Management method and system for intelligent replenishment based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence and the like, and provides an intelligent replenishment management method and system based on artificial intelligence, and the method comprises the steps: calculating the historical order data of a target product on an e-commerce platform, and calculating the historical inventory data of the target product in each regional warehouse; and analyzing the historical order data and the historical inventory data through a machine learning model to obtain the correlation characteristics of the orders and the inventory of the target product in different time periods and different areas, training an intelligent inventory allocation model through the correlation characteristics of the orders and the inventory, so that the intelligent inventory allocation model learns to automatically adjust the inventory allocation, and the intelligent inventory allocation efficiency is improved. And according to the regional replenishment demand of the target product, the trained intelligent inventory allocation model is utilized to analyze the current order of the target product and the current inventory state of the target product in each regional warehouse to generate an inventory allocation plan of the target product in each regional warehouse, so that replenishment management adapts to the demand change, and the replenishment efficiency is improved. And regional requirements are quickly responded.

Description

Technical Field

[0001] The present invention relates to the technical fields of artificial intelligence, Internet services, etc., and in particular to a management method and system for intelligent replenishment based on artificial intelligence. Background Art

[0002] In current replenishment management technologies, inventory management mostly relies on simple sales forecasts and fails to take into account regional inventory data, resulting in uneven supply and demand distribution. Moreover, inventory management often relies on fixed safety stock levels or simple linear demand forecasting models, which are difficult to adapt to the diversification of demand changes and difficult to respond quickly to regional demand.

[0003] In summary, existing replenishment management technologies have technical problems such as difficulty in adapting to changes in demand and difficulty in quickly responding to regional demand. Summary of the invention

[0004] In view of the deficiencies in the above-mentioned prior art, the present invention provides an intelligent replenishment management method and system based on artificial intelligence, so as to enable replenishment management to adapt to changes in demand and quickly respond to regional demand.

[0005] In a first aspect, the present invention provides a management method for intelligent replenishment based on artificial intelligence, comprising:

[0006] Calculate the historical order data of the target product on the e-commerce platform, and calculate the historical inventory data of the target product in each regional warehouse;

[0007] Analyze the historical order data and the historical inventory data through a machine learning model to obtain correlation features between orders and inventories of the target product in different regions at different time periods;

[0008] Training an intelligent inventory transfer model through the correlation features of the order and inventory, so that the intelligent inventory transfer model learns to automatically adjust inventory transfer to adapt to the regional replenishment demand of the target product;

[0009] The trained intelligent inventory transfer model is used to analyze the current order of the target product and the current inventory status of the target product in each regional warehouse to generate an inventory transfer plan for the target product in each regional warehouse.

[0010] In a second aspect, the present invention provides an artificial intelligence-based intelligent replenishment management system, and the artificial intelligence-based intelligent replenishment management system applies the above-mentioned artificial intelligence-based intelligent replenishment management method.

[0011] Compared with the prior art, the present invention has the following beneficial effects:

[0012] The present invention provides a management method and system for intelligent replenishment based on artificial intelligence. The method calculates historical order data generated by a target product on an e-commerce platform and historical inventory data of the target product in each regional warehouse, and analyzes the historical order data and the historical inventory data through a machine learning model to obtain correlation features between orders and inventories of the target product in different regions at different time periods. An intelligent inventory transfer model is trained through the correlation features between the orders and inventories, so that the intelligent inventory transfer model learns to automatically adjust inventory transfers to adapt to regional replenishment needs of the target product. The trained intelligent inventory transfer model is used to analyze the current order of the target product and the current inventory status of the target product in each regional warehouse to generate an inventory transfer plan for the target product in each regional warehouse, thereby enabling replenishment management to adapt to demand changes and quickly respond to regional needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention and do not constitute an improper limitation of the present invention. Some specific embodiments of the present invention will be described in detail in an exemplary and non-restrictive manner with reference to the drawings. The same reference numerals in the drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. In the drawings:

[0014] Figure 1 It is a flow chart of a management method for intelligent replenishment based on artificial intelligence according to an embodiment of the present invention;

[0015] Figure 2 It is a schematic diagram of the architecture of an intelligent replenishment management system based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only an embodiment of a part of the present invention, not all 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.

[0017] Embodiment 1

[0018] See also Figure 1-Figure 2 , this embodiment provides a management method for intelligent replenishment based on artificial intelligence, comprising the following steps:

[0019] S101. Calculate the historical order data of the target product on the e-commerce platform, and calculate the historical inventory data of the target product in each regional warehouse; wherein the historical order data may come from the e-commerce platform, and the historical inventory data may come from the warehouse management system.

[0020] S102, analyzing the historical order data and the historical inventory data by using a machine learning model to obtain correlation features of orders and inventories of the target product in different regions at different time periods;

[0021] S103, training an intelligent inventory transfer model through the correlation features of the order and inventory, so that the intelligent inventory transfer model learns to automatically adjust inventory transfer to adapt to the regional replenishment demand of the target product;

[0022] S104: Analyze the current order of the target product and the current inventory status of the target product in each regional warehouse by using the trained intelligent inventory transfer model to generate an inventory transfer plan for the target product in each regional warehouse.

[0023] It should be noted that in step S101, the historical order data generated by the target product on the e-commerce platform is calculated, and the historical inventory data of the target product in each regional warehouse is calculated, which can provide a data basis for subsequent analysis and model training. By calculating the historical order data and inventory data of the target product in different regions and time periods, the demand change trend and inventory status of the target product in each region can be obtained. For example, a target product (such as mobile phone accessories) is sold on an e-commerce platform, and the demand is different in different regions and time periods. By analyzing the order data of the past 6 months, it can be found that the order volume in some regions has increased significantly during holidays; and the inventory data can show that there is often a shortage or backlog of inventory in some regions. This information will help the model in the subsequent steps to understand the product demand and inventory situation. In step S102, the historical order data and the historical inventory data are analyzed by a machine learning model to obtain the correlation features of the orders and inventories of the target product in different regions in different time periods, and to extract the correlation features between historical orders and inventories. Changes in demand and inventory are often related to specific time, holidays, promotional activities and other factors. Machine learning can help discover these implicit patterns and provide a basis for subsequent inventory transfer models. For example, in the above-mentioned example of mobile phone accessories, after analysis by the machine learning model, it may be found that the demand in a certain area increased greatly during the Double Eleven period, and then the inventory decreased rapidly. At this time, the associated features may include the pattern of "festival-demand increase-inventory reduction", which will help the intelligent inventory allocation model to predict demand and allocate inventory in advance before the arrival of future festivals. In step S103, the intelligent inventory allocation model is trained through the associated features of the order and inventory, so that the intelligent inventory allocation model learns to automatically adjust the inventory allocation and adapt to the regional replenishment demand of the target product, thereby realizing the training of the intelligent inventory allocation model and enabling it to have the ability to automatically adjust inventory. The intelligent inventory allocation model grasps the demand pattern and inventory changes in different regions by learning the associated features of orders and inventory, so that it can adaptively adjust the inventory allocation in actual applications to reduce the situation of insufficient inventory or backlog. For example, through the aforementioned associated features, the intelligent inventory allocation model can learn that there will be a trend of inventory shortage in a certain area before the festival. In the future, when similar order and inventory patterns appear, the intelligent inventory allocation model can generate allocation requirements in advance to ensure that the inventory in the area is sufficient to avoid orders that cannot be met due to out-of-stock. In step S104, the trained intelligent inventory transfer model is used to analyze the current order of the target product and the current inventory status of the target product in each regional warehouse to generate an inventory transfer plan for the target product in each regional warehouse, thereby realizing real-time analysis of the current order and current inventory status using the trained intelligent inventory transfer model to adapt to the current replenishment demand and ensure that the inventory level of each regional warehouse remains dynamically balanced. For example, suppose that the current inventory status of a region is low, and the order volume suddenly increases.The intelligent inventory transfer model can generate a transfer plan based on the current order data and inventory status, and transfer part of the inventory from the area with sufficient inventory to the area with insufficient inventory, thereby avoiding out-of-stock situations caused by sudden increases in demand.

[0024] In some further embodiments, the correlation characteristics of the order and inventory include: correlation characteristics of order demand volatility and inventory consumption rate, correlation characteristics of safety stock and order fulfillment rate, correlation characteristics of order periodicity and inventory replenishment frequency, correlation characteristics of order burstiness and emergency inventory transfer, correlation characteristics of inventory turnover rate and seasonality of order demand, and correlation characteristics of inventory backlog and order demand saturation. It should be noted that order demand volatility and inventory consumption rate reflect the impact of changing demand trends on inventory levels. When demand fluctuates greatly, the inventory consumption rate is usually high, and inventory needs to be replenished in time; when demand is relatively stable, the inventory consumption rate is low, and a stable inventory level can be maintained. By analyzing the correlation between order volatility and inventory consumption rate, the intelligent inventory transfer model can learn to judge the stability or volatility of demand in different regions, and then determine the frequency and batch of replenishment or transfer. If the order volatility is high, the intelligent inventory transfer model can predict that inventory consumption will accelerate in the future, so as to transfer or replenish in advance. For example, the order demand in a certain area fluctuates greatly, especially when the order volume surges during the promotion period, resulting in a rapid increase in inventory consumption rate. After the intelligent inventory transfer model learns to identify this correlation feature, it can increase inventory before the peak demand period to avoid order delays due to insufficient inventory. In addition, the relationship between safety stock and order fulfillment rate reflects the degree to which inventory levels guarantee timely delivery of orders. A higher safety stock level can improve the order fulfillment rate, while a lower safety stock may lead to a decrease in the order fulfillment rate, affecting customer experience. By associating safety stock with order fulfillment rate, the intelligent inventory transfer model learns to dynamically adjust the inventory level to ensure that sufficient inventory is maintained to meet orders during high demand, helping the intelligent inventory transfer model learn to prioritize areas with insufficient inventory to prevent orders from being unfulfilled due to out-of-stock. For example, the safety stock level in a certain area is set low, resulting in an inability to meet demand when orders surge, and a decrease in order fulfillment rate. In addition, the relationship between the cyclical changes in orders and the frequency of inventory replenishment can reflect the impact of seasonal or cyclical demand on inventory. An increase in cyclical orders will lead to rapid inventory consumption, requiring more frequent inventory replenishment; while a decrease in cyclical orders can reduce the replenishment frequency. The association between order periodicity and inventory replenishment frequency can help the intelligent inventory transfer model learn to determine the best time to replenish inventory and avoid inventory shortages during peak demand periods. For example, in winter, the order volume of hand warmers increases cyclically, and inventory consumption also accelerates. After the intelligent inventory transfer model learns to identify the association between cyclical changes in orders and the frequency of inventory replenishment, it can increase the frequency of inventory replenishment for hand warmers before winter to ensure sufficient inventory during the peak season. In addition, the association between sudden changes in orders (such as temporary promotions, surges in market demand) and emergency inventory transfers reflects the pressure on inventory caused by a surge in orders in the short term. This sudden increase in order volume usually leads to a rapid decline in inventory, requiring emergency transfers to meet demand.By associating sudden changes in orders with urgent inventory transfers, the intelligent inventory transfer model can learn to quickly identify sudden order demands and make transfer decisions to prevent insufficient inventory caused by sudden demands. For example, during the "Double 12" promotion, the number of smartphone orders in a certain area increased suddenly, and the inventory decreased rapidly. After the intelligent inventory transfer model learns to identify the association characteristics of order suddenness and urgent inventory transfer, it can transfer inventory from neighboring warehouses to ensure that orders are met during the promotion period. In addition, the relationship between inventory turnover rate and seasonal changes in order demand can reflect the speed of inventory consumption when the number of orders increases in a specific season. This association feature can reveal the impact of seasonal order demand on inventory turnover rate. By analyzing the association between inventory turnover rate and seasonality of order demand, the intelligent inventory transfer model learns to prepare inventory in advance in a specific season, thereby preventing inventory shortages caused by increased seasonal demand. For example, the number of refrigerator orders usually increases in summer, resulting in an increase in inventory turnover rate. After the intelligent inventory transfer model learns to identify the association between this seasonal demand and inventory turnover rate, it increases the inventory of refrigerators before the arrival of summer to meet seasonal demand. In addition, the relationship between inventory backlog and order demand saturation reflects the situation where inventory cannot be consumed in time when order demand reaches saturation. When order demand decreases or tends to be saturated, the possibility of inventory backlog increases, and destocking measures need to be taken. The association between inventory backlog and order demand saturation can help the intelligent inventory transfer model learn to reduce unnecessary replenishment or transfer when demand is saturated, and reduce backlog costs through inventory destocking strategies. For example, the order demand in a certain area tends to be saturated, resulting in slower inventory turnover and inventory backlog. After the intelligent inventory transfer model learns to determine the correlation characteristics between inventory backlog and order demand saturation, it can take destocking measures to transfer excess inventory to other areas with higher demand to reduce backlogs.

[0025] In some preferred embodiments, before calculating the historical order data, a target product screening formula is set to screen the target product, and the target product screening formula is f(Product);

[0026] f(Product)=Category∩Brand∩Price_Range∩Sales_Range;

[0027] In the target product screening formula, Category is the category label of the target product, such as electronic products, clothing, etc.; Brand is the brand label of the target product; Price_Range is the price range of the target product; Sales_Range is the sales range of the target product, which is used to screen products that meet the sales standards; when f(Product)=1, the target product is screened.

[0028] It should be noted that the target product screening formula screens eligible target products by setting a series of conditions (category, brand, price range and sales range), so as to ensure that the screened products fully meet specific business needs, thereby improving the accuracy of data analysis. By setting different tags, the screening conditions can be flexibly adjusted to ensure that the selection of target products is more accurate. For example, an e-commerce platform may need to analyze the sales of a specific category of products under a certain brand, while paying attention to specific price and sales ranges. Through multiple screening conditions, it is possible to focus on specific product types without being affected by other unrelated products, thereby providing more customized analysis results. For example, if the target products are limited to electronic accessories with lower prices, only the order and inventory data of electronic accessories with prices that meet the conditions need to be analyzed, reducing the amount of calculation and storage requirements. For example, suppose an e-commerce platform wants to analyze the smartphone accessories that have performed well in sales in the past year, so as to make reasonable inventory allocation and stocking plans in the next promotion activities. The following screening conditions can be set: Category is selected as mobile phone accessories, Brand is selected as a specific brand (such as Brand A), Price_Range is set to products with a price range of 50-200 yuan, and Sales_Range is set to products with sales of more than 1,000 units. When f(Product) = 1, it means that all conditions are met and the product will be identified as the target product. Through this screening process, you can focus on the best-selling mobile phone accessories of Brand A, thereby providing data support for subsequent analysis.

[0029] In some preferred embodiments, when calculating the historical order data and the historical inventory data, a time window is selected according to demand to analyze the order data and inventory data of the target product in different time periods, and the time window is defined as T;

[0030] T={t1,t2,...,t n}; T contains each time period, t n Indicates each time period.

[0031] It should be noted that the time window setting allows historical order and inventory data to be divided into different time periods. Through such segmentation, the demand change trend and inventory status of the target product in different time periods can be analyzed, so as to better understand the cyclical or seasonal fluctuations in demand and inventory. In addition, the flexibility of the time window allows different time granularities to be selected according to different business needs. For example, an e-commerce platform can analyze the sales and inventory changes of target products based on days, weeks, months, or even quarters. This flexibility can better adapt to the analysis needs in different scenarios, such as short-term daily inventory transfers and long-term seasonal stocking planning.

[0032] In some preferred embodiments, when calculating the historical order data, the order quantity is calculated using a calculation formula, and the calculation formula for the order quantity is:

[0033]

[0034] In the calculation formula of the order quantity, O i，t，r represents the order volume of target product i in region r within time t; N t，r N t represents the total number of orders in region r at time t; S i，j Indicates whether the target product i is included in order j. If so, S i，j =1, otherwise 0.

[0035] It should be noted that the order quantity calculation formula is used to calculate the order quantity of the target product i in the region r within the time t. By checking whether each order j contains the target product one by one, it is ensured that only the order quantity related to the target product is calculated, thereby improving the data accuracy and ensuring that the subsequent inventory and transfer analysis is more reliable.

[0036] In some preferred embodiments, when calculating the historical inventory data, the calculation is performed using a calculation formula for the inventory amount, and the calculation formula for the inventory amount is:

[0037]

[0038] In the calculation formula of the inventory, I i，t，r represents the inventory of target product i in the warehouse of region r within time t, L t，r represents the total number of inventory records in region r at time t, S i，k represents the inventory quantity of target product i in inventory record k.

[0039] It should be noted that the inventory calculation formula is used to calculate the total inventory of target product i in region r within time t, by accumulating each inventory record one by one, to ensure that the calculated inventory is a true reflection of the actual inventory record. In addition, the formula counts the inventory according to time t and region r respectively, which can refine the inventory situation in each region and time period. The inventory demand and replenishment frequency may vary in different regions, so inventory statistics divided by region and time can understand the specific inventory status of each region, thereby providing support for more refined inventory management.

[0040] In some preferred embodiments, the historical inventory data includes regional inventory in and out records, and the regional inventory in and out records are calculated using a calculation formula for net inventory change, and the calculation formula for net inventory change is:

[0041]

[0042] In the calculation formula of the net inventory change, ΔI i，t，r represents the net inventory change of target product i in region r within time t, represents the total inventory quantity of target product i in region r within time t, Represents the total shipment quantity of target product i in region r within time t.

[0043] It should be noted that the calculation formula for the net inventory change is used to calculate the net inventory change of region r in i within time t. By calculating the difference between the total inventory inflow and the total outflow, the inventory change in the time period can be obtained, and the dynamic adjustment of the inventory in each time period can be captured, which is convenient for identifying the inventory consumption trend and replenishment demand. For example, if the net inventory change of a region is negative in multiple consecutive time periods, it may mean that the inventory demand in the region is high, the inventory is consumed quickly, and timely replenishment is required; conversely, if the net inventory change is positive, it indicates that there may be a backlog of inventory.

[0044] In some preferred embodiments, when analyzing the historical order data and the historical inventory data by a machine learning model, it includes: using a linear regression model to fit the order quantity. i，t，r and inventory I i，t，r The relationship between them is specifically:

[0045] O i，t，r =β0+β1·I i，t，r +∈;

[0046] β0 and β1 are regression coefficients, indicating the strength of the impact of inventory on orders; ∈ is the error term, which is used to represent the part that the linear regression model cannot explain;

[0047] The linear regression model is trained, and the association between historical orders and inventories is analyzed by the trained linear regression model to obtain the association characteristics between orders and inventories of the target product in different regions at different time periods.

[0048] It should be noted that the linear regression model is used to fit the order volume O i，t，r and inventory I i，t，rThe relationship between can easily quantify the impact of inventory on order volume and analyze the correlation between inventory and demand. The linear regression model can obtain the regression coefficient β1, which represents the intensity of the impact of inventory on order volume. If β1 is positive, it indicates that an increase in inventory may promote an increase in order volume; if β1 is negative, it indicates that an increase in inventory may not bring about an increase in order volume, and may even lead to inventory backlogs. By quantitatively analyzing the impact of inventory on demand, the system can optimize inventory strategies accordingly. The error term ∈ represents the part that the linear regression model cannot explain. It can capture the fluctuations between orders and inventory due to randomness or other external factors. For example, promotional activities, holiday demand fluctuations, etc., these factors may affect the order volume, but cannot be directly explained from inventory data. By introducing the error term, the linear regression model can more realistically reflect the actual order situation and make the analysis results more representative.

[0049] In some preferred embodiments, training the linear regression model includes: using mean square error as a loss function to minimize the difference between the predicted value and the actual order quantity, and optimizing the regression coefficients β0 and β1 using a gradient descent method to minimize the loss function; the loss function is:

[0050]

[0051] Wherein, n represents the total number of samples for training the linear regression model.

[0052] It should be noted that using mean square error (MSE) as the loss function, the prediction errors are squared, accumulated and averaged, so that the impact of larger errors can be amplified. This means that the linear regression model will pay more attention to samples with larger prediction errors during the training process, prompting the model to further optimize the fitting of large error samples, avoid bias for a few abnormal samples, and improve the stability and robustness of the linear regression model. In addition, MSE averages the errors of all samples, so that the linear regression model considers the errors of all samples when adjusting parameters, and will not overly emphasize the impact of a specific sample. In this way, the linear regression model can maintain overall balance during the parameter adjustment process, avoid overfitting to specific samples, and thus improve the generalization ability of the linear regression model on new data.

[0053] In some preferred embodiments, when training the intelligent inventory transfer model through the association features of the order and inventory, it includes:

[0054] Define the related features of the orders and inventories of the target product in different time periods and regions as relatedfeatures, and define the input feature vector X t，r ;

[0055] X t，r =[Oi，t，r , I i，t，r , related features];

[0056] Construct a multi-layer perceptron model f and input the feature vector X through the input layer of the multi-layer perceptron model f t，r Train the multi-layer perceptron model f, using the actual required allocation quantity D of the target product i in region r within time t during training i，t，r As label supervision training, the output layer of the multilayer perceptron model f outputs the predicted allocation volume of the target product i in region r within time t predicted by the multilayer perceptron model f Multilayer Perceptron Model f and Predicted Allocation Quantity satisfy:

[0057] θ is the parameter of the multilayer perceptron model f.

[0058] It should be noted that, in this embodiment, an input feature vector is defined, and the order quantity, inventory quantity and related associated features are combined into an input vector to capture the demand and inventory situation of the target product in different time periods and different regions. These associated features (such as seasonality, promotional activities, etc.) can reflect complex demand patterns, so that the multi-layer perceptron model can better understand the relationship between orders and inventories, learn more comprehensive information, and thus improve the accuracy of the allocation quantity prediction. In addition, the multi-layer perceptron (MLP) model is a neural network structure that can fit complex nonlinear relationships. Unlike linear regression, the MLP model can capture more complex relationships between inputs and outputs through the connection of multiple layers of neurons and nonlinear activation functions. This feature is particularly suitable for inventory allocation problems, because the relationship between order quantity and inventory quantity is often not linear, but is affected by multiple factors. The MLP model can identify these complex patterns and provide stronger support for allocation decisions.

[0059] In some further preferred embodiments, the current order of the target product and the current inventory status of the target product in each regional warehouse are analyzed by using the trained intelligent inventory transfer model to generate an inventory transfer plan for the target product in each regional warehouse, including: obtaining the current order of the target product and the current inventory status of the target product in each regional warehouse in real time; analyzing the real-time current order and the current inventory status obtained in real time by using the trained intelligent inventory transfer model to generate an inventory transfer plan for the target product in each regional warehouse, thereby achieving a more flexible response to actual demand fluctuations, generating a dynamic transfer plan, reducing transfer delays, and ensuring that inventory always matches actual demand.

[0060] In some further preferred embodiments, the generated inventory transfer plan for the target product in each regional warehouse includes a safety inventory replenishment plan, a demand forecast replenishment plan, an emergency transfer plan, a cost optimization transfer plan, an inventory liquidation plan, and a conventional replenishment transfer plan with decreasing priorities; when the inventory of the target product in each regional warehouse is transferred, the inventory transfer plan with a high priority is executed first. It should be noted that by sorting the transfer plans by priority, it is ensured that the demand that has the greatest impact on business operations is met first. Plans with high priority (such as safety inventory replenishment, demand forecast replenishment, and emergency transfer) are closely related to order fulfillment rate and customer satisfaction. For example, suppose that the demand for smartphones in a certain area surges, causing the inventory to fall below the safety inventory line. At this time, the safety inventory replenishment plan is executed first, and the inventory is transferred from other regional warehouses in time to ensure that the safety inventory line is maintained, so as to avoid customer orders from being unable to be met due to out-of-stock. Under the premise that the safety inventory is guaranteed, responding to future predicted demand fluctuations in advance helps prevent future out-of-stock or insufficient inventory. The demand forecast plan can allocate inventory in advance according to sales forecasts to avoid future supply pressures and make the transfer process more orderly. For example, during the upcoming "Double Eleven" promotion, it is predicted that the demand for tablets in a certain area will increase significantly. At this time, the demand forecast replenishment plan is prioritized to allocate sufficient inventory to the area before the peak of the promotion to ensure that there is sufficient inventory during the event to meet market demand. In special circumstances (such as sudden order surges or unexpected inventory shortages), the emergency allocation plan can flexibly and quickly respond to demand and prevent supply chain disruptions. Emergency allocation has a lower priority than safety stock and demand forecasts to ensure that inventory is dynamically responded to sudden market fluctuations. For example, the inventory of smart watches in a certain area is rapidly consumed due to unexpected market demand growth. At this time, the emergency allocation plan is triggered to quickly allocate inventory from neighboring areas to ensure that the area can meet the demand of sudden orders and avoid order backlogs or customer loss. When inventory and demand are relatively stable, the transportation cost during the allocation process can be reduced through the cost optimization plan. Although it has a lower priority, after meeting basic inventory needs, the cost optimization plan can help companies save transportation costs and maximize cost-effectiveness. For example, when demand is relatively stable, it is planned to transfer excess inventory in certain areas to another area, execute a cost-optimized transfer plan, select the transfer path with the lowest transportation cost, and reduce replenishment costs by merging batches, so that the transfer operation can reduce costs while meeting demand. The inventory reduction plan has a lower priority, but when inventory is overstocked in some areas, inventory can be reduced through transfer. Inventory reduction helps improve inventory turnover, reduce long-term storage costs, and reduce the risk of inventory aging or loss. For example, there is an excess of headphone inventory in one area, while other areas are understocked. After meeting high-priority demand, execute the inventory reduction plan to transfer the excess inventory in the area to the out-of-stock area, reduce the backlog, and improve the overall inventory turnover efficiency.The purpose of the regular replenishment plan is to maintain the basic inventory level in each region to ensure smooth operation in the absence of special needs or emergencies. It has the lowest priority because it does not urgently affect the order fulfillment rate and is only executed when there are no other high-priority plans. For example, in daily inventory management, the inventory of Bluetooth speakers in a certain area gradually decreases to the predetermined replenishment point. At this time, after meeting the high-priority plan, the regular replenishment transfer plan is executed to transfer appropriate amounts of inventory from other warehouses to maintain the basic inventory level in the area.

[0061] Embodiment 2

[0062] See also Figure 1-Figure 2 This embodiment provides an intelligent replenishment management system based on artificial intelligence, which runs on a server. The intelligent replenishment management system based on artificial intelligence adopts the intelligent replenishment management method based on artificial intelligence described in any of the above embodiments, calculates the historical order data generated by the target product on the e-commerce platform, and calculates the historical inventory data of the target product in each regional warehouse, analyzes the historical order data and the historical inventory data through a machine learning model to obtain the correlation characteristics of the orders and inventories of the target product in different regions in different time periods, trains an intelligent inventory transfer model through the correlation characteristics of the orders and inventories, so that the intelligent inventory transfer model learns to automatically adjust the inventory transfer to adapt to the regional replenishment needs of the target product, and uses the trained intelligent inventory transfer model to analyze the current order of the target product and the current inventory status of the target product in each regional warehouse to generate the inventory transfer plan of the target product in each regional warehouse, so as to realize replenishment management to adapt to demand changes and quickly respond to regional needs. Among them, the historical order data can come from the e-commerce platform, and the historical inventory data can come from the warehouse management system.

[0063] It should be pointed out that the above embodiments are only preferred specific implementation modes of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. The protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A management method for intelligent replenishment based on artificial intelligence, characterized in that: include: Calculate the historical order data of the target product on the e-commerce platform, and calculate the historical inventory data of the target product in each regional warehouse; Analyze the historical order data and the historical inventory data through a machine learning model to obtain correlation features between orders and inventories of the target product in different regions at different time periods; Training an intelligent inventory transfer model through the correlation features of the order and inventory, so that the intelligent inventory transfer model learns to automatically adjust inventory transfer to adapt to the regional replenishment demand of the target product; The trained intelligent inventory transfer model is used to analyze the current order of the target product and the current inventory status of the target product in each regional warehouse to generate an inventory transfer plan for the target product in each regional warehouse.

2. The management method of intelligent replenishment based on artificial intelligence according to claim 1, characterized in that: Also includes: Before calculating the historical order data, a target product screening formula is set to screen the target product, and the target product screening formula is f(Product); f(Product)=Category∩Brand∩Price_Range∩Sales_Range; In the target product screening formula, Category is the category label of the target product, such as electronic products, clothing, etc.; Brand is the brand label of the target product; Price_Range is the price range of the target product; Sales_Range is the sales range of the target product, which is used to screen products that meet the sales standards; when f(Product)=1, the target product is screened.

3. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 2, characterized in that: When calculating the historical order data and the historical inventory data, a time window is selected according to demand to analyze the order data and inventory data of the target product in different time periods, and the time window is defined as T; T={t1,t2,...,t n }; T contains each time period, t n Indicates each time period.

4. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 3, characterized in that: When calculating the historical order data, the order quantity is calculated using the calculation formula, which is: In the calculation formula of the order quantity, O i，t，r represents the order volume of target product i in region r within time t; N t，r N t represents the total number of orders in region r at time t; S i，j Indicates whether the target product i is included in order j. If so, S i，j =1, otherwise 0.

5. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 4, characterized in that: When calculating the historical inventory data, the inventory calculation formula is used for calculation, and the inventory calculation formula is: In the calculation formula of the inventory, I i，t，r represents the inventory of target product i in the warehouse of region r within time t, L t，r represents the total number of inventory records in region r at time t, S i，k represents the inventory quantity of target product i in inventory record k.

6. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 5, characterized in that: The historical inventory data includes regional in-and-out records, which are calculated using a net inventory change calculation formula: In the calculation formula of the net inventory change, ΔI i,t,r represents the net inventory change of target product i in region r within time t, represents the total inventory quantity of target product i in region r within time t, Represents the total shipment quantity of target product i in region r within time t.

7. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 5, characterized in that: When analyzing the historical order data and the historical inventory data by a machine learning model, it includes: using a linear regression model to fit the order quantity. i，t，r and inventory I i，t，r The relationship between them is specifically: Oh i，t，r =β0+β1·I i，t，r +∈; β0 and β1 are regression coefficients, indicating the strength of the impact of inventory on orders; ∈ is the error term, which is used to represent the part that the linear regression model cannot explain; The linear regression model is trained, and the association between historical orders and inventories is analyzed by the trained linear regression model to obtain the association characteristics between orders and inventories of the target product in different regions at different time periods.

8. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 7, characterized in that: Training the linear regression model includes: using mean square error as a loss function to minimize the difference between the predicted value and the actual order quantity, and using the gradient descent method to optimize the regression coefficients β0 and β1 to minimize the loss function; the loss function is: Wherein, n represents the total number of samples for training the linear regression model.

9. The management method of intelligent replenishment based on artificial intelligence as claimed in claim 5, characterized in that: When training the intelligent inventory transfer model through the association features of the order and inventory, it includes: Define the related features of the target product's orders and inventories in different regions at different time periods as relatedfeatures, and define the input feature vector X t，r ; X t，r =[O i，t，r ,I i，t，r ,related features]; Construct a multi-layer perceptron model f and input the feature vector X through the input layer of the multi-layer perceptron model f t，r Train the multi-layer perceptron model f, using the actual required allocation quantity D of the target product i in region r within time t during training i，t，r As label supervision training, the output layer of the multilayer perceptron model f outputs the predicted allocation volume of the target product i in region r within time t predicted by the multilayer perceptron model f Multilayer Perceptron Model f and Predicted Allocation Quantity satisfy: θ is the parameter of the multilayer perceptron model f.

10. An artificial intelligence-based intelligent replenishment management system, characterized in that: The artificial intelligence-based smart replenishment management system applies the artificial intelligence-based smart replenishment management method as described in any one of claims 1-9.

Images