of variation, amplitude of variation, customer concentration, sales rhythm, and release interval.

[0008] Therefore, based on the segmentation results, it is possible to extract products for which the demand forecast results need to be confirmed by sales personnel, and by effectively combining the business experience of sales personnel, it is possible to improve the accuracy of demand forecasting.

[0009] In the above-mentioned smart supply chain system, the sales forecasting device may also use one of the forecasting methods of point forecasting, interval forecasting, and hybrid forecasting combining point forecasting and interval forecasting for each commodity that is subdivided by the commodity classification device into commodities that need to confirm the demand forecast results, and predict the sales volume corresponding to the at least one forecast time for each commodity.

[0010] Therefore, by combining factors such as the product classification results and the time to be predicted, and using the most suitable prediction method, a more accurate prediction result corresponding to the required prediction time can be obtained.

[0011] In the above-mentioned smart supply chain system, the intelligent replenishment device may also generate a replenishment plan based on the prediction results of the sales forecasting device and the product characteristic analysis indicators calculated by the product classification device, using a replenishment model based on point prediction, a hybrid replenishment model based on point prediction and interval prediction, and a replenishment model based on interval prediction.

[0012] Therefore, based on the characteristics of different products, different replenishment cycles and other factors, the most suitable replenishment model is used to generate a replenishment plan, which can provide replenishment decisions that better meet demand.

[0013] In the above-mentioned smart supply chain system, the sales forecasting device may also include: a point prediction unit, which performs point prediction based on a long short-term memory network; and an interval prediction unit, which performs probability density interval prediction; the smart supply chain system may also include: a model selection unit, which selects whether the sales forecast is performed by the point prediction unit, the interval prediction unit, or both the point prediction unit and the interval prediction unit in the sales forecasting device based on at least one of the application object, the prediction time, and the classification result of the commodity classification device.

[0014] Therefore, it is possible to select the most appropriate forecasting method for sales forecasting based on the application object (such as sales stores, suppliers, manufacturers, etc.), forecast time (such as 1 day, 1 week, 1 month, 1 year, etc.), classification results, etc.

[0015] In the above-mentioned smart supply chain system, the intelligent replenishment device may also include: a point prediction replenishment unit, which uses a replenishment model based on point prediction to generate a replenishment plan with regular and irregular quantities or irregular and irregular quantities; a mixed replenishment unit, which uses a mixed replenishment model based on point prediction and interval prediction to generate a replenishment plan with regular and irregular quantities; and an interval prediction replenishment unit, which uses a replenishment model based on interval prediction to generate a replenishment plan with irregular and irregular quantities, wherein the model selection unit selects which of the point prediction replenishment unit, the mixed replenishment unit, and the interval prediction replenishment unit in the intelligent replenishment device to generate the replenishment plan based on at least one of an application object, a prediction time, a classification result of the product classification device, and a prediction result of the sales forecasting device.

[0016] In this way, the most appropriate replenishment model can be selected to generate a replenishment plan based on the application object (such as sales stores, suppliers, manufacturers, etc.), forecast time (such as 1 day, 1 week, 1 month, 1 year, etc.), classification results, forecast results, etc.

[0017] In the above-mentioned smart supply chain system, when the application object is a sales store, the product classification device is subdivided into each product for which a demand forecast result needs to be confirmed. The model selection unit selects the point prediction unit in the sales forecast device to perform sales forecasting, builds a point prediction model of a long short-term memory network corresponding to at least one forecast time, and predicts the sales volume corresponding to the at least one forecast time for each product. Then, the model selection unit selects the point prediction replenishment unit in the intelligent replenishment device to generate a replenishment plan, and uses the replenishment model based on point prediction to generate a replenishment plan with regular or irregular quantity.

[0018] In this way, the smart supply chain system can be effectively applied to sales outlets.

[0019] In the above-mentioned smart supply chain system, when the application object is a supplier, the product classification device is subdivided into each product for which demand forecast results need to be confirmed. The model selection unit selects the point prediction unit and the interval prediction unit in the sales forecast device to perform sales forecasts, builds a point prediction model of a long short-term memory network corresponding to at least one prediction time among multiple prediction times, and a probability density interval prediction model corresponding to at least one other prediction time among the multiple prediction times, and predicts the sales volume corresponding to the multiple prediction times for each product. Then, the model selection unit selects the hybrid replenishment unit in the intelligent replenishment device to generate a replenishment plan, and uses the hybrid replenishment model based on point prediction and interval prediction to generate a regular and quantitative replenishment plan.

[0020] In this way, the smart supply chain system can be effectively applied to suppliers.

[0021] In the above-mentioned smart supply chain system, when the application object is a manufacturer, the commodity classification device is subdivided into each commodity for which a demand forecast result needs to be confirmed, and the model selection unit selects the interval prediction unit in the sales forecast device to perform sales forecasting, builds a probability density interval prediction model corresponding to at least one forecast time, and predicts the sales volume corresponding to the at least one forecast time for each commodity, and then the model selection unit selects the interval prediction replenishment unit in the intelligent replenishment device to generate a replenishment plan, and uses the replenishment model based on interval prediction to generate an irregular and indefinite replenishment plan.

[0022] In this way, the smart supply chain system can be effectively applied to manufacturers.

[0023] One embodiment of the present invention also provides a server platform for managing the supply chain of multiple commodities, comprising a processor, a memory, and an interface, and capable of communicating data with a client device via the interface. The server platform is characterized in that the processor executes the following processing by executing a program stored in the memory: commodity classification processing, classifying multiple commodities based on historical data; sales forecasting processing, making sales forecasts for each commodity based on historical data and the classification results of the commodity classification processing for multiple commodities; and intelligent replenishment processing, applying an automatic replenishment model based on the prediction results of the sales forecasting processing to generate replenishment decisions, and the replenishment decisions are sent to the client device via the interface.

[0024] The above-mentioned specific methods and effects of the smart supply chain system of the present invention can also be achieved through the above-mentioned server platform, the method executed by the above-mentioned smart supply chain system, the program that enables the computer to execute the above-mentioned method, or the recording medium that stores the above-mentioned program. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a conceptual diagram showing application objects to which the smart supply chain system according to the first embodiment of the present invention can be applied.

[0026] Figure 2 This is a functional block diagram showing the smart supply chain system according to the first embodiment of the present invention.

[0027] Figure 3 This is a block diagram showing an example of the hardware configuration of the smart supply chain system according to the first embodiment of the present invention.

[0028] Figure 4 This is a block diagram showing another example of the hardware configuration of the smart supply chain system according to the first embodiment of the present invention.

[0029] Figure 5 This is a flowchart showing a method executed by the smart supply chain system according to the first embodiment of the present invention.

[0030] Figure 6 This is a functional structure block diagram of the smart supply chain system according to the second embodiment of the present invention.

[0031] Figure 7 This is a table showing an example of a model selection reference table referred to by the model selection unit according to the second embodiment of the present invention.

[0032] Figure 8 This is a flowchart showing an application example (specific example 1) in which the smart supply chain system of the present invention is applied to a sales store.

[0033] Figure 9 This is a flowchart showing an application example (specific example 2) in which the smart supply chain system of the present invention is applied to a supplier.

[0034] Figure 10 This is a flowchart showing an application example (specific example 3) in which the smart supply chain system of the present invention is applied to a manufacturer.

[0035] Description of reference numerals:

[0036] 100, 100': intelligent supply chain system; 10: commodity classification device; 20: sales forecasting device; 21: point forecasting unit; 22: interval forecasting unit; 30: intelligent replenishment device; 31: point forecasting replenishment unit; 32: mixed replenishment unit; 33: interval forecasting replenishment unit; 40: model selection unit; 110, 110A, 110B: processor; 120, 120A, 120B: memory; 130, 130A, 130B: interface; 140: input device; 150: display unit; 160, 160A, 160B: bus; 100A: server platform; 100B: client device; 200: database. DETAILED DESCRIPTION

[0037] The present invention is described in more detail below with reference to the accompanying drawings, embodiments, and specific examples. The following descriptions are merely examples for facilitating understanding of the present invention and are not intended to limit the scope of the present invention. In specific embodiments, the components of the apparatus and system may be modified, deleted, or added based on actual circumstances, and the steps of the method may be modified, deleted, added, or reordered based on actual circumstances.

[0038] (First embodiment)

[0039] First, the application objects to which the smart supply chain system 100 of this embodiment can be applied are described. Figure 1 This is a conceptual diagram showing the application objects to which the smart supply chain system of the first embodiment of the present invention can be applied. The application objects of the smart supply chain system 100 include at least one of a sales store, a supplier, and a manufacturer, wherein the relationship between the sales store, supplier, and manufacturer is as follows: Figure 1 As shown, arrows indicate the direction of goods flow. Here, three sales outlets, Sales Outlet 1, Sales Outlet 2, and Sales Outlet 3, two suppliers, Supplier 1 and Supplier 2, and three manufacturers, Manufacturer 1, Manufacturer 2, and Manufacturer 3, are shown. However, the number and relationships of sales outlets, suppliers, and manufacturers are not limited to this. In practice, there may be a large number of sales outlets, suppliers, and manufacturers, as well as more complex relationships.

[0040] Sales outlets need to monitor the sales of each product to develop sales and inventory strategies. Suppliers source goods from various manufacturers and distribute them to multiple sales outlets, requiring them to place orders with manufacturers and arrange distribution plans for the next phase of sales outlets. Manufacturers produce goods and supply them to multiple suppliers, who then supply them to sales outlets. Each product has a different life cycle, necessitating forecasts of future demand to guide future production plans.

[0041] To address at least one of the above requirements, this embodiment provides a smart supply chain system 100 . Figure 2 1 is a functional block diagram showing the smart supply chain system 100 according to the first embodiment of the present invention. Figure 2 As shown, the smart supply chain system 100 includes a commodity classification device 10, a sales forecasting device 20 and an intelligent replenishment device 30. Figure 1 Each device can be implemented as a functional module by executing an application stored in a memory by a processor, or as a separate hardware structure. Two specific examples of specific implementations are described below.

[0042] like Figure 2 As shown, the product classification device 10 classifies multiple products based on historical data. For example, the product classification device 10 further considers the temporal characteristics of products in addition to the traditional ABC classification method. For example, the recent sales time interval is added as an analysis dimension, and the ABC classification results based on the historical sales data and the recent sales time interval are used as two dimensions to classify multiple products using the R-ABC classification method. This further refines the ABC classification, allowing more valuable products to be selected from the traditional ABC classification.

[0043] Here, the historical data includes the time of shipment, month / week, destination warehouse number, product number, shipment quantity, etc. For example, when the application object is a sales store, the historical data includes the sales time, month / week, product number, sales volume, etc. The historical data may also include other data as needed. In addition, since for suppliers / manufacturers, "shipping" to a certain customer is equivalent to "sales", the "shipping" of suppliers / manufacturers and "sales" of sales stores are uniformly expressed as "sales" in this manual. For example, "sales time" includes not only the sales time in the sales store, but also the shipment time in the supplier / manufacturer, and "sales volume" includes not only the sales volume in the sales store, but also the shipment volume in the supplier / manufacturer.

[0044] Furthermore, in order to further clarify the sales characteristics of the goods, the commodity classification device 10 can also calculate the commodity characteristic analysis indicators of each commodity based on historical data. The commodity characteristic analysis indicators include, for example, at least one of the following indicators: demand probability, sales frequency, coefficient of variation, amplitude of variation, customer concentration, sales rhythm, release interval, etc. Thus, based on the classification results (preliminary classification results) of the above-mentioned improved R-ABC classification method, the calculated commodity characteristic analysis indicators are used to perform multi-dimensional segmentation as the final classification results. Furthermore, based on the segmentation results, the commodities whose demand forecast results need to be confirmed by the sales staff are extracted. Thus, by effectively combining the business experience of the sales staff, the accuracy of the demand forecast can be improved.

[0045] The sales forecasting device 20 generates sales forecasts for each product based on historical data and the classification results for the plurality of products by the product classification device 10. For example, the sales forecasting device 20 predicts sales volume for a predetermined period of time (e.g., one day, one week, one month, one year, etc.) for each product, such as the products for which demand forecast results are required to be confirmed, extracted by the product classification device 10, based on the classification results by the product classification device 10.

[0046] Specifically, the sales forecasting device 20 can use, for example, point prediction based on a long short-term memory network to build a long short-term memory network point prediction model corresponding to the forecast time (e.g., 1 day, 1 week, or 1 year), perform data collection, data processing, model construction, model evaluation, and model deployment, and output a forecast result corresponding to the above-mentioned forecast time. Alternatively, the sales forecasting device 20 can also use probability density interval prediction based on deep learning (e.g., DeepAR method), build a probability density interval prediction model corresponding to the forecast time (e.g., 1 month), perform data collection, data processing, model construction, forecast result processing, and output a forecast result corresponding to the above-mentioned forecast time.

[0047] Furthermore, the sales forecasting device 20 can also utilize a hybrid forecasting method that combines the aforementioned point forecasting and interval forecasting to build multiple forecasting models corresponding to each of the multiple forecasting time periods, and output forecast results corresponding to each of the multiple forecasting time periods. For example, the sales forecasting device 20 can build a long short-term memory network point forecasting model corresponding to one day, a probability density interval forecasting model corresponding to one month, and a long short-term memory network point forecasting model corresponding to one year, and output forecast results corresponding to one day, one month, and one year, respectively. By combining factors such as the product classification results and the time to be forecasted, and utilizing the most appropriate forecasting method, a more accurate forecast result corresponding to the desired forecasting time can be obtained.

[0048] Based on the forecast results from the sales forecasting device 20 and the product characteristic evaluation indicators, the intelligent replenishment device 30 applies an automatic replenishment model to generate replenishment decisions. For example, the intelligent replenishment device 30 formulates different replenishment strategies based on the characteristics of different products, generating replenishment plans for "regular and irregular quantity," "regular and irregular quantity," and "irregular and irregular quantity."

[0049] Specifically, the intelligent replenishment device 30 can utilize a replenishment model based on point predictions to generate a "regular, irregular quantity" or "irregular, irregular quantity" replenishment plan. Alternatively, the intelligent replenishment device 30 can utilize a hybrid replenishment model based on point and interval predictions to generate a "regular and irregular quantity" replenishment plan. Alternatively, the intelligent replenishment device 30 can utilize a replenishment model based on interval predictions to generate an "irregular, irregular quantity" replenishment plan. Generating a replenishment plan using the most appropriate replenishment model based on factors such as the characteristics of different products and different replenishment cycles can provide replenishment decisions that better meet demand.

[0050] in, Figure 2 This example illustrates a connection between the smart supply chain system 100 and an external database 200, obtaining historical data from the external database 200. However, this embodiment is not limited thereto; the database 200 may also be stored in a built-in storage device within the smart supply chain system 100. Furthermore, the connection between the smart supply chain system 100 and the external database 200 may be a wired or wireless connection, which is not limited here.

[0051] According to the intelligent supply chain system 100 of this embodiment, supply chain management is more than just the supply chain. Relying on big data and information systems, it effectively integrates three key areas: product classification, sales forecasting, and intelligent replenishment. This system integrates internal and external data elements, enabling coordinated action across various systems under information guidance, maximizing service energy and unleashing service capabilities in an orderly manner. This enables precise service, transparent, flexible, and agile supply chains, and enhanced collaboration across various functions.

[0052] Two specific examples of the hardware structure of the smart supply chain system 100 according to this embodiment are described below. Figure 3 1 is a block diagram showing an example of the hardware structure of the smart supply chain system 100 according to the first embodiment of the present invention. Figure 3 As shown, the smart supply chain system 100 is implemented by a computer system having a processor 110, a memory 120, an interface 130, an input device 140 and a display unit 150. Figure 2 The processor 110 , the memory 120 , the interface 130 , the input device 140 and the display unit 150 are connected to each other via a bus 160 .

[0053] Specifically, the processor 110 , such as a CPU or microprocessor, implements the functions of each device in the smart supply chain system 100 by executing application programs stored in the memory 120 . The interface 130 , such as a communication interface, enables data communication with the database 200 . The input device 140 , such as a keyboard, mouse, microphone, or other input device, allows the user to input commands. The display unit 150 , such as a liquid crystal display, can display images related to the processing process and results of the smart supply chain system 100 .

[0054] The smart supply chain system 100 based on the hardware structure of this example can be set up at sales stores, suppliers, and manufacturers as application objects, and used by sales stores, suppliers, and manufacturers as users.

[0055] Figure 4 FIG. 1 is a block diagram showing another example of the hardware structure of the smart supply chain system 100 according to the first embodiment of the present invention. Figure 4 As shown, the smart supply chain system 100 includes a server platform 100A and a client device 100B capable of data communication with each other. The server platform 100A includes a processor 110A, a memory 120A, and an interface 130A interconnected via a bus 160A. The client device 100B includes a processor 110B, a memory 120B, an interface 130B, an input device 140, and a display unit 150 interconnected via a bus 160B.

[0056] Specifically, the processor 110A of the server platform 100A, such as a CPU or microprocessor, executes application programs stored in the memory 120A to implement the functions of the various devices of the smart supply chain system 100. Furthermore, the interface 130A of the server platform 100A, such as a communication interface, enables data communication with the interface 130B of the client device 100B and / or the database 200.

[0057] The processor 110B of the client device 100B, such as a CPU or microprocessor, implements the user interface by executing application programs stored in the memory 120B. The input device 140, such as a keyboard, mouse, or microphone, allows the user to input commands. The display unit 150, such as a liquid crystal display, can display images related to the processing process and results of the smart supply chain system 100.

[0058] In the smart supply chain system 100 based on the hardware structure of this example, the server platform 100A can be installed at the service provider, and the client device 100B can be installed at the sales outlets, suppliers, and manufacturers that are the application targets. Thus, the service provider provides the above-mentioned functions of the smart supply chain system 100 to the sales outlets, suppliers, and manufacturers that are the users.

[0059] The following describes the method performed by the smart supply chain system 100 of this embodiment. Figure 5 This is a flowchart showing a method executed by the smart supply chain system 100 according to the first embodiment of the present invention. Here, an example of the method executed by the smart supply chain system 100 according to this embodiment is specifically described, but the method executed by the smart supply chain system 100 according to this embodiment is not limited to the following example.

[0060] like Figure 5 As shown, in step S1, the product classification device 100 classifies products so that corresponding sales and inventory strategies can be adopted for different product categories. Specifically, based on the aforementioned R-ABC classification method, various product characteristic analysis indicators are combined to subdivide the products and select products whose demand forecast results require confirmation by sales personnel.

[0061] In step S2, the sales forecasting device 20 performs sales forecasting to formulate a demand plan. As described above, based on the classification results in step S1, the sales forecasting device 20 uses, for example, point forecasting based on a long short-term memory network or probability density interval forecasting based on the DeepAR algorithm for commodities for which demand forecast results need to be confirmed.

[0062] In step S3, the intelligent replenishment device 30 applies the automatic replenishment model to a specific scenario to generate a replenishment decision. As described above, the intelligent replenishment device 30 can utilize a replenishment model based on point prediction, a hybrid replenishment model based on point and interval prediction, or a replenishment model based on interval prediction, depending on the prediction method and prediction results in step S2.

[0063] (Second embodiment)

[0064] The second embodiment of the present invention will be described in detail below. Figure 6 1 is a functional block diagram showing a smart supply chain system 100' according to a second embodiment of the present invention. Figure 6 As shown, the smart supply chain system 100 ′ of this embodiment is based on the smart supply chain system 100 of the first embodiment, with the functions of the sales forecasting device 20 and the smart replenishment device 30 modified, and a model selection unit 40 added. Detailed description is given below.

[0065] like Figure 6As shown, the sales forecasting device 20 includes a point prediction unit 21 and an interval prediction unit 22. The point prediction unit 21 performs point predictions based on a long short-term memory network, while the interval prediction unit 22 performs probability density interval predictions. Furthermore, the intelligent replenishment device 30 includes a point prediction replenishment unit 31, a mixed replenishment unit 32, and an interval prediction replenishment unit 33. The point prediction replenishment unit 31 uses a replenishment model based on point predictions to generate regular or irregular replenishment plans. The mixed replenishment unit 32 uses a hybrid replenishment model based on point and interval predictions to generate regular and fixed replenishment plans. The interval prediction replenishment unit 33 uses a replenishment model based on interval predictions to generate irregular replenishment plans.

[0066] Furthermore, the smart supply chain system 100' further includes a model selection unit 40. The model selection unit 40 can be implemented by hardware independent of the product classification device 10, the sales forecasting device 20, and the intelligent replenishment device 30, or can be implemented by the same hardware as any one of the product classification device 10, the sales forecasting device 20, and the intelligent replenishment device 30. Furthermore, this embodiment can be applied to Figure 3 The hardware structure shown can also be applied to Figure 4 The hardware structure shown.

[0067] The model selection unit 40 selects whether the sales forecast is to be performed by the point prediction unit 21, the interval prediction unit 22, or both the point prediction unit 21 and the interval prediction unit 22 in the sales forecasting device 20, based on at least one of the application target (e.g., sales outlet, supplier, manufacturer, etc.), the prediction time (e.g., one day, one week, one month, one year, etc.), and the classification result of the commodity classification device 10. For example, the time required for prediction can be determined based on whether the application target is a sales outlet, supplier, or manufacturer, and then the selection can be made as to whether the point prediction unit 21 performs point prediction based on the long short-term memory network or whether the interval prediction unit 22 performs probability density interval prediction. This allows the most appropriate prediction method to be selected for sales forecasting based on the application target, prediction time, classification result, etc.

[0068] Furthermore, the model selection unit 40 selects which of the point-based replenishment unit 31, the mixed replenishment unit 32, or the interval-based replenishment unit 33 in the intelligent replenishment device 30 will generate the replenishment plan based on at least one of the application target, the forecast time, the classification results of the product classification device 10, and the forecast results of the sales forecast device 20. For example, the most appropriate replenishment model can be selected to generate the replenishment plan based on whether the application target is a sales store, supplier, or manufacturer, as well as the forecast model and forecast results selected in the sales forecast device 20.

[0069] The model selection section 40 can select the model according to the pre-stored model selection reference table, for example. The selection method of the model selection section 40 is not limited to this, and the selection result can be directly input from the outside, or the selection result based on the model selection reference table can be further modified by external input, or a selection scheme can be generated by applying machine learning to the past selection result. Figure 7 is a table showing an example of a model selection reference table referred to by the model selection section 40 of the second embodiment of the present application, and will be described in detail below.

[0070] As shown in Figure 7 , in the case where the application object of the intelligent supply chain system 100' is a sales store, the sales store needs to monitor the sales of each product, so as to formulate sales and inventory strategies. Therefore, in combination with the requirements of the replenishment plan, the product classification device 10 is subdivided into each product for which the demand prediction result needs to be confirmed, and the demand situation of each product is simultaneously predicted on a daily and weekly basis. The model selection section 40 selects the point prediction model of the long short-term memory network corresponding to the daily and weekly prediction of each product by the point prediction section 21 in the sales prediction device 20, and predicts the sales corresponding to 1 day and 1 week for each product. Furthermore, the model selection section 40 selects the replenishment plan generated by the point prediction replenishment section 31 in the intelligent replenishment device 30, and generates a replenishment plan of a regular and irregular quantity or a irregular and irregular quantity by using the replenishment model based on the point prediction. In this way, the appropriate prediction time and prediction model can be determined in combination with the characteristics of the sales store as the application object, and the appropriate replenishment model can be used, so that the intelligent supply chain system 100' can be effectively applied to the sales store.

[0071] In addition, as shown in Figure 7As shown, in the case where the application target is a supplier, they source goods from various manufacturers and distribute them to multiple stores. Therefore, they need to place orders with the manufacturers and simultaneously arrange distribution plans for the stores. Since the supplier connects manufacturers and stores, they need to understand both long-term and short-term demand for their products. Therefore, in accordance with the requirements of the replenishment plan, the product classification device 10 is subdivided into each product for which demand forecast results need to be confirmed. Daily, monthly, and annual demand forecasts are then simultaneously performed for each product. The model selection unit 40 selects both the point prediction unit 21 and the interval prediction unit 22 in the sales forecasting device 20 to perform sales forecasting. The point prediction unit 21 constructs a point prediction model using a long short-term memory network corresponding to one day and one year, while the interval prediction unit 22 constructs a probability density interval prediction model corresponding to one month. Sales volume for each product is predicted for one day, one month, and one year, respectively. Furthermore, the model selection unit 40 selects the hybrid replenishment unit 32 in the intelligent replenishment device 30 to generate a replenishment plan. This hybrid replenishment model, based on both point and interval predictions, generates a regular and quantitative replenishment plan. Therefore, it is possible to determine appropriate forecast time and forecast model based on the characteristics of suppliers as application objects, and use appropriate replenishment models to effectively apply the smart supply chain system 100 ′ to suppliers.

[0072] In addition, if Figure 7 As shown, in the case where the application target is a manufacturer, the manufacturer produces a variety of goods and supplies them to multiple different suppliers, who then supply the goods to sales outlets. Each product is in a different life cycle, and future demand forecasts are needed to guide future production plans. Because each batch of production produces a large number of goods, supply can be provided over a long period of time. Therefore, the product classification device is subdivided into each product for which demand forecast results need to be confirmed, and monthly demand forecasts are performed. The model selection unit 40 selects the interval prediction unit 22 in the sales forecasting device 20 to perform sales forecasting, builds a probability density interval prediction model corresponding to one month, and predicts sales volume for each product corresponding to one month. Furthermore, the model selection unit 40 selects the interval prediction and replenishment unit 33 in the intelligent replenishment device 30 to generate a replenishment plan. Using the replenishment model based on interval prediction, a replenishment plan with irregular periods and quantities is generated. In this way, the characteristics of the manufacturer as the application target can be combined to determine the appropriate forecast time and forecast model, and using the appropriate replenishment model, the intelligent supply chain system 100' can be effectively applied to the manufacturer.

[0073] The above-mentioned forecast time is merely an example, and other forecast times may be used depending on the situation. Alternatively, the forecast time may be specified by the user. Specifically, the model selection unit 40 may select an appropriate forecast model and replenishment model based on the input application and forecast time.

[0074] (Specific example 1)

[0075] The following describes a specific example of the present invention. Here, this specific example is implemented based on the second embodiment, but when the model to be used in sales forecasting and intelligent replenishment is determined in advance, this specific example can also be applied to the first embodiment. In addition, this specific example can be applied to Figure 3 The hardware structure shown can also be applied to Figure 4 The hardware structure shown.

[0076] This example is used to illustrate the application of the model in sales stores. Figure 8 This is a flowchart showing an application example (specific example 1) in which the smart supply chain system 100 ′ of the present invention is applied to a sales store.

[0077] like Figure 8 As shown, in step S101 , the commodity classification device 10 classifies commodities using the R-ABC classification method.

[0078] Specifically, suppose the sales volume of each product at a particular store over the past three months is sorted from highest to lowest. The percentage of each product's sales volume to the total sales volume is calculated, and then the cumulative percentage is calculated. According to the ABC classification system, a cumulative percentage of 0%-80% is classified as Category A, a cumulative percentage of 80%-90% is classified as Category B, and a cumulative percentage of 90%-100% is classified as Category C. The percentages here can be adjusted based on actual circumstances.

[0079] However, traditional ABC analysis doesn't consider the temporal nature of products. For example, even if a product's sales total is 100, the importance of the product will differ depending on whether this 100 includes recent sales records or sales records from two months ago. Adding the dimension of "recent sales interval" effectively addresses this issue. Based on the ABC classification, products are divided into three levels along the temporal dimension, according to their "recent sales interval." The number of levels into which products are divided according to the temporal dimension is not limited to this and can be set as appropriate. For example, the classification results based on the R-ABC classification method are shown in the last column of Table 1.

[0080] Table 1:

[0081]

[0082]

[0083] Furthermore, assuming that the store sales records are as shown in Table 2, calculate the values ​​of the following product characteristic analysis indicators.

[0084] Table 2:

[0085] Store Number Product Number Sales Day Year moon week Customer membership number quantity SP01 09 2020 / 1 / 3 2020 1 1 C0001 5 SP01 09 2020 / 1 / 6 2020 1 2 C0002 20 SP01 09 2020 / 1 / 14 2020 1 3 C0001 6 SP01 09 2020 / 1 / 23 2020 1 4 C0001 10 SP01 09 2020 / 1 / 31 2020 1 5 C0002 30 SP01 09 2020 / 2 / 7 2020 2 6 C0001 6 SP01 09 2020 / 2 / 16 2020 2 8 C0003 10 SP01 09 2020 / 2 / 20 2020 2 8 C0001 8 SP01 09 2020 / 2 / 26 2020 2 9 C0001 5

[0086] Regarding demand probability, calculate the probability of actual sales results within a unit period. Regarding sales frequency, within a specified period, count sales results on a weekly basis, and count the number of weeks with actual sales results as the frequency. Regarding the coefficient of variation, use the coefficient of variation to determine the magnitude of demand fluctuations. Regarding the amplitude of variation, confirm the fluctuations in total demand within a specified period. Regarding customer concentration, confirm whether sales within a specified period are concentrated on a small number of customers. Regarding sales cadence, count the time interval from each sales record to the next sales record on a commodity basis. Regarding release interval, count the time interval from the earliest sales date of the commodity to the current analysis time point on a commodity basis. The longer the interval, the longer the release time, and the shorter the interval, the shorter the release time.

[0087] The characteristics of item 09 are as follows: Product 09 has high overall sales volume and recent sales records, making it the current top-selling product. It has a high sales frequency and a small demand volatility coefficient, but its overall monthly fluctuation is large. Customer concentration is high, with C0002 as the primary customer. The average sales cycle is 7 days, and the release interval is 67 days. Based on the pre-defined rule table, we can determine that item 09 is a top-selling product group with a high sales frequency and a relatively sufficient data sample. Future demand can be predicted through data analysis and machine learning.

[0088] Next, in step S102, the model selection unit 40 selects a model to be used in sales forecasting and intelligent replenishment. Specifically, the model selection unit 40 selects a model to be used in sales forecasting and intelligent replenishment. Figure 7 The model selection reference table shown, for this specific example, selects both daily and weekly forecasts for each product's demand, and selects the point forecasting unit 21 in the sales forecasting device 20 to perform sales forecasting. Here, two sales volume forecasting models are constructed based on the long-short-term memory model. Furthermore, the model selection unit 40 selects the point forecasting unit 31 in the intelligent replenishment device 30 to establish a regular variable-quantity replenishment model and an irregular variable-quantity replenishment model.

[0089] Next, in step S103, the sales forecasting device 20 obtains the sales history data of each sales store, and its attributes are shown in Table 3.

[0090] Table 3:

[0091]

[0092] Next, we process the data, using the Keras algorithm to build long-short-term memory (LSTM) network models for daily and weekly forecasts. We then perform post-processing on the forecast results based on the forecast subject and output them in appropriate formats, such as web pages and reports. In this example, we construct a forecast results table in the database. The structure and example are shown in Table 4.

[0093] Table 4:

[0094]

[0095]

[0096] Next, in step S104, the intelligent replenishment device 30 calculates the maximum inventory level. Based on the prediction results from step S103, the periodic, variable-quantity replenishment model based on point forecasting is applied to calculate the minimum inventory level under the periodic replenishment model. If the current inventory level is lower than the minimum inventory level, a replenishment plan is generated to replenish the inventory to the calculated maximum inventory level. Furthermore, based on the prediction results from step S103, the irregular, variable-quantity replenishment model based on point forecasting is applied to calculate the minimum inventory level under the irregular replenishment model. If the current inventory level is lower than the minimum inventory level, a replenishment plan is generated to replenish the inventory to the calculated maximum inventory level.

[0097] In this example, assuming the company expects a two-week inventory cycle, we use the demand probability and sales frequency to determine the maximum inventory level for each product, along with the inventory turnover period and safety factor, as shown in Table 5. The safety factor can be adjusted based on actual conditions.

[0098] Inventory turnover period for each product = demand probability for each product * 2 (rounded to the nearest integer)

[0099] The period in this example is 2 months (9 weeks), and the sales frequency is divided into five levels: high, medium-high, medium, medium-low, and low according to the following rules.

[0100] Table 5:

[0101] Sales frequency Demand probability Inventory turnover period Service Level Safety factor 8~9 (high) ≧0.8 2 weeks 99.90％ 3.1 6-8 (medium to high) ≧0.6and<0.8 2 weeks 99％ 2.33 3-6 (medium) ≧0.4and<0.6 1 week 98％ 2 2-3 (medium-low) ≧0.2and<0.4 1 week 97％ 1.88 0~2 (low) <0.2 1 week 95％ 1.65

[0102] In this example, item 09 is a high-frequency item, with an inventory turnover period of 2 weeks and a safety factor of 3.1. Therefore, the maximum inventory for item 09 is calculated according to the formula:

[0103] Maximum inventory level = demand forecast within the future inventory turnover period + safety stock (SS)

[0104] = Demand forecast for the next two weeks + (3.1 * standard deviation of daily demand performance over the past two weeks * √14) = (40 + 47) + (3.1 * 2.05 * 3.74) = 111

[0105] Next, using the periodic variable quantity replenishment model, we regularly check the current inventory status at a fixed time each week. Assuming the standard replenishment lead time is 3 days, we use the following formula to calculate the minimum inventory under the periodic replenishment model:

[0106] Minimum inventory turnover period (weeks) = Standard replenishment lead time / 7 days (rounded up to the nearest decimal point) = 3 / 7 = 1 week

[0107] Minimum inventory level = Demand forecast for the minimum inventory turnover period in the future + Safety stock (SS) = Demand forecast for the next week + (3.1 * Standard deviation of demand performance in the past week in daily units * √7) = 40 + 0 = 40

[0108] If the current inventory is less than 40, assuming the current inventory is 30, the production replenishment plan, the replenishment quantity is 111-30=81

[0109] Next, using the irregular replenishment model, check the current inventory status in real time every day. Assuming the standard replenishment lead time is 3 days, use the following formula to calculate the minimum inventory under the irregular replenishment model:

[0110] Minimum inventory = Demand forecast for the future standard replenishment lead time + Safety stock (SS) = Demand forecast for the next three days + (3.1 * Standard deviation of actual demand for the past three days * √3) = (5 + 10 + 6) + 0 = 21

[0111] If the inventory for the day is less than 21, assuming the current inventory is 18, generate a replenishment plan with a replenishment quantity of 111-18=93

[0112] (Specific example 2)

[0113] The following describes a specific example of the present invention. Here, this specific example is implemented based on the second embodiment, but when the model to be used in sales forecasting and intelligent replenishment is determined in advance, this specific example can also be applied to the first embodiment. In addition, this specific example can be applied to Figure 3 The hardware structure shown can also be applied to Figure 4 The hardware structure shown.

[0114] This example is used to illustrate the application of the model in suppliers. Figure 9 This is a flowchart showing an application example (concrete example 2) in which the smart supply chain system 100 ′ of the present invention is applied to a supplier.

[0115] like Figure 9 As shown, in step S201, the commodity classification device 10 classifies commodities using the R-ABC classification method. This step can be compared with the specific example 1 ( Figure 8 The process is the same as step S101, and the detailed description is omitted here.

[0116] Next, in step S202, the model selection unit 40 selects a model to be used in sales forecasting and intelligent replenishment. Specifically, the model selection unit 40 selects a model to be used in sales forecasting and intelligent replenishment. Figure 7The model selection reference table shown is selected for the case of the supplier in this specific example, the demand status of each commodity is simultaneously predicted on a daily, monthly, and yearly basis, and the sales forecast in the sales forecasting device 20 is performed by the point forecasting unit 21 and the interval forecasting unit 22 together, where two sales forecasting models are constructed for daily and yearly forecasts based on the long short-term memory model, and a sales forecasting model is constructed for monthly forecasts based on the probability density interval forecasting model. Furthermore, the model selection unit 40 selects the periodic and quantitative replenishment model established by the mixed replenishment unit 31 in the intelligent replenishment device 30.

[0117] Next, in step S203, the sales forecasting device 20 obtains historical data for the different stores supplied by the supplier, the attributes of which are as shown in Table 6.

[0118] Table 6:

[0119]

[0120] Data processing is then performed, the long short-term memory network model for daily and yearly forecasts is constructed using the keras algorithm, the probability density interval forecasting model for monthly forecasts is constructed using the DeepAR model, and the prediction results are processed accordingly according to the different prediction subjects, and output in the form of a webpage, report, or the like. In this example, a prediction result table is constructed in the database, the structure of which is as shown in Tables 1 and 2.

[0121] Table 7:

[0122]

[0123] Table 8:

[0124]

[0125] Next, in step S204, the intelligent replenishment device 30 generates a replenishment plan based on the prediction results of step S203. For a supplier, it is generally necessary to try to meet customer order demand, but there is no direct contact with the end market, and market feedback delays may occur, so the replenishment needs to have the ability to cope with risks to a certain extent, while controlling costs to achieve optimal operation.

[0126] For example, for the replenishment scenario of this supplier, the service level coefficient is defined as 95%, the current inventory is 45 pieces, the order lead time is 5 days, the cost of each order is 50 yuan / time, and the holding cost of each commodity is 30 yuan / piece. The replenishment quantity and the demand in the period are both selected as the prediction median M.

[0127] (1). Calculate the optimal economic replenishment cycle based on the EOQ (Economic Order Quantity) model, taking the above commodity 09 as an example:

[0128]

[0129] c e is the ordering cost per period, C t is the storage cost of each commodity during the period, and D is the demand during the period (predicted median M).

[0130] Optimal economic replenishment cycle = 365 / (annual demand forecast / optimal economic replenishment quantity) = 365 / (3985 / 69) = 6

[0131] (2) Safety stock establishment

[0132] Safety stock of item 09 = (monthly demand standard deviation / 30*6)*service level factor = (23 / 30*6)*1.68=8

[0133] (3) Establishment of restocking points

[0134] The restocking point for the above product 09 = sales forecast within the replenishment lead time + safety stock = 33 + 8 = 41

[0135] (4) Determine the replenishment quantity

[0136] The replenishment quantity of the above product 09 on March 1, 2020 = 2 * demand forecast during the replenishment lead time + demand forecast during the replenishment cycle + safety stock – current inventory = 2 * 33 + (325 / 30 * 6) + 8 - 45 = 94

[0137] Taking the predicted result data as an example, its replenishment plan is shown in Table 3.

[0138] Table 9:

[0139]

[0140] This means that for the above-mentioned product 09, regular replenishment will be carried out every 6 days during the period from 2020.3.1 to 2020.4.1. If the inventory is less than 41 during the replenishment period, replenishment will begin. The first replenishment on 2020.3.1 was 94 pieces.

[0141] (Concrete example 3)

[0142] The following describes a specific example of the present invention. Here, this specific example is implemented based on the second embodiment, but when the model to be used in sales forecasting and intelligent replenishment is determined in advance, this specific example can also be applied to the first embodiment. In addition, this specific example can be applied to Figure 3 The hardware structure shown can also be applied to Figure 4 The hardware structure shown.

[0143] This example is used to illustrate the application of the model in a manufacturer. Figure 10 This is a flowchart showing an application example (specific example 3) in which the smart supply chain system 100 ′ of the present invention is applied to a manufacturer.

[0144] like Figure 10 As shown, in step S301, the commodity classification device 10 classifies commodities using the R-ABC classification method. This step can be compared with the specific example 1 ( Figure 8 The process is the same as step S101, and the detailed description is omitted here.

[0145] Next, in step S302, the model selection unit 40 selects a model to be used in sales forecasting and intelligent replenishment. Specifically, the model selection unit 40 selects a model to be used in sales forecasting and intelligent replenishment. Figure 7 The model selection reference table shown, for this specific example, applies to a manufacturer. It selects to perform a monthly demand forecast for each product simultaneously. It also selects to have the interval forecasting unit 22 in the sales forecasting device 20 perform sales forecasts, building a monthly sales volume forecast model based on the probability density interval forecast model. Furthermore, the model selection unit 40 selects to have the interval forecasting replenishment unit 33 in the intelligent replenishment device 30 build an irregular replenishment model.

[0146] Next, in step S303 , the sales forecasting device 20 obtains the sales history data of each product of the manufacturer, the attributes of which are shown in Table 10.

[0147] Table 10:

[0148]

[0149] Next, data processing is performed, using the DeepAR model to construct a probability density interval prediction model for the monthly forecast. The forecast results are then post-processed based on the forecast subject and output in appropriate formats such as web pages and reports. In this example, a forecast results table is constructed in the database. The structure and example are shown in Table 11. This table provides data support for subsequent replenishment strategies.

[0150] Table 11:

[0151]

[0152] Next, in step S304, the intelligent replenishment device 30 generates a replenishment plan based on the forecast results from step S303. For this manufacturer, in a scenario where they need to supply goods to downstream suppliers, monthly forecast data is used. For example, a service level coefficient of 95% is defined, an order lead time of 5 days, an order cost of 50 yuan per order, and a holding cost of 10 yuan per item are defined. The forecast median M is used for both the daily inventory quantity and the periodic demand.

[0153] First, determine the optimal replenishment point, taking product 02 as an example:

[0154] Safety stock of item 02 = Standard deviation of demand for each SKU * Service level coefficient = 3021 * 1.68 = 5075

[0155] The stock quantity of product 02 during the order lead time = the number of days of order lead time * the daily stock quantity (forecast median M) = 5 * (10025 / 22) = 2278 (Note: there are 22 working days in a month)

[0156] Dynamic replenishment point for item 02 = safety stock + stock quantity during order lead time = 7353

[0157] Next, we determine the optimal economic replenishment quantity based on the EOQ model, taking item 02 as an example:

[0158]

[0159] C e is the ordering cost per period, C t is the storage cost of each commodity during the period, and D is the demand during the period (predicted median M).

[0160] Taking the forecast data as an example, the replenishment plan is shown in Table 12:

[0161] Table 12:

[0162] Prediction granularity Product Number Forecast start date Forecast end date Replenishment Point Replenishment Monthly Forecast 02 2020.3.1 2020.4.1 7353 63 Monthly Forecast 04 2020.3.1 2020.4.1 9593 126

[0163] This means that if the inventory of item 02 is less than 7353 between March 1, 2020, and April 1, 2020, replenishment will begin at 63 units per replenishment. If the inventory of item 04 is less than 9593 between March 1, 2020, and April 1, 2020, replenishment will begin at 126 units per replenishment.

[0164] The specific embodiments and specific examples of the present invention are described above with reference to the accompanying drawings. The specific embodiments and specific examples described above are merely specific examples of the present invention, which are used to understand the present invention and are not intended to limit the scope of the present invention. Those skilled in the art can make various modifications, combinations, and reasonable omissions of elements to the specific embodiments and specific examples based on the technical ideas of the present invention, and the resulting methods are also included in the scope of the present invention. For example, the above-mentioned embodiments and specific examples can be combined with each other, and the embodiments formed by the combination are also included in the scope of the present invention.

Claims

1. A smart supply chain system that manages the supply chains of multiple commodities, characterized by: have: A commodity classification device that classifies a variety of commodities based on historical data; a sales forecasting device for forecasting sales of each commodity based on historical data and the classification results of the commodity classification device for the plurality of commodities; and The intelligent replenishment device generates replenishment decisions based on the prediction results of the sales forecasting device and applies an automatic replenishment model. The sales forecasting device comprises: Point prediction unit, which performs point prediction based on long short-term memory network; and The interval prediction department performs probability density interval prediction; The smart supply chain system also has: a model selection unit that selects, based on at least one of an application object, a forecast time, and a classification result of the commodity classification unit, whether the sales forecast is to be performed by the point forecast unit, the interval forecast unit, or both the point forecast unit and the interval forecast unit in the sales forecast device; The commodity classification device classifies a plurality of commodities using the ABC classification results based on the sales data as historical data and the time interval between recent sales as two dimensions to obtain a preliminary classification result, and calculates the commodity characteristic analysis index of each commodity based on the historical data. The calculated commodity characteristic analysis index is further used to perform multi-dimensional segmentation on the preliminary classification result to obtain the final classification result. The commodity characteristic analysis indicator includes at least one of sales frequency, coefficient of variation, amplitude of variation, customer concentration, sales rhythm, and release interval.

2. The intelligent supply chain system according to claim 1, characterized in that: The sales forecasting device uses one of the prediction methods of point prediction, interval prediction, and hybrid prediction combining point prediction and interval prediction for at least one prediction time for each commodity that is subdivided by the commodity classification device into commodities that need to confirm the demand forecast results, and predicts the sales volume corresponding to the at least one prediction time for each commodity.

3. The intelligent supply chain system according to claim 2, characterized in that: The intelligent replenishment device generates a replenishment plan based on the prediction results of the sales forecasting device and the commodity characteristic analysis index calculated by the commodity classification device, using one of the replenishment models selected from the group consisting of a point prediction-based replenishment model, a hybrid replenishment model based on point prediction and interval prediction, and a replenishment model based on interval prediction.

4. The intelligent supply chain system according to claim 1, wherein: The intelligent replenishment device has: The point forecast replenishment department uses a replenishment model based on point forecasts to generate regular or irregular replenishment plans. The hybrid replenishment department generates regular and quantitative replenishment plans using a hybrid replenishment model based on point and interval forecasts; and The interval forecast replenishment department uses the replenishment model based on interval forecast to generate irregular replenishment plans. The model selection unit selects which of the point prediction replenishment unit, the mixed replenishment unit, and the interval prediction replenishment unit is to generate a replenishment plan in the intelligent replenishment device based on at least one of an application object, a forecast time, a classification result of the product classification unit, and a forecast result of the sales forecast unit.

5. The intelligent supply chain system according to claim 4, characterized in that: In the case where the application object is a sales store, For each commodity that needs to be confirmed as a demand forecast result by the commodity classification device, the model selection unit selects the point prediction unit in the sales forecast device to perform sales forecasting, builds a point prediction model of a long short-term memory network corresponding to at least one forecast time, and predicts the sales volume corresponding to the at least one forecast time for each commodity. Furthermore, the model selection unit selects the point prediction replenishment unit in the intelligent replenishment device to generate a replenishment plan, and uses a replenishment model based on point prediction to generate a replenishment plan with regular or irregular quantities.

6. The intelligent supply chain system according to claim 4, characterized in that: In the case where the application object is a supplier, For each commodity that needs to be confirmed in the demand forecast result, the model selection unit selects the sales forecast to be performed by both the point forecast unit and the interval forecast unit in the sales forecast device, builds a point forecast model of a long short-term memory network corresponding to at least one forecast time among a plurality of forecast times, and a probability density interval forecast model corresponding to at least one other forecast time among the plurality of forecast times, and predicts the sales volume corresponding to each of the plurality of forecast times for each commodity. Furthermore, the model selection unit selects the hybrid replenishment unit in the intelligent replenishment device to generate a replenishment plan, and uses a hybrid replenishment model based on point prediction and interval prediction to generate a regular and quantitative replenishment plan.

7. The intelligent supply chain system according to claim 4, wherein: In the case where the application object is a manufacturer, For each commodity that needs to be confirmed as a demand forecast result by the commodity classification device, the model selection unit selects the interval prediction unit in the sales forecast device to perform sales forecasting, builds a probability density interval forecasting model corresponding to at least one forecast time, and predicts the sales volume corresponding to the at least one forecast time for each commodity. Furthermore, the model selection unit selects the interval prediction replenishment unit in the intelligent replenishment device to generate a replenishment plan, and uses the replenishment model based on interval prediction to generate an irregular and indefinite replenishment plan.

8. A server platform for managing the supply chain of multiple commodities, comprising a processor, a memory, and an interface, capable of communicating data with a client device via the interface, characterized in that: The processor executes the following processing by executing the program stored in the memory: Commodity classification processing, classifying a variety of commodities based on historical data; Sales forecasting processing, based on historical data and the classification results of the commodity classification processing for the plurality of commodities, making sales forecasts for each commodity; and Intelligent replenishment processing, based on the forecast results of the sales forecast processing, applies the automatic replenishment model to generate replenishment decisions, The replenishment decision is sent to the client device via the interface, The sales forecasting process can: Point prediction based on long short-term memory networks; and Make probability density interval predictions; The processor further performs the following processing by executing the program stored in the memory: Based on at least one of an application object, a forecast time, and a classification result of the commodity classification process, selecting whether to perform sales forecasting by the point forecast, the probability density interval forecast, or both the point forecast and the probability density interval forecast in the sales forecasting process; In the commodity classification process, a plurality of commodities are classified using the ABC classification results based on the sales data as historical data and the time interval of the most recent sales as two dimensions to obtain preliminary classification results, and commodity characteristic analysis indicators of each commodity are calculated based on the historical data. The calculated commodity characteristic analysis indicators are further used to perform multi-dimensional segmentation on the preliminary classification results to obtain the final classification results. The commodity characteristic analysis indicator includes at least one of sales frequency, coefficient of variation, amplitude of variation, customer concentration, sales rhythm, and release interval.

Images