Model training method and apparatus, inventory allocation method and apparatus, electronic device, storage medium and program product

Abstract

Disclosed in the embodiments of the present application are a model training method and apparatus, an inventory allocation method and apparatus, an electronic device, a storage medium and a program product. The model training method comprises: acquiring a first sample quantity demanded of a target item in a first warehouse center and a second sample quantity demanded thereof in a second warehouse center within a sample historical time period, and acquiring an inventory estimation module and an allocation simulation module; using the inventory estimation module to obtain a first estimated inventory of the target item at a simulation historical time point on the basis of the first sample quantity demanded, and to obtain a second estimated inventory of the target item at the simulation historical time point on the basis of the second sample quantity demanded; and using the allocation simulation module to perform allocation simulation of the first warehouse center and the second warehouse center on the basis of the first estimated inventory and the second estimated inventory, so as to adjust parameters in the inventory estimation module on the basis of an allocation simulation result to obtain a target inventory estimation module. The technical solution of the embodiments of the present application can implement end-to-end inventory allocation decision-making.

Description

A model training, inventory transfer method, apparatus, electronic device, storage medium, and program product.

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese Patent Application No. 202411554148.X, filed on November 1, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to the field of warehousing and logistics technology, and in particular to a model training, inventory transfer method, apparatus, electronic device, storage medium, and program product. Background Technology

[0004] Inventory management is the cornerstone of the entire supply chain. Excellent inventory management can lead to cost reduction and efficiency improvement. Especially when facing a massive amount of goods, how to arrange inventory layout, allocate inventory more rationally, and place items closer to consumers in advance to meet the growing demand of consumers for fast delivery is a common challenge faced by the current retail e-commerce industry.

[0005] For inventory allocation decisions, the current industry technology involves breaking down the inventory allocation system into multiple modules and determining the final inventory allocation decision through the sequential decision-making of each module.

[0006] In the process of realizing this disclosure, the inventors discovered the following technical problems in the prior art: it cannot bring about globally optimal decisions, the system is highly complex and has high maintenance costs, and individual modules cannot be quickly optimized. Summary of the Invention

[0007] This disclosure provides a model training, inventory allocation method, apparatus, electronic device, storage medium, and program product to achieve end-to-end inventory allocation decision-making.

[0008] According to one aspect of this disclosure, a model training method is provided, which may include:

[0009] The system obtains the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and obtains the model to be trained, wherein the model to be trained includes an inventory prediction module and an allocation simulation module.

[0010] Using the inventory forecasting module, based on the first sample demand, the first estimated inventory of the target item at the simulated historical time point is obtained, and based on the second sample demand, the second estimated inventory of the target item at the simulated historical time point is obtained.

[0011] Using the allocation simulation module, based on the first estimated inventory and the second estimated inventory, an allocation simulation is performed between the first and second warehousing centers. The parameters in the inventory estimation module are adjusted according to the allocation simulation results to obtain the target inventory estimation module.

[0012] According to another aspect of this disclosure, an inventory transfer method is provided, which may include:

[0013] The system obtains the first target demand quantity of the target item in the first warehouse center and the second target demand quantity in the second warehouse center within the target historical time period, as well as a target inventory estimation module trained according to the model training method described in any embodiment of this disclosure.

[0014] Using the target inventory forecasting module, based on the first target demand, the first target inventory of the target item at the inventory transfer time point is forecasted, and based on the second target demand, the second target inventory of the target item at the inventory transfer time point is forecasted.

[0015] Based on the first target inventory and the second target inventory, determine the inventory transfer decision between the first and second warehousing centers, and transfer inventory between the first and second warehousing centers for the target items according to the inventory transfer decision.

[0016] According to another aspect of this disclosure, a model training apparatus is provided, which may include:

[0017] The first acquisition module is used to acquire the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and to acquire the model to be trained, wherein the model to be trained includes an inventory estimation module and an allocation simulation module.

[0018] The inventory estimation module is used to obtain the first estimated inventory of the target item at the simulation historical time point based on the first sample demand, and to obtain the second estimated inventory of the target item at the simulation historical time point based on the second sample demand.

[0019] The model training module is used to simulate the transfer between the first and second storage centers based on the first and second estimated inventory using the transfer simulation module. The parameters in the inventory estimation module are adjusted according to the transfer simulation results to obtain the target inventory estimation module.

[0020] According to another aspect of this disclosure, an inventory transfer device is provided, which may include:

[0021] The second acquisition module is used to acquire the first target demand quantity of the target item in the first storage center and the second target demand quantity in the second storage center during the target historical time period, as well as the target inventory estimation module trained according to the model training method described in any embodiment of this disclosure.

[0022] The target inventory module is used to estimate the first target inventory of the target item at the inventory transfer time point based on the first target demand, and to estimate the second target inventory of the target item at the inventory transfer time point based on the second target demand, using the target inventory estimation module.

[0023] The inventory transfer module is used to determine the inventory transfer decision between the first and second warehousing centers based on the first target inventory and the second target inventory, so as to transfer the inventory between the first and second warehousing centers for the target items according to the inventory transfer decision.

[0024] According to another aspect of this disclosure, an electronic device is provided that may include:

[0025] At least one processor; and

[0026] A memory that is communicatively connected to at least one processor; wherein,

[0027] The memory stores a computer program that can be executed by at least one processor, such that when the at least one processor executes the program, it implements the model training method or inventory transfer method provided in any embodiment of this disclosure.

[0028] According to another aspect of this disclosure, a computer-readable storage medium is provided that stores computer instructions thereon, which are used to cause a processor to execute and implement the model training method or inventory transfer method provided in any embodiment of this disclosure.

[0029] According to another aspect of this disclosure, a computer program product is provided, on which a computer program is stored, which, when executed by a processor, implements the model training method or inventory allocation method provided in any embodiment of this disclosure.

[0030] The technical solution of this disclosure involves obtaining the first sample demand quantity of the target item in the RDC and the second sample demand quantity in the FDC during a sample historical time period. Demand quantity is an important factor in estimating the required inventory. A training model is also obtained, comprising an inventory estimation module and a transfer simulation module, for model training. Using the inventory estimation module, based on the first sample demand quantity, a first estimated inventory of the target item at the simulated historical time point is obtained, and based on the second sample demand quantity, a second estimated inventory of the target item at the simulated historical time point is obtained, thus estimating the required inventory of the target item in the RDC and FDC. Then, using the transfer simulation module, based on the first and second estimated inventories, a transfer simulation is performed between the RDC and FDC for the target item. Based on the obtained transfer simulation results, the parameters in the inventory estimation module can be adjusted to obtain the target inventory estimation module. The above technical solution solves the problems caused by module splitting, such as the inability to make globally optimal decisions, high system complexity and maintenance costs, and the inability to quickly optimize individual modules, by training an end-to-end target inventory estimation module that can be used for inventory allocation decisions.

[0031] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 is a flowchart of a model training method provided according to an embodiment of the present disclosure;

[0034] Figure 2 is a flowchart of another model training method provided according to an embodiment of the present disclosure;

[0035] Figure 3 is a flowchart of another model training method provided according to an embodiment of the present disclosure;

[0036] Figure 4 is a schematic diagram of a model in another model training method provided according to an embodiment of the present disclosure;

[0037] Figure 5 is a flowchart of an inventory transfer method according to an embodiment of the present disclosure;

[0038] Figure 6 is a structural block diagram of a model training device provided according to an embodiment of the present disclosure;

[0039] Figure 7 is a structural block diagram of an inventory transfer device according to an embodiment of the present disclosure;

[0040] Figure 8 is a schematic diagram of the structure of an electronic device that implements the model training method or inventory transfer method of the present disclosure. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present disclosure, the technical solutions of the present disclosure will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present disclosure, and not all embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present disclosure.

[0042] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The same applies to "target," "original," etc., and will not be repeated here. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0043] Before introducing the embodiments of this disclosure, the application scenarios of the embodiments of this disclosure will be described by way of example. For example, the currently applied inventory transfer system has problems such as the inability to achieve the globally optimal decision due to each module making decisions independently, the high system complexity and high maintenance cost due to the decision-making of multiple modules, and the inability of a single module to be quickly optimized because the optimization of a single module requires the cooperation of other modules.

[0044] In response to this, this disclosure proposes a technical solution for making inventory allocation decisions in an end-to-end manner, which eliminates the need for module splitting and thus solves a series of problems caused by module splitting.

[0045] Figure 1 is a flowchart of a model training method provided in an embodiment of this disclosure. This embodiment is applicable to training a target inventory forecasting module capable of achieving end-to-end inventory allocation decisions. The method can be executed by the model training apparatus provided in this embodiment, which can be implemented in software and / or hardware and can be integrated into an electronic device, such as various user terminals or servers.

[0046] Referring to Figure 1, the method of this embodiment specifically includes the following steps:

[0047] S110. Obtain the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and obtain the model to be trained.

[0048] The models to be trained include an inventory estimation module and an allocation simulation module.

[0049] The target items can be understood as items that need to be transferred between the first and second warehousing centers. More specifically, they can be understood as items that need to be transferred from the first warehousing center to the second warehousing center, or vice versa. The specific direction of the inventory transfer depends on the actual situation and is not specifically limited here.

[0050] The aforementioned first and second warehousing centers are different warehousing centers used at least for storing target items. Based on this, and in conjunction with the application scenarios that may be involved in the embodiments of this disclosure, the first and second warehousing centers may optionally be any two of the following warehousing centers: Central Distribution Center (CDC), Regional Distribution Center (RDC), and Front Distribution Center (FDC). Of course, they may also be other warehousing centers, which can be set according to actual needs and are not specifically limited here.

[0051] Based on this, in order to more vividly understand the model training process described in the embodiments of this disclosure, the following will take the example of the first warehouse center being RDC and the second warehouse center being FDC. However, it should be emphasized that this is just an example and not a specific limitation on these two warehouse centers.

[0052] The historical time period of the sample can be understood as the time period that has already occurred and is used in the model training process.

[0053] The first sample demand can be understood as time-series data representing the demand for the target item in the RDC during the sample's historical time period. Optionally, in conjunction with the application scenarios that may be involved in the embodiments of this disclosure, the demand can be understood as sales volume.

[0054] Similarly, the second sample demand can be understood as time-series data representing the demand for the target item in the FDC within the sample's historical time period.

[0055] The model to be trained can be understood as a model with training requirements. This model may include an inventory forecasting module and a transfer simulation module. The inventory forecasting module can be understood as a module used for inventory forecasting, specifically for forecasting required inventory, which is related to demand. Optionally, this module may include a demand forecasting unit and an inventory forecasting unit. The transfer simulation module is connected to the inventory forecasting module and is used to simulate the inventory transfer process of the target item between the RDC and FDC based on the inventory forecasting results output by the inventory forecasting module. In practical applications, the model may also optionally include a loss calculation module connected to the transfer simulation module to calculate losses based on the transfer simulation results output by the transfer simulation module; etc., without specific limitations here.

[0056] Obtain the first sample requirement, the second sample requirement, and the model to be trained. Considering the application scenarios that may be involved in the embodiments of this disclosure, optionally, the number of RDCs is usually one, while the number of FDCs can be one or more. In this step, the second sample requirement corresponding to each FDC can be obtained.

[0057] S120. Using the inventory estimation module, based on the first sample demand, obtain the first estimated inventory of the target item at the simulation historical time point, and based on the second sample demand, obtain the second estimated inventory of the target item at the simulation historical time point.

[0058] Here, the simulation history time point can be understood as a point in time that has occurred during the simulation process. In this embodiment, optionally, the time point may be a certain time point within the sample's historical time period, such as a relatively late time point within the sample's historical time period; it may also be a time point later than all time points within the sample's historical time period; etc., without specific limitation. Further optionally, the number of time points can be one or more; if there are multiple time points, each time point can be processed separately.

[0059] The first estimated inventory can be understood as the inventory required by the target item in the RDC at a simulated historical time point. The inventory is estimated using the inventory estimation module based on the demand of the first sample.

[0060] The same logic applies to the second estimated inventory, so I won't go into details here.

[0061] S130. Using the transfer simulation module, based on the first estimated inventory and the second estimated inventory, perform a transfer simulation between the first storage center and the second storage center, so as to adjust the parameters in the inventory estimation module according to the transfer simulation results, and obtain the target inventory estimation module.

[0062] Specifically, a transfer simulation module is used to simulate inventory transfer between the RDC and FDC for the target item, based on a first estimated inventory and a second estimated inventory. Furthermore, considering the application scenarios that may be involved in the embodiments of this disclosure, before inventory transfer, taking into account that the target item may be stored in the RDC and / or FDC, to avoid invalid transfers, inventory transfer simulation can be performed based on the first estimated inventory and the second estimated inventory, combined with a first initial inventory and the second initial inventory. The first initial inventory can be understood as the quantity of the target item stored in the RDC before the inventory transfer, and the second initial inventory can be understood as the quantity of the target item stored in the FDC before the inventory transfer.

[0063] Furthermore, based on the allocation simulation results output by the allocation simulation module, the parameters in the inventory estimation module of the model to be trained can be adjusted to obtain the target inventory estimation module. Subsequently, the trained target inventory estimation module can be used for inventory estimation, and inventory allocation decisions can be made based on this.

[0064] In practical applications, optionally, corresponding target inventory prediction modules can be trained for different target items or different stock keeping units (SKUs); alternatively, a single target inventory prediction module can be trained for at least some SKUs; etc., without specific limitations. For the latter, during both the model training and inference phases, item information representing the SKU can be input into the model to help the model determine the SKU currently being processed. This item information can be, for example, the item name and / or the item brand.

[0065] The technical solution of this disclosure involves obtaining the first sample demand quantity of the target item in the RDC and the second sample demand quantity in the FDC during a sample historical time period. Demand quantity is an important factor in estimating the required inventory. A training model is also obtained, comprising an inventory estimation module and a transfer simulation module, for model training. Using the inventory estimation module, based on the first sample demand quantity, a first estimated inventory of the target item at the simulated historical time point is obtained, and based on the second sample demand quantity, a second estimated inventory of the target item at the simulated historical time point is obtained, thus estimating the required inventory of the target item in the RDC and FDC. Then, using the transfer simulation module, based on the first and second estimated inventories, a transfer simulation is performed between the RDC and FDC for the target item. Based on the obtained transfer simulation results, the parameters in the inventory estimation module can be adjusted to obtain the target inventory estimation module. The above technical solution solves the problems caused by module splitting, such as the inability to make globally optimal decisions, high system complexity and maintenance costs, and the inability to quickly optimize individual modules, by training an end-to-end target inventory estimation module that can be used for inventory allocation decisions.

[0066] Figure 2 is a flowchart of another model training method provided in this embodiment. This embodiment is based on and optimized from the above-described technical solutions. In this embodiment, optionally, the inventory estimation module includes a demand forecasting unit and an inventory estimation unit; using the inventory estimation module, based on the first sample demand, to obtain the first estimated inventory of the target item at the simulated historical time point includes: inputting the first sample demand into the demand forecasting unit to obtain the first predicted demand of the target item at the associated simulated historical time point; inputting the first predicted demand into the inventory estimation unit to obtain the first estimated inventory of the target item at the simulated historical time point. The explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here.

[0067] Referring to Figure 2, the method in this embodiment may specifically include the following steps:

[0068] S210. Obtain the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and obtain the model to be trained;

[0069] The model to be trained includes an inventory forecasting module and an allocation simulation module. The inventory forecasting module includes a demand forecasting unit and an inventory forecasting unit.

[0070] As mentioned above, demand is a crucial factor in inventory forecasting. Therefore, inventory forecasting is achieved through a demand forecasting unit and an inventory forecasting unit.

[0071] S220. Input the first sample demand into the demand prediction unit to obtain the first predicted demand of the target item related to the simulation historical time point, and input the second sample demand into the demand prediction unit to obtain the second predicted demand of the target item related to the simulation historical time point.

[0072] The first predicted demand can be understood as the predicted demand for the target item in the RDC. Based on this, and considering the application scenarios that may be involved in the embodiments of this disclosure, this demand may be the demand for the target item in the RDC at a simulated historical time point, or it may be the demand in the RDC during a time period including the simulated historical time point. This can be set according to the actual situation and is not specifically limited here. The first sample demand is input into the demand prediction unit to obtain the first predicted demand.

[0073] The second forecast demand is similar to the first forecast demand, and no specific limitations are specified here.

[0074] S230. Input the first predicted demand into the inventory estimation unit to obtain the first estimated inventory of the target item at the simulated historical time point, and input the second predicted demand into the inventory estimation unit to obtain the second estimated inventory of the target item at the simulated historical time point.

[0075] Specifically, the inventory estimation unit processes the demand forecast results (i.e., the first forecast demand or the second forecast demand) output by the demand forecast unit to obtain the estimated inventory (i.e., the first estimated inventory corresponding to the first forecast demand or the second estimated inventory corresponding to the second forecast demand).

[0076] S240. Using the transfer simulation module, based on the first estimated inventory and the second estimated inventory, perform a transfer simulation between the first storage center and the second storage center, so as to adjust the parameters in the inventory estimation module according to the transfer simulation results, and obtain the target inventory estimation module.

[0077] The technical solution of this disclosure divides inventory forecasting into two parts: demand forecasting and inventory forecasting. The inventory forecasting is performed using the demand forecasting results, and the resulting forecasted inventory is interpretable, thereby ensuring the accuracy of inventory forecasting.

[0078] Based on this, and considering the application scenarios that may be involved in the embodiments of this disclosure, taking sales volume as an example, the relevant content of sales volume forecasting is described below.

[0079] In the sales forecasting process, the input data may include historical sales at the SKU level, historical sales at the third-level category level, historical sales at the brand level, and historical sales of SKUs at the national level. In addition, it may also include promotional and marketing activity factors that affect sales and / or item information that represents SKUs, etc., without specific limitations.

[0080] Furthermore, after obtaining these input data, preprocessing tasks such as outlier cleanup and feature engineering can be performed. For example, statistical features can be obtained through one or more aggregation methods, which may include at least one of max, min, avg, std, sum, and count; one-hot transformation can be performed on the provided data frame and specified columns, supporting simultaneous transformation of multiple columns; missing values ​​can be filled using methods such as mean, median, or random sampling; and normalization can be performed, such as transforming the data to a specified range to change the overall mean and standard deviation, normal standardization, and robust standardization (e.g., data standardization based on quartiles). These can be selected according to actual needs and are not specifically limited here.

[0081] Furthermore, the demand forecasting unit (i.e., the sales forecasting unit) can operate through the following two steps:

[0082] Step 1: Information Encoding, which may include the following parts: recent sales encoder, influencing activity encoder, and concurrent sales encoder. In practical applications, these encoders can optionally adopt a transformer structure, which consists of multiple stacked multi-head attention mechanisms. Each multi-head attention mechanism consists of three linear layers stacked with multiple attention mechanisms. The entire structure increases the depth of the network by stacking network structures, giving the network a stronger learning and representation ability.

[0083] Specifically, the recent sales encoder can input recent sales, time information, and item information into the transformer structure, encode them, and output encoded data of the predicted length.

[0084] The active encoder can input price data into the transformer structure, encode it, and output encoded data of the predicted length.

[0085] The same-period sales encoder can input historical same-period sales into the transformer structure, encode them, and output encoded data of the predicted length.

[0086] Step 2: Merge and decode the information.

[0087] Specifically, the encoded data output from the three encoders are merged along the same dimension, and then processed through a multi-layer linear network with a ReLU activation function. Sales predictions at multiple quantiles are output via an accumulation method, ensuring that the prediction for each high quantile is greater than or equal to the prediction for a relatively low quantile. For the accumulation method, for example, q_1 represents the lowest quantile, then q_2 quantile = q_1 quantile + X, where X represents a number greater than or equal to 0, which can be learned through the network.

[0088] Based on this, the sales forecasting loss function can be expressed as follows: Where, q i Let q represent the i-th quantile. i and q i Similarly, y n This represents the historical sales volume of the nth item. This represents the predicted sales volume of the nth item.

[0089] Regarding inventory forecasting, and considering the application scenarios that may be involved in the embodiments of this disclosure, optionally, the inventory forecasting unit includes a basic inventory forecasting subunit, a fluctuating inventory forecasting subunit, and an inventory forecasting subunit.

[0090] Inputting the first predicted demand into the inventory estimation unit yields the first estimated inventory of the target item at the simulated historical time point, including:

[0091] Input the first forecasted demand into the basic inventory estimation sub-unit to obtain the estimated basic inventory;

[0092] Input the first predicted demand, the first actual demand corresponding to the first predicted demand in the first sample demand, and the prediction deviation obtained based on the first predicted demand and the first actual demand into the fluctuation inventory estimation sub-unit to obtain the estimated fluctuation inventory.

[0093] Input the estimated base inventory and the estimated fluctuating inventory into the inventory estimation sub-unit to obtain the first estimated inventory of the target item at the simulated historical time point.

[0094] Here, we take RDC as an example. The related processing of FDC is similar and will not be repeated here.

[0095] The first predicted demand is input into the basic inventory estimation sub-unit to obtain the estimated basic inventory of the target item in the RDC. In combination with the application scenarios that may be involved in the embodiments of this disclosure, the estimated basic inventory may be the basic inventory at the simulated historical time point; or it may be the basic inventory within the time period including the simulated historical time point, such as the basic inventory within the allocation leading time (alt); etc., without specific limitations.

[0096] The first sample demand includes multiple first actual demand quantities (i.e., actual demand quantities occurring within the RDC) within the historical time period of the sample. In this technical solution, the first actual demand quantity can be understood as the first actual demand quantity corresponding to the first predicted demand quantity among these multiple first actual demand quantities. For example, considering that during the model inference stage, the first predicted demand quantity is the demand quantity at a future point in time, and the first actual demand quantity at that future point in time does not exist, in order to ensure the consistency between model training and model inference, the first actual demand quantity used in this technical solution can be the first actual demand quantity at a point in time one prediction period prior to the time point corresponding to the first predicted demand quantity. The prediction deviation can be understood as the difference between the first predicted demand quantity and the first actual demand quantity. Inputting the first predicted demand quantity, the first actual demand quantity, and the prediction deviation into the fluctuation inventory estimation sub-unit yields the estimated fluctuation inventory.

[0097] Furthermore, the estimated basic inventory and estimated fluctuating inventory are input into the inventory estimation sub-unit to obtain the first estimated inventory of the target item at the simulated historical time point.

[0098] The above technical solution takes into account the volatility of inventory. Therefore, by combining the estimated basic inventory with the estimated fluctuating inventory to calculate the first estimated inventory, the accuracy of the first estimated inventory is improved.

[0099] Based on this, optionally, for the allocation lead time corresponding to the historical time point in the simulation, the estimated basic inventory is the estimated basic inventory within the allocation lead time, and the estimated fluctuation inventory is the estimated fluctuation inventory within the allocation lead time.

[0100] The inventory forecasting unit also includes a masked inventory forecasting subunit;

[0101] Inputting the first predicted demand into the inventory estimation unit yields the first estimated inventory of the target item at a simulated historical time point, which also includes:

[0102] Using the masked inventory estimation sub-unit, the masked demand during the allocation lead time is masked from the first predicted demand, and the masked demand is summed to obtain the estimated masked inventory.

[0103] Accordingly, the estimated base inventory and the estimated fluctuating inventory are input into the inventory estimation sub-unit to obtain the first estimated inventory of the target item at the simulated historical time point, including:

[0104] Input the estimated base inventory, estimated fluctuation inventory, and estimated mask inventory into the inventory estimation sub-unit to obtain the first estimated inventory of the target item at the simulated historical time point.

[0105] Specifically, for the alt value corresponding to the historical time point in the simulation, the estimated basic inventory can be the estimated basic inventory within alt, and the estimated fluctuation inventory can be the estimated fluctuation inventory within alt.

[0106] Based on this, the masked inventory estimation sub-unit masks the first actual demand located within the alt value from multiple first actual demand values ​​represented by the first predicted demand. For ease of distinction, this first actual demand value is referred to as the masked demand value. Then, all the obtained masked demand values ​​are summed to obtain the estimated masked inventory. In this way, the first estimated inventory can be obtained by combining the estimated base inventory, the estimated fluctuation inventory, and the estimated masked inventory.

[0107] The above technical solution further improves the accuracy of the first estimated inventory by introducing alt and combining it with mask summation for inventory estimation.

[0108] Based on this, and in conjunction with the application scenarios that may be involved in the embodiments of this disclosure, for example, inventory can be divided into two types: safety stock Ss and target stock Ti. Ss of a distribution center (DC) (such as RDC or FDC) can be understood as the inventory required to ensure that the target items shipped from the DC on the same day are available in the DC. Ti can be understood as the inventory required to ensure that the target items shipped from the DC within a certain transfer frequency are available in the DC. By setting Ti, the number of inventory transfers can be reduced.

[0109] Based on this, and still using sales volume as an example, inventory forecasting can be performed in the following ways:

[0110] 1) Sales forecasting can be achieved by generating the estimated base inventory within alt and the estimated base inventory f within alt+7 using a Multilayer Perceptron (MLP). mlp (alt+7), where 7 refers to 7 days, meaning the default allocation frequency is weekly. 2) Based on predicted sales, actual sales, and prediction deviation, the transformer structure learns the demand fluctuations within alt and generates the estimated fluctuation inventory f. diff (α). 3) Sum the predicted sales for each day within alt to obtain the estimated masked inventory sum(alt). Therefore, Ss = f mlp (alt)+f diff (α)+sum(alt), Ti=f mlp (alt+7)+Ss, complete the inventory forecast.

[0111] Figure 3 is a flowchart of another model training method provided in this embodiment. This embodiment is based on the above-mentioned technical solutions and optimized. In this embodiment, optionally, the model to be trained also includes a loss calculation module connected to the transfer simulation module. The target item is to be transferred from the first storage center to the second storage center. Using the transfer simulation module, based on the first estimated inventory and the second estimated inventory, a transfer simulation is performed between the first storage center and the second storage center to adjust the parameters in the inventory estimation module according to the transfer simulation results, thereby obtaining the target inventory estimation module. This can include: using the transfer simulation module to perform the following three steps to perform the transfer simulation between the first storage center and the second storage center: based on the first estimated inventory and the second estimated inventory... The process involves obtaining the inventory transfer volume from the first storage center to the second storage center; obtaining the first simulated demand volume based on the inventory transfer volume and the first actual demand volume of the target item at the first storage center at the simulated historical time point; obtaining the second simulated demand volume based on the inventory transfer volume and the second actual demand volume of the target item at the second storage center at the simulated historical time point; and using the loss calculation module, calculating the loss based on the first actual demand volume, the first simulated demand volume, the second actual demand volume, and the second simulated demand volume, adjusting the parameters in the inventory forecasting module according to the loss calculation results to obtain the target inventory forecasting module. Explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here.

[0112] Referring to Figure 3, the method in this embodiment may specifically include the following steps:

[0113] S310. Obtain the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and obtain the model to be trained;

[0114] The model to be trained includes an inventory estimation module, a transfer simulation module, and a loss calculation module connected to the transfer simulation module; the target items are to be transferred from the first storage center to the second storage center.

[0115] S320. Using the inventory estimation module, based on the first sample demand, obtain the first estimated inventory of the target item at the simulation historical time point, and based on the second sample demand, obtain the second estimated inventory of the target item at the simulation historical time point.

[0116] S330. Using the transfer simulation module, execute S340-S360 below to perform transfer simulation between the first and second storage centers.

[0117] S340. Based on the first estimated inventory and the second estimated inventory, obtain the inventory transfer amount from the first storage center to the second storage center.

[0118] Taking the transfer of a target item from the first storage center to the second storage center as an example, the first estimated inventory represents the inventory required for the target item at the RDC, and the second estimated inventory represents the inventory required for the target item at the FDC. The inventory transfer amount from the RDC to the FDC can be obtained from these two estimates. It should be noted that when there are multiple FDCs under an RDC, the inventory transfer amount from the RDC to each FDC can be calculated separately.

[0119] Based on this, as explained above, before inventory transfer, considering that the target items may be stored in the RDC and / or FDC, the inventory transfer amount can be determined based on the first estimated inventory and the second estimated inventory, combined with the first initial inventory and the second initial inventory, in order to avoid invalid transfers.

[0120] The inventory transfer amount determined above is the inventory transfer decision. Inventory transfer simulation can be carried out according to this inventory transfer decision to obtain the first demand loss and the second demand loss in the subsequent steps.

[0121] S350. Based on the inventory transfer volume and the first real demand for the target item at the first storage center at the simulated historical time point, obtain the first simulated demand.

[0122] The first actual demand can be understood as the actual demand for the target item in the RDC at the simulated historical time point, i.e., the actual quantity of the target item shipped out. The first simulated demand can be understood as the simulated demand for the target item in the RDC at the simulated historical time point.

[0123] The first simulated demand can be obtained based on the inventory transfer amount and the first actual demand. For example, the first simulated demand can be the minimum value between the inventory remaining in RDC after the inventory transfer and the first actual demand. Optionally, if there is a purchase for the target item, the purchase order quantity should be added to this inventory. Taking demand as sales volume as an example, RDC simulated sales volume = min(RDC inventory + purchase order quantity - inventory transfer amount, RDC actual sales volume), where RDC simulated sales volume is the first simulated demand, and RDC inventory is the first initial inventory.

[0124] Based on this, optionally, at the end of the day, the remaining inventory of RDC can be expressed by the following formula: RDC Inventory = RDC Inventory + Purchase Order Quantity - Inventory Transfer Quantity - Transferred Sales - Actual RDC Sales, where the RDC inventory on the left side of "=" is the remaining inventory, which is also the first initial inventory for the next day, and the RDC inventory on the right side of "=" is the first initial inventory for the day. The transferred sales will be explained later.

[0125] S360. Based on the inventory transfer volume and the second real demand volume of the target item at the second storage center at the simulated historical time point, obtain the second simulated demand volume.

[0126] The second simulation requirement is largely the same as the first simulation requirement.

[0127] For example, FDC simulated sales = min(FDC actual sales, FDC inventory + FDC transferred goods), where FDC simulated sales is the second simulated demand, FDC actual sales is the second actual demand, FDC inventory is the second initial inventory, and FDC transferred goods is the inventory transfer amount.

[0128] Optionally, at the end of the day, the remaining FDC inventory can be represented by the following formula: FDC Inventory = max(FDC Inventory + FDC Transfer Arrivals - FDC Actual Sales, 0). Here, the FDC inventory on the left side of the "=" sign is the remaining inventory, and the FDC inventory on the right side of the "=" sign is the second initial inventory for the day.

[0129] S370. Using the loss calculation module, based on the first actual demand, the first simulated demand, the second actual demand, and the second simulated demand, perform loss calculation to adjust the parameters in the inventory forecasting module according to the loss calculation results, and obtain the target inventory forecasting module.

[0130] The system utilizes a loss calculation module to calculate losses related to four demand quantities. Then, based on the loss calculation results, the parameters in the inventory forecasting module are adjusted to minimize the calculated losses, resulting in the target inventory forecasting module. Optionally, for an inventory forecasting module that includes both a demand forecasting unit and an inventory forecasting unit, the loss calculation module can calculate not only the losses related to demand quantities but also the losses related to forecast deviations. This combined approach improves the accuracy of parameter adjustments. The forecast deviation can be calculated based on the predicted demand and the actual demand at the same historical training point in time.

[0131] The technical solution of this disclosure calculates the simulated demand of RDC and FDC, thereby combining the actual demand to calculate the loss of demand and adjust the parameters accordingly, thereby minimizing the loss of demand and ensuring that all target items that need to be shipped out can be shipped out.

[0132] Based on this, an optional technical solution is provided, wherein the first estimated inventory includes a first estimated safety stock and a first estimated target inventory, and the second estimated inventory may include a second estimated safety stock and a second target inventory. Based on this, the inventory transfer volume from the first storage center to the second storage center is obtained according to the first estimated inventory and the second inventory, including:

[0133] Based on the first estimated safety stock and the second estimated safety stock, the first transfer amount from RDC to FDC is obtained. Inventory transfer simulation is performed based on the first transfer amount to obtain the first simulated inventory of RDC.

[0134] If the first simulated inventory is greater than the first estimated safety stock, the second transfer amount from RDC to FDC can be obtained based on the first simulated inventory, the first estimated target inventory, and the second estimated target inventory.

[0135] The inventory transfer amount is obtained based on the first and second transfer amounts.

[0136] The first allocation amount can be determined based on the first estimated safety stock and the second estimated safety stock. Furthermore, it can be determined by combining the first initial inventory and the second initial inventory. Based on the first allocation amount, inventory is transferred from the RDC to the FDC, simulating the first simulated inventory of the RDC after the inventory transfer. In other words, given a limited initial inventory, inventory is first allocated between the RDC (referring to the RDC itself, excluding the area covered by the RDC) and each FDC based on the safety stock. Typically, the RDC itself has the highest priority, followed by allocation to the FDCs.

[0137] Furthermore, if the first simulated inventory is greater than the first estimated safety stock, this indicates that there is a surplus of inventory in the RDC while ensuring the safety of the local inventory in the RDC. At this time, the second transfer amount from the RDC to the FDC can be obtained based on the first simulated inventory, the first estimated target inventory, and the second estimated target inventory. Compared with directly transferring inventory according to the first transfer amount, the application of this second transfer amount helps to reduce the number of inventory transfers from the RDC to the FDC.

[0138] Furthermore, based on the first and second allocation quantities, the inventory allocation quantity is obtained. For example, the sum of the first and second allocation quantities can be used as the inventory allocation quantity. This inventory allocation quantity can minimize the number of inventory transfers while ensuring inventory safety.

[0139] Another optional technical solution utilizes a loss calculation module to perform loss calculations based on a first actual demand, a first simulated demand, a second actual demand, and a second simulated demand. The parameters in the inventory forecasting module are then adjusted according to the loss calculation results to obtain the target inventory forecasting module, which includes:

[0140] Using the loss calculation module, the first demand loss is obtained based on the first simulated demand and the first actual demand, and the second demand loss is obtained based on the second simulated demand and the second actual demand.

[0141] Based on the first demand loss and the second demand loss, the loss calculation results are obtained. The parameters in the inventory forecasting module are then adjusted according to the loss calculation results to obtain the target inventory forecasting module.

[0142] The first demand loss can be understood as the demand that the RDC should have provided but could not provide after inventory transfers were made according to the inventory allocation plan; that is, the quantity of target items that should have been shipped but were not. The second demand loss has a similar meaning and will not be elaborated here.

[0143] Using the loss calculation module, the second demand loss is obtained based on the second simulated demand and the second actual demand. For example, the maximum value between the difference between the second simulated demand and the second actual demand and 0 can be used as the second demand loss. For instance, FDC sales loss = max(FDC actual sales - FDC simulated sales, 0), where FDC sales loss is the second demand loss.

[0144] The process for obtaining the first demand loss may be the same as or different from the process for obtaining the second demand loss. This is primarily due to the possibility of FDC (Fulfilled Direct Sales) shifting sales volume, as detailed below:

[0145] Using the allocation simulation module, based on the second real demand and the inventory allocation, the transfer demand of the target item at the second storage center at the simulated historical time point can be obtained. At this time, the first demand loss can be obtained based on the first simulated demand, the first real demand, and the transfer demand.

[0146] The transferred demand can be understood as the demand that should have been shipped from FDC at a given historical time point in the simulation, but was instead shipped from RDC due to insufficient FDC inventory. Using the allocation simulation module, based on the second actual demand and inventory allocation, and further combined with the second initial inventory, the transferred demand can be obtained. For example, FDC transferred sales = max(FDC actual sales - FDC inventory - FDC allocated goods, 0), where FDC transferred sales are the transferred demand, and FDC inventory is the second initial inventory. Then, the first demand loss can be expressed as: RDC sales loss = max(RDC actual sales - RDC simulated sales - FDC transferred sales, 0), where RDC sales loss is the first demand loss. Considering the transferred demand improves the accuracy of determining the first demand loss.

[0147] Based on this, optionally, since the existence of transferred demand will affect the timeliness of delivery, when calculating the loss, not only the first demand loss and the second demand loss, but also the transferred demand can be taken into account, and the loss calculation result can be obtained based on these three factors.

[0148] The above technical solution calculates the first demand loss and the second demand loss, and then obtains the loss calculation result based on these. The loss calculation result can characterize the demand loss situation, which is an important factor in evaluating the quality of inventory forecasting. Therefore, the parameters in the inventory forecasting module can be adjusted based on the loss calculation result, thereby improving the model training effect and ensuring the accuracy of inventory forecasting.

[0149] Based on this, if there are multiple optional historical time points in the simulation, then the loss calculation results are obtained according to the first demand loss and the second demand loss, including:

[0150] The first demand loss is obtained by accumulating the first demand loss corresponding to multiple simulation historical time points, and the second demand loss is obtained by accumulating the second demand loss corresponding to multiple simulation historical time points.

[0151] The loss calculation results are obtained based on the first demand loss and the second demand loss.

[0152] Compared to a single simulated historical time point, loss calculation results based on multiple simulated historical time points are more universally applicable, and the parameters adjusted accordingly can better ensure the accuracy of inventory forecasting. Therefore, we accumulate the first demand loss corresponding to each of the multiple simulated historical time points to obtain the first cumulative loss, and accumulate the corresponding second demand loss to obtain the second cumulative loss. Furthermore, the loss calculation result can be obtained based on these two cumulative losses.

[0153] The above technical solution takes into account that each decision is not isolated and is affected by previous decisions, and that the current decision will affect subsequent decisions. Therefore, by accumulating the demand losses corresponding to multiple historical simulation time points, and adjusting the parameters based on the accumulated results, the loss is minimized through simulation in the process decision-making to adjust the parameters. The adjusted parameters can better ensure the accuracy of inventory forecasting.

[0154] Based on this, in order to better understand the model training process described in the above technical solutions as a whole, specific examples will be used to illustrate it below.

[0155] For example, see Figure 4, where the blue background represents FDC and the orange background represents RDC. Model training can be divided into two stages. The first stage is to train sales prediction unit 1 using historical sales data to obtain a preliminarily trained sales prediction unit 2. The second stage is to train sales prediction unit 2 and inventory estimation unit 1 using sample sales data, initial inventory, and purchase order quantities. The resulting sales prediction unit 3 and inventory estimation unit 2 can be used for end-to-end inventory allocation decisions, ensuring the global optimality of inventory allocation. In practical applications, optionally, the historical time period corresponding to the historical sales data can be a part of the historical time period corresponding to the sample sales data. For example, the sales data of the past 21 days can be used for the first stage of model training, and the sales data of the past 42 days can be used for the second stage of model training and simulation. This can avoid data skipping.

[0156] Specifically, the second phase of training is as follows:

[0157] First, the predicted sales volume is output through the sales forecasting unit 2; then, Ss and Ti are output through the inventory estimation unit 1; and then, the simulated sales volume is output through the inventory simulation module.

[0158] The above steps are repeated s times, meaning after the first execution, the process repeats s-1 times. Then, the cumulative sales loss and prediction bias loss are calculated, and the model minimizes these two losses: loss = w1*loss1 + w2*loss2 + w3*loss3 + loss sale (y,y q );

[0159] Where loss1 represents RDC sales loss, loss2 represents FDC sales loss, loss3 represents FDC transferred sales, and w1, w2, and w3 are hyperparameters; loss sale To predict the bias loss, y represents the predicted sales volume. q This represents the actual sales volume, from which the globally optimal inventory allocation decision can be obtained.

[0160] It should be noted that, firstly, the model to be trained is a large recurrent neural network (RNN) structure, that is, it has recurrent connections, which allows the network to retain and utilize information from previous time steps. The simulation process is achieved through this RNN structure.

[0161] Second, the aforementioned model minimizes regional sales losses and rationally allocates limited inventory using Ss, ensuring that inventory is allocated to the most urgent distribution centers (DCs). The local RDC is treated as a DC, and the local Ss of the RDC is calculated based on the local demand of the RDC. Each allocation to the RDC retains sufficient inventory to counteract local fluctuations. The RDC and FDC output forecasts simultaneously, and the RDC makes decisions simultaneously. By calculating regional sales losses through a simulation mechanism, the inconsistency between forecast deviations and regional spot rates and local sales fulfillment rates of FDCs can be reduced. This solves problems such as unbalanced inventory caused by forecast inconsistencies between RDCs and FDCs, and between FDCs themselves.

[0162] Figure 5 is a flowchart of an inventory transfer method provided in an embodiment of this disclosure. This embodiment is applicable to situations where inventory transfer decisions are made in an end-to-end manner. The method can be executed by the inventory transfer device provided in this embodiment, which can be implemented in software and / or hardware and can be integrated into an electronic device, such as various user terminals or servers.

[0163] Referring to Figure 5, the method in this embodiment may specifically include the following steps:

[0164] S410. Obtain the first target demand quantity of the target item in the first storage center and the second target demand quantity in the second storage center within the target historical time period, and the target inventory estimation module trained according to the model training method described in any embodiment of this disclosure.

[0165] The target historical time period can be understood as the time period that has already occurred and is used in the model reasoning process.

[0166] The first target demand can be understood as time-series data representing the demand for the target item in the RDC within the target historical time period. The second target demand is similar and will not be elaborated here.

[0167] Obtain the first target demand, the second target demand, and the pre-trained target inventory estimation module.

[0168] S420. Using the target inventory estimation module, based on the first target demand, estimate the first target inventory of the target item at the inventory transfer time point, and based on the second target demand, estimate the second target inventory of the target item at the inventory transfer time point.

[0169] The inventory transfer time point can be understood as the time point when inventory transfer needs to be carried out.

[0170] The first target inventory can be understood as the inventory required by the target item in the RDC at the time of inventory transfer. The target inventory estimation module is used to estimate this inventory based on the first target demand.

[0171] The same logic applies to the second target inventory, so I won't go into details here.

[0172] S430. Based on the first target inventory and the second target inventory, determine the inventory transfer decision between the first storage center and the second storage center, so as to transfer the inventory between the first storage center and the second storage center for the target item according to the inventory transfer decision.

[0173] As explained above, the first target inventory represents the inventory required by the RDC for the target item, and the second target inventory represents the inventory required by the FDC for the target item. Therefore, based on the two, and especially by combining the initial inventory of the RDC and FDC for the target item, the inventory transfer amount of the target item from the RDC to the FDC can be obtained, and this inventory transfer amount can be used as the inventory transfer decision.

[0174] Furthermore, based on this inventory allocation decision, inventory transfers can be made between the RDC and FDC for the target item. At this point, the inventory transfer process for the target item at the specified inventory transfer time is complete.

[0175] In light of the application scenarios that may be involved in the embodiments of this disclosure, the above technical solution can optionally be encapsulated in a data-driven end-to-end inventory transfer system. This system makes decisions simultaneously for all data centers (DCs), thereby achieving optimal end-to-end inventory transfer in terms of timeliness while maintaining controllable costs.

[0176] The technical solution of this disclosure realizes end-to-end inventory allocation decision-making by using a pre-trained target inventory estimation module to estimate inventory.

[0177] Figure 6 is a structural block diagram of the model training apparatus provided in the embodiments of this disclosure. This apparatus is used to execute the model training method provided in any of the above embodiments. This apparatus and the model training methods of the above embodiments belong to the same inventive concept. Details not described in detail in the embodiments of the model training apparatus can be referred to the embodiments of the above model training methods. Referring to Figure 6, the apparatus may specifically include: a first acquisition module 510, an estimated inventory module 520, and a model training module 530.

[0178] The first acquisition module 510 is used to acquire the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center during the sample historical time period, and to acquire the model to be trained, wherein the model to be trained includes an inventory estimation module and an allocation simulation module.

[0179] The inventory estimation module 520 is used to obtain the first estimated inventory of the target item at the simulation historical time point based on the first sample demand quantity and the second estimated inventory of the target item at the simulation historical time point based on the second sample demand quantity.

[0180] The model training module 530 is used to perform transfer simulation between the first and second storage centers based on the first and second estimated inventory using the transfer simulation module, so as to adjust the parameters in the inventory estimation module according to the transfer simulation results and obtain the target inventory estimation module.

[0181] Optionally, the inventory forecasting module includes a demand forecasting unit and an inventory forecasting unit;

[0182] The estimated inventory module 520 may include:

[0183] The first predicted demand quantity acquisition submodule is used to input the first sample demand quantity into the demand quantity prediction unit to obtain the first predicted demand quantity of the target item related to the simulation historical time point.

[0184] The first estimated inventory submodule can be used to input the first predicted demand into the inventory estimation unit to obtain the first estimated inventory of the target item at a simulated historical time point.

[0185] Optionally, based on the above-mentioned device, the inventory forecasting unit includes a basic inventory forecasting subunit, a fluctuating inventory forecasting subunit, and an inventory forecasting subunit.

[0186] The first estimated inventory submodule may include:

[0187] The estimated basic inventory unit is used to input the first forecasted demand into the basic inventory estimation sub-unit to obtain the estimated basic inventory.

[0188] The estimated fluctuation inventory unit is used to input the first predicted demand, the first actual demand corresponding to the first predicted demand in the first sample demand, and the prediction deviation obtained based on the first predicted demand and the first actual demand into the fluctuation inventory estimation sub-unit to obtain the estimated fluctuation inventory.

[0189] The first estimated inventory acquisition unit is used to input the estimated basic inventory and the estimated fluctuating inventory into the inventory estimation sub-unit to obtain the first estimated inventory of the target item at the simulated historical time point.

[0190] Optionally, based on the above device, for the allocation lead time corresponding to the simulation historical time point, the estimated basic inventory is the estimated basic inventory within the allocation lead time, and the estimated fluctuation inventory is the estimated fluctuation inventory within the allocation lead time; the inventory estimation unit also includes a mask inventory estimation subunit.

[0191] The first estimated inventory submodule may also include:

[0192] The estimated mask inventory unit is used to mask the mask demand during the allocation lead time from the first predicted demand and sum the mask demand to obtain the estimated mask inventory.

[0193] Accordingly, the first estimated inventory unit is specifically used for:

[0194] Input the estimated base inventory, estimated fluctuation inventory, and estimated mask inventory into the inventory estimation sub-unit to obtain the first estimated inventory of the target item at the simulated historical time point.

[0195] Optionally, the model to be trained may also include a loss calculation module connected to the transfer simulation module, whereby the target item is to be transferred from the first storage center to the second storage center.

[0196] Model training module 530 may include:

[0197] The inventory transfer volume acquisition submodule is used to obtain the inventory transfer volume from the first storage center to the second storage center based on the first estimated inventory and the second estimated inventory using the transfer simulation module.

[0198] The first demand loss submodule is used to obtain the first simulated demand by using the allocation simulation module, based on the inventory allocation and the first real demand of the target item at the first storage center at the simulation historical time point.

[0199] The second demand loss submodule is used to obtain the second simulated demand by using the allocation simulation module, based on the inventory allocation and the second real demand of the target item at the second warehouse center at the simulation historical time point.

[0200] The target inventory forecasting module obtains a sub-module, which uses the loss calculation module to perform loss calculation based on the first actual demand, the first simulated demand, the second actual demand, and the second simulated demand. The parameters in the inventory forecasting module are adjusted according to the loss calculation results to obtain the target inventory forecasting module.

[0201] Optionally, based on the above-described apparatus, the target inventory estimation module may include a sub-module, which may include:

[0202] The second demand loss obtaining unit is used to obtain the first demand loss based on the first simulated demand and the first actual demand using the loss calculation module, and to obtain the second demand loss based on the second simulated demand and the second actual demand.

[0203] The target inventory forecasting module obtains a unit used to calculate the loss based on the first demand loss and the second demand loss, and then adjusts the parameters in the inventory forecasting module according to the loss calculation results to obtain the target inventory forecasting module.

[0204] Based on the above-described apparatus, optionally, the model training module 530 may further include:

[0205] The demand transfer submodule is used to obtain the demand transfer quantity of the target item at the second warehouse center at the simulated historical time point by using the allocation simulation module, based on the second real demand quantity and the inventory allocation quantity.

[0206] The second unit for obtaining demand loss may include:

[0207] The first demand loss sub-unit is used to obtain the first demand loss based on the first simulated demand, the first actual demand, and the transferred demand.

[0208] Alternatively, the number of historical time points in the simulation can be multiple, and the target inventory estimation module can obtain units that may include:

[0209] The second cumulative loss sub-unit is used to accumulate the first demand loss corresponding to multiple simulation historical time points to obtain the first cumulative loss, and to accumulate the second demand loss corresponding to multiple simulation historical time points to obtain the second cumulative loss.

[0210] The loss calculation results are used to obtain sub-units, which are used to obtain the loss calculation results based on the first cumulative loss and the second cumulative loss.

[0211] Optionally, based on the above-mentioned device, the first estimated inventory includes a first estimated safety stock and a first estimated target inventory, and the second estimated inventory includes a second estimated safety stock and a second estimated target inventory.

[0212] The inventory transfer quantity acquisition submodule may include:

[0213] The first simulated inventory obtaining unit is used to obtain the first transfer amount from the first storage center to the second storage center based on the first estimated safety stock and the second estimated safety stock, and to perform inventory transfer simulation based on the first transfer amount to obtain the first simulated inventory of the first storage center.

[0214] The second allocation unit is used to obtain the second allocation amount from the first storage center to the second storage center when the first simulated inventory is greater than the first estimated safety stock, based on the first simulated inventory, the first estimated target inventory and the second estimated target inventory.

[0215] The inventory transfer quantity obtaining unit is used to obtain the inventory transfer quantity based on the first transfer quantity and the second transfer quantity.

[0216] The model training device provided in this embodiment achieves end-to-end inventory allocation decisions through the cooperation of various modules.

[0217] The model training apparatus provided in this disclosure can execute the model training method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the execution method.

[0218] It is worth noting that in the embodiments of the above-mentioned model training device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the protection scope of this disclosure.

[0219] Figure 7 is a structural block diagram of the inventory transfer device provided in the embodiments of this disclosure. This device is used to execute the inventory transfer method provided in any of the above embodiments. This device and the inventory transfer methods of the above embodiments belong to the same inventive concept. Details not described in detail in the embodiments of the inventory transfer device can be referred to the embodiments of the above inventory transfer methods. Referring to Figure 7, the device may specifically include: a second acquisition module 610, a target inventory module 620, and an inventory transfer module 630.

[0220] The second acquisition module 610 is used to acquire the first target demand quantity of the target item in the first storage center and the second target demand quantity in the second storage center during the target historical time period, as well as the target inventory estimation module trained according to the model training method described in any embodiment of this disclosure.

[0221] The target inventory module 620 is used to estimate the first target inventory of the target item at the inventory transfer time point based on the first target demand, and to estimate the second target inventory of the target item at the inventory transfer time point based on the second target demand, using the target inventory estimation module.

[0222] The inventory transfer module 630 is used to determine the inventory transfer decision between the first storage center and the second storage center based on the first target inventory and the second target inventory, so as to transfer the inventory between the first storage center and the second storage center for the target item according to the inventory transfer decision.

[0223] The inventory transfer device provided in this embodiment achieves end-to-end inventory transfer decisions through the cooperation of various modules.

[0224] The inventory transfer device provided in this disclosure can execute the inventory transfer method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects for executing the method.

[0225] It is worth noting that in the above-mentioned embodiments of the inventory transfer device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of this disclosure.

[0226] Figure 8 illustrates a schematic diagram of the structure of an electronic device 10 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0227] As shown in Figure 8, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer programs stored in the ROM 12 or loaded into the RAM 13 from storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0228] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0229] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as model training methods or inventory transfer methods.

[0230] In some embodiments, the model training method or inventory transfer method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the model training method or inventory transfer method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the model training method or inventory transfer method by any other suitable means (e.g., by means of firmware).

[0231] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0232] Computer programs used to implement the methods of this disclosure may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0233] In the context of this disclosure, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0234] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0235] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0236] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0237] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this disclosure can be achieved, and this is not limited herein.

[0238] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

A model training method, comprising: The system obtains the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and obtains the model to be trained, wherein the model to be trained includes an inventory estimation module and an allocation simulation module. Using the inventory estimation module, based on the first sample demand, a first estimated inventory of the target item at the simulated historical time point is obtained; and based on the second sample demand, a second estimated inventory of the target item at the simulated historical time point is obtained; and Using the aforementioned transfer simulation module, based on the first estimated inventory and the second estimated inventory, a transfer simulation is performed between the first storage center and the second storage center. The parameters in the inventory estimation module are adjusted according to the transfer simulation results to obtain the target inventory estimation module. According to the method of claim 1, wherein, The inventory forecasting module includes a demand forecasting unit and an inventory forecasting unit; The step of using the inventory estimation module to obtain the first estimated inventory of the target item at a simulated historical time point based on the first sample demand includes: The first sample demand is input into the demand prediction unit to obtain the first predicted demand of the target item related to the simulation historical time point. as well as The first predicted demand is input into the inventory estimation unit to obtain the first estimated inventory of the target item at the simulated historical time point. The method according to claim 2, wherein, The inventory forecasting unit includes a basic inventory forecasting subunit, a fluctuating inventory forecasting subunit, and an inventory forecasting subunit. The step of inputting the first predicted demand into the inventory estimation unit to obtain the first estimated inventory of the target item at the simulated historical time point includes: Input the first predicted demand into the basic inventory estimation subunit to obtain the estimated basic inventory; The first predicted demand, the first actual demand corresponding to the first predicted demand in the first sample demand, and the prediction deviation obtained based on the first predicted demand and the first actual demand are input into the fluctuation inventory estimation subunit to obtain the estimated fluctuation inventory; and The estimated basic inventory and the estimated fluctuating inventory are input into the inventory estimation subunit to obtain the first estimated inventory of the target item at the simulated historical time point. The method according to claim 3, wherein, For the allocation lead time corresponding to the simulated historical time point, the estimated basic inventory is the estimated basic inventory within the allocation lead time, and the estimated fluctuation inventory is the estimated fluctuation inventory within the allocation lead time. The inventory estimation unit also includes a masked inventory estimation subunit; The step of inputting the first predicted demand into the inventory estimation unit to obtain the first estimated inventory of the target item at the simulated historical time point further includes: Using the masked inventory estimation subunit, the masked demand during the allocation lead time is masked from the first predicted demand, and the masked demand is summed to obtain the estimated masked inventory. Accordingly, inputting the estimated basic inventory and the estimated fluctuating inventory into the inventory estimation subunit to obtain the first estimated inventory of the target item at the simulated historical time point includes: The estimated base inventory, the estimated fluctuation inventory, and the estimated mask inventory are input into the inventory estimation subunit to obtain the first estimated inventory of the target item at the simulated historical time point. According to the method of claim 1, wherein, The model to be trained also includes a loss calculation module connected to the transfer simulation module, wherein the target item is to be transferred from the first storage center to the second storage center; The method of using the transfer simulation module to perform transfer simulation between the first and second warehousing centers based on the first and second estimated inventory, and adjusting the parameters in the inventory estimation module according to the transfer simulation results to obtain the target inventory estimation module, includes: Using the aforementioned transfer simulation module, the following three steps are performed to simulate the transfer between the first and second warehousing centers: Based on the first estimated inventory and the second estimated inventory, the inventory transfer volume from the first warehousing center to the second warehousing center is obtained; The first simulated demand is obtained based on the inventory transfer volume and the first real demand for the target item at the first storage center at the simulated historical time point. Based on the inventory transfer volume and the second actual demand for the target item at the second storage center at the simulated historical time point, the second simulated demand is obtained; and Using the loss calculation module, loss is calculated based on the first actual demand, the first simulated demand, the second actual demand, and the second simulated demand. The parameters in the inventory forecasting module are then adjusted according to the loss calculation results to obtain the target inventory forecasting module. The method according to claim 5, wherein, The loss calculation module performs loss calculations based on the first actual demand, the first simulated demand, the second actual demand, and the second simulated demand. The parameters in the inventory forecasting module are adjusted according to the loss calculation results to obtain the target inventory forecasting module, which includes: Using the loss calculation module, a first demand loss is obtained based on the first simulated demand and the first actual demand, and a second demand loss is obtained based on the second simulated demand and the second actual demand; and Based on the first demand loss and the second demand loss, a loss calculation result is obtained, and the parameters in the inventory forecasting module are adjusted according to the loss calculation result to obtain the target inventory forecasting module. The method according to claim 6, wherein, The allocation simulation module is also used to perform the following steps: Based on the second actual demand and the inventory transfer volume, the transfer demand of the target item in the second warehouse center at the simulated historical time point is obtained; Accordingly, obtaining the first demand loss based on the first simulated demand and the first actual demand includes: The first demand loss is obtained based on the first simulated demand, the first actual demand, and the transferred demand. The method according to claim 6, wherein, The number of simulated historical time points is multiple. The loss calculation result obtained based on the first demand loss and the second demand loss includes: The first cumulative loss is obtained by accumulating the first demand loss corresponding to each of the multiple simulation historical time points, and the second cumulative loss is obtained by accumulating the second demand loss corresponding to each of the multiple simulation historical time points; and The loss calculation result is obtained based on the first cumulative loss and the second cumulative loss. The method according to claim 5, wherein, The first estimated inventory includes a first estimated safety stock and a first estimated target inventory, and the second estimated inventory includes a second estimated safety stock and a second estimated target inventory; The step of obtaining the inventory transfer volume from the first warehousing center to the second warehousing center based on the first estimated inventory and the second estimated inventory includes: Based on the first estimated safety stock and the second estimated safety stock, a first transfer volume from the first storage center to the second storage center is obtained, and inventory transfer simulation is performed based on the first transfer volume to obtain the first simulated inventory of the first storage center. If the first simulated inventory is greater than the first estimated safety stock, a second transfer quantity from the first storage center to the second storage center is obtained based on the first simulated inventory, the first estimated target inventory, and the second estimated target inventory; and The inventory transfer amount is obtained based on the first transfer amount and the second transfer amount. An inventory transfer method includes: The first target demand quantity of the target item in the first warehouse center and the second target demand quantity in the second warehouse center within the target historical time period are obtained, as well as the target inventory estimation module trained according to the model training method of any one of claims 1-9. Using the target inventory estimation module, based on the first target demand, the first target inventory of the target item at the inventory transfer time point is estimated, and based on the second target demand, the second target inventory of the target item at the inventory transfer time point is estimated. Based on the first target inventory and the second target inventory, an inventory transfer decision is determined between the first warehousing center and the second warehousing center, so as to transfer inventory between the first warehousing center and the second warehousing center for the target item according to the inventory transfer decision. A model training device, comprising: The first acquisition module is used to acquire the first sample demand quantity of the target item in the first storage center and the second sample demand quantity in the second storage center within the sample historical time period, and to acquire the model to be trained, wherein the model to be trained includes an inventory estimation module and an allocation simulation module. An inventory estimation module is used to obtain, based on the first sample demand, a first estimated inventory of the target item at a simulated historical time point, and based on the second sample demand, a second estimated inventory of the target item at the simulated historical time point. The model training module is used to perform transfer simulation between the first storage center and the second storage center based on the first estimated inventory and the second estimated inventory using the transfer simulation module, so as to adjust the parameters in the inventory estimation module according to the transfer simulation results and obtain the target inventory estimation module. An inventory transfer device, comprising: The second acquisition module is used to acquire the first target demand quantity of the target item in the first storage center and the second target demand quantity in the second storage center during the target historical time period, as well as the target inventory estimation module trained according to the model training method of any one of claims 1-9. The target inventory module is used to estimate the first target inventory of the target item at the inventory transfer time point based on the first target demand, and to estimate the second target inventory of the target item at the inventory transfer time point based on the second target demand. The inventory transfer module is used to determine an inventory transfer decision between the first warehousing center and the second warehousing center based on the first target inventory and the second target inventory, so as to transfer inventory between the first warehousing center and the second warehousing center for the target item according to the inventory transfer decision. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the model training method as described in any one of claims 1-9, or the inventory transfer method as described in claim 10. A computer-readable storage medium, wherein, The computer-readable storage medium stores computer instructions that cause a processor to execute the model training method as described in any one of claims 1-9, or the inventory transfer method as described in claim 10. A computer program product, wherein, The computer program product includes a computer program that, when executed by a processor, implements the model training method as described in any one of claims 1-9, or the inventory transfer method as described in claim 10.

Images