Model training method, inventory article replenishment method, device, equipment and medium

Abstract

The embodiment of the invention discloses a model training method, an inventory article replenishment method and device, equipment and a medium. A specific embodiment of the method comprises the steps of obtaining first historical inventory article data and historical inventory decision feedback information; generating a historical article circulation predicted quantity under each quantile; according to the historical article circulation predicted quantity under each quantile, generating inventory replenishment information under each quantile; according to the historical inventory decision feedback information, historical replenishment optimal decision information is generated; and performing model training on the initial article prediction flow quantity generation model and the initial replenishment strategy information generation model to generate an article prediction flow quantity generation model and a replenishment strategy information generation model. The implementation mode is related to artificial intelligence, and the predicted article flow quantity and the replenishment strategy information can be accurately generated by utilizing the predicted article flow quantity generation model and the replenishment strategy information generation model, so that accurate replenishment of inventory articles is realized.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a model training method, an inventory item replenishment method, an apparatus, a device, and a medium. Background Art

[0002] At present, inventory management is the cornerstone of the logistics supply chain. Accurate inventory management can reduce storage costs or transportation costs and improve efficiency. How to effectively achieve accurate inventory replenishment for inventory items has become a key development direction in the current logistics field. For the replenishment of inventory items, the usual method is to generate a model through the relevant end-to-end replenishment strategy to directly output the corresponding replenishment decision based on the existing warehouse storage situation.

[0003] However, the inventors have found that when the above method is used to carry out replenishment processing, the following technical problems often occur:

[0004] The replenishment strategy generation model has uncertainties, and there are also various uncertainties in demand, which often lead to deterioration in turnover or spot rate. Replenishment processing is carried out based on the replenishment strategy output by the replenishment strategy generation model, resulting in inaccurate replenishment.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0006] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] Some embodiments of the present disclosure propose model training methods, inventory item replenishment methods, devices, equipment and media to solve the technical problems mentioned in the above background technology section.

[0008] In a first aspect, some embodiments of the present disclosure provide a model training method, including: acquiring first historical inventory item data and historical inventory decision feedback information; generating historical item turnover prediction quantities at each quantile using an initial item prediction turnover generation model based on the first historical inventory item data; generating inventory replenishment information at each quantile based on the historical item turnover prediction quantities at each quantile; generating historical replenishment optimal decision information based on the historical inventory decision feedback information; training the initial item prediction turnover generation model and the initial replenishment strategy information generation model based on the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information to generate an item prediction turnover generation model and a replenishment strategy information generation model.

[0009] Optionally, the above-mentioned initial item predicted turnover volume generation model includes: an initial encoding layer and an initial decoding layer; and the above-mentioned first historical inventory item data is based on the initial item predicted turnover volume generation model to generate the historical item turnover prediction volume at each quantile, including: preprocessing the above-mentioned first historical inventory item data to generate preprocessed data; inputting the above-mentioned preprocessed data into the above-mentioned initial encoding layer to generate encoding information; inputting the above-mentioned encoding information into the above-mentioned initial decoding layer to generate the historical item turnover prediction volume at each of the above-mentioned quantiles.

[0010] Optionally, the above-mentioned initial coding layer includes: an initial recent item circulation volume coding layer, an initial flow rate coding layer, an initial value information coding layer and an initial concurrent item circulation coding layer; and the above-mentioned pre-processed data is input into the above-mentioned initial coding layer to generate coding information, including: inputting the recent item circulation data included in the above-mentioned pre-processed data into the above-mentioned initial recent item circulation volume coding layer to generate recent item circulation coding information; inputting the flow rate data included in the above-mentioned pre-processed data into the above-mentioned initial flow rate coding layer to generate flow rate coding information; inputting the value data included in the above-mentioned pre-processed data into the above-mentioned initial value information coding layer to generate value coding information; inputting the concurrent item circulation data included in the above-mentioned pre-processed data into the above-mentioned initial concurrent item circulation coding layer to generate concurrent item circulation coding information; generating the above-mentioned coding information according to the above-mentioned recent item circulation coding information, the above-mentioned flow rate coding information, the above-mentioned value coding information and the above-mentioned concurrent item circulation coding information.

[0011] Optionally, the above inventory replenishment information is inventory support days; and

[0012] The generating of inventory replenishment information at each of the above-mentioned quantiles according to the historical item turnover forecast quantity at each of the above-mentioned quantiles includes: determining the number of inventory support days at each of the above-mentioned quantiles according to the historical item turnover forecast quantity at each of the above-mentioned quantiles.

[0013] Optionally, the above-mentioned historical replenishment optimal decision information is the historical replenishment optimal decision days; and the above-mentioned generating the historical replenishment optimal decision information based on the above-mentioned historical inventory decision feedback information includes: determining whether the inventory decision corresponding to the above-mentioned historical inventory decision feedback information includes a manual decision; in response to the determination that it includes, determining the replenishment quantity corresponding to the above-mentioned manual decision; generating the inventory support days corresponding to the above-mentioned replenishment quantity as the first inventory support days; determining the inventory support days under the model decision in the above-mentioned inventory decision as the second inventory support days; generating the above-mentioned historical replenishment optimal decision days based on the above-mentioned first inventory support days and the above-mentioned second inventory support days.

[0014] Optionally, the above-mentioned initial item forecast turnover generation model and initial replenishment strategy information generation model are trained according to the historical item turnover forecast quantity at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, so as to generate an item forecast turnover generation model and a replenishment strategy information generation model, including: matching the inventory support days at each quantile with the historical replenishment optimal decision days to generate a target forecast quantile; obtaining the actual inventory support days at each quantile; matching the actual inventory at each quantile The supporting days are matched with the above-mentioned historical replenishment optimal decision days to generate a target real quantile; the quantile loss information for the above-mentioned target predicted quantile and the above-mentioned target real quantile is generated; the item circulation loss information is generated according to the historical item circulation real quantity under the above-mentioned each quantile and the historical item circulation predicted quantity under the above-mentioned each quantile; the initial item predicted circulation quantity generation model and the initial replenishment strategy information generation model are trained according to the above-mentioned quantile loss information and the above-mentioned item circulation loss information to generate an item predicted circulation quantity generation model and a replenishment strategy information generation model.

[0015] Optionally, the above-mentioned training of the initial item predicted turnover volume generation model and the initial replenishment strategy information generation model based on the above-mentioned quantile loss information and the above-mentioned item circulation loss information to generate the item predicted turnover volume generation model and the replenishment strategy information generation model includes: acquiring replenishment parameter data; generating a total inventory forecast according to the above-mentioned replenishment parameter data and the historical item circulation forecast under each quantile; generating inventory loss information for the above-mentioned total inventory forecast; training of the initial item predicted turnover volume generation model and the initial replenishment strategy information generation model based on the above-mentioned quantile loss information, the above-mentioned item circulation loss information and the above-mentioned inventory loss information to generate the item predicted turnover volume generation model and the replenishment strategy information generation model.

[0016] In a second aspect, some embodiments of the present disclosure provide a model training device, including: a first acquisition unit, configured to acquire first historical inventory item data and historical inventory decision feedback information; a first generation unit, configured to generate historical item turnover prediction quantities at each quantile based on the above-mentioned first historical inventory item data using an initial item predicted turnover quantity generation model; a second generation unit, configured to generate inventory replenishment information at each quantile based on the historical item turnover prediction quantities at each quantile; a third generation unit, configured to generate historical replenishment optimal decision information based on the above-mentioned historical inventory decision feedback information; a training unit, configured to perform model training on the above-mentioned initial item predicted turnover quantity generation model and the initial replenishment strategy information generation model based on the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile and the above-mentioned historical replenishment optimal decision information, so as to generate an item predicted turnover quantity generation model and a replenishment strategy information generation model.

[0017] Optionally, the above-mentioned initial item predicted turnover generation model includes: an initial encoding layer and an initial decoding layer; and the first generation unit can be configured to perform data preprocessing on the above-mentioned first historical inventory item data to generate preprocessed data; input the above-mentioned preprocessed data into the above-mentioned initial encoding layer to generate encoding information; input the above-mentioned encoding information into the above-mentioned initial decoding layer to generate the historical item turnover prediction quantity at the above-mentioned each quantile point.

[0018] Optionally, the above-mentioned initial coding layer includes: an initial recent item circulation volume coding layer, an initial flow rate coding layer, an initial value information coding layer and an initial concurrent item circulation coding layer; and the first generating unit can be configured to input the recent item circulation data included in the above-mentioned preprocessing data into the above-mentioned initial recent item circulation volume coding layer to generate recent item circulation coding information; input the flow rate data included in the above-mentioned preprocessing data into the above-mentioned initial flow rate coding layer to generate flow rate coding information; input the value data included in the above-mentioned preprocessing data into the above-mentioned initial value information coding layer to generate value coding information; input the concurrent item circulation data included in the above-mentioned preprocessing data into the above-mentioned initial concurrent item circulation coding layer to generate concurrent item circulation coding information; generate the above-mentioned coding information according to the above-mentioned recent item circulation coding information, the above-mentioned flow rate coding information, the above-mentioned value coding information and the above-mentioned concurrent item circulation coding information.

[0019] Optionally, the above-mentioned inventory replenishment information is the inventory support days; and the second generating unit can be configured to determine and generate the inventory support days at the above-mentioned each quantile according to the historical item turnover forecast quantity at the above-mentioned each quantile.

[0020] Optionally, the above-mentioned historical replenishment optimal decision information is the historical replenishment optimal decision days; and the third generating unit can be configured to determine whether the inventory decision corresponding to the above-mentioned historical inventory decision feedback information includes a manual decision; in response to the determination, determine the replenishment quantity corresponding to the above-mentioned manual decision; generate the inventory support days corresponding to the above-mentioned replenishment quantity as the first inventory support days; determine the inventory support days under the model decision in the above-mentioned inventory decision as the second inventory support days; and generate the above-mentioned historical replenishment optimal decision days based on the above-mentioned first inventory support days and the above-mentioned second inventory support days.

[0021] Optionally, the training unit may be configured to match the inventory support days at each of the above quantiles with the historical replenishment optimal decision days in number of days to generate a target prediction quantile; obtain the actual inventory support days at each of the above quantiles; match the actual inventory support days at each of the above quantiles with the historical replenishment optimal decision days in number of days to generate a target real quantile; generate quantile loss information for the target prediction quantile and the target real quantile; generate item circulation loss information based on the historical item circulation actual quantity at each of the above quantiles and the historical item circulation predicted quantity at each of the above quantiles; perform model training on the initial item prediction circulation quantity generation model and the initial replenishment strategy information generation model based on the quantile loss information and the item circulation loss information to generate an item prediction circulation quantity generation model and a replenishment strategy information generation model.

[0022] Optionally, the training unit can be configured to obtain replenishment parameter data; generate a total inventory forecast based on the above replenishment parameter data and the historical item turnover forecast at the above each quantile; generate inventory loss information for the above total inventory forecast; and perform model training on the above initial item forecast turnover generation model and the initial replenishment strategy information generation model based on the above quantile loss information, the above item turnover loss information and the above inventory loss information to generate an item forecast turnover generation model and a replenishment strategy information generation model.

[0023] In a third aspect, some embodiments of the present disclosure provide a method for replenishing inventory items, comprising: acquiring second historical inventory item data; generating future item turnover predictions at each quantile using a pre-trained item turnover prediction generation model based on the second historical inventory item data, wherein the item turnover prediction generation model is generated based on a model training method; generating future replenishment strategy information using a pre-trained replenishment strategy information generation model based on the future item turnover predictions at each quantile, wherein the future replenishment strategy information includes: a future optimal quantile, wherein the item turnover prediction generation model is generated based on a model training method; and performing replenishment processing on the inventory items based on the future optimal quantile.

[0024] Optionally, performing replenishment processing for the inventory items according to the future optimal quantile includes: determining a future optimal replenishment quantity corresponding to the future optimal quantile; and performing replenishment processing for the inventory items according to the future optimal replenishment quantity.

[0025] In a fourth aspect, some embodiments of the present disclosure provide an inventory item replenishment device, comprising: a second acquisition unit, configured to acquire second historical inventory item data; a fourth generation unit, configured to generate future item turnover predictions at each quantile based on the second historical inventory item data using a pre-trained item turnover prediction generation model, wherein the item turnover prediction generation model is generated based on a model training method; a fifth generation unit, configured to generate future replenishment strategy information based on the future item turnover predictions at each quantile based on a pre-trained replenishment strategy information generation model, wherein the future replenishment strategy information includes: a future optimal quantile, and the item turnover prediction generation model is generated based on a model training method; an execution unit, configured to execute replenishment processing for the inventory items based on the future optimal quantile.

[0026] In a fifth aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the implementation methods in the first and third aspects.

[0027] In a sixth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any one of the implementation modes of the first aspect and the third aspect is implemented.

[0028] In a seventh aspect, some embodiments of the present disclosure provide a computer program product, including a computer program, which, when executed by a processor, implements the method described in any one of the implementation modes of the first and third aspects above.

[0029] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, using the item prediction turnover generation model and the replenishment strategy information generation model, the item prediction turnover and replenishment strategy information can be accurately generated to achieve accurate replenishment of inventory items. Specifically, the reason why the relevant replenishment strategy information is not accurate enough is that there is uncertainty in the replenishment strategy generation model, and there are various uncertainties in demand, and turnover deterioration or spot rate deterioration often occurs. Replenishment processing is carried out according to the replenishment strategy output by the replenishment strategy generation model, resulting in inaccurate replenishment. Based on this, the model training method of some embodiments of the present disclosure first obtains the first historical inventory item data and the historical inventory decision feedback information. Here, the first historical inventory item data obtained is the inventory data related to the inventory item, which is used as the data basis to facilitate the generation of the historical item turnover prediction at each subsequent quantile. The historical inventory decision feedback information obtained can fully reflect the specific inventory feedback situation under the execution of the historical inventory decision. Through the inventory feedback, we can fully understand whether the replenishment quantity generated by the model is insufficient or excessive in the historical time, so as to adjust and solve the problem in the next inventory decision, so as to continuously improve the accuracy of the replenishment quantity output by the model, so that the subsequent feedback information of the historical inventory decision can be based on the training process of the initial item prediction turnover generation model and the initial replenishment strategy information generation model, and fully consider the problems existing in the historical decision, so as to obtain a more accurate model for subsequent output. Then, according to the above-mentioned first historical inventory item data, the initial item prediction turnover generation model can be used to accurately generate the historical item turnover prediction quantity at each quantile. Here, by determining the historical item turnover prediction quantity at each quantile, the circulation distribution of the item turnover prediction quantity at the predetermined historical time can be more fully reflected. In addition, it is convenient to determine the inventory replenishment information at each quantile in the future. Furthermore, based on the historical predicted item turnover at each quantile, the inventory replenishment information at each quantile can be accurately generated, so as to accurately generate model training data for the initial item predicted turnover generation model and the initial replenishment strategy information generation model training with the historical predicted item turnover at each quantile. Next, based on the historical inventory decision feedback information, the historical replenishment optimal decision information is generated as the strategy label of the optimal strategy for replenishment in historical time, which is convenient for subsequent model training. Finally, based on the historical predicted item turnover at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, the initial item predicted turnover generation model and the initial replenishment strategy information generation model are trained to generate the item predicted turnover generation model and the replenishment strategy information generation model.In summary, based on the historical inventory decision feedback information, as well as based on the generated historical item turnover predictions at each quantile and the inventory replenishment information at each quantile, the initial item prediction turnover generation model and the initial replenishment strategy information generation model can be effectively and accurately trained to obtain a more accurate item prediction turnover generation model and a replenishment strategy information generation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0031] Figure 1 is a schematic diagram of an application scenario of a model training method according to some embodiments of the present disclosure;

[0032] Figure 2 is a flowchart of some embodiments of the model training method according to the present disclosure;

[0033] Figure 3 is a flowchart of other embodiments of the model training method according to the present disclosure;

[0034] Figure 4 is a flow chart of some embodiments of a method for replenishing inventory items according to the present disclosure;

[0035] Figure 5 is a schematic diagram of the structure of some embodiments of the model training device according to the present disclosure;

[0036] Figure 6 is a schematic structural diagram of some embodiments of the inventory item replenishment device according to the present disclosure;

[0037] Figure 7 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0038] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0039] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0040] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0041] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0042] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0043] With regard to the collection, storage, and use of user personal information (such as the first historical inventory item data and the second historical inventory item data) involved in this disclosure, before performing the corresponding operations, the relevant organizations or individuals shall fulfill obligations including conducting personal information security impact assessments, fulfilling the obligation to inform the personal information subject, and obtaining the authorization and consent of the personal information subject in advance.

[0044] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0045] Figure 1 It is a schematic diagram of an application scenario of the model training method according to some embodiments of the present disclosure.

[0046] exist Figure 1In the application scenario, first, the electronic device 101 can obtain the first historical inventory item data 101 and the historical inventory decision feedback information 102. In this application scenario, the first historical inventory item data 101 can be: "Basic information of items", "Item sales data", "Item delivery time". The historical inventory decision feedback information 102 can be: "There is a manual order", "There is a shortage in the last replenishment order execution cycle". Then, the electronic device 101 can generate the historical item turnover prediction 103 at each quantile based on the above first historical inventory item data 101 and the initial item prediction turnover generation model 106. In this application scenario, the historical item circulation prediction amount 103 at each quantile may be: "0.1 quantile: 399", "0.2 quantile: 599", "0.3 quantile: 739", "0.4 quantile: 909", "0.5 quantile: 1229", "0.6 quantile: 1459", "0.7 quantile: 1899", "0.8 quantile: 2199", "0.9 quantile: 2499". Then, the electronic device 101 may generate the inventory replenishment information 104 at each quantile according to the historical item circulation prediction amount 103 at each quantile. In this application scenario, the inventory replenishment information 104 at each quantile may be: "0.1 quantile: 2 days", "0.2 quantile: 4 days", "0.3 quantile: 6 days", "0.4 quantile: 7 days", "0.5 quantile: 12 days", "0.6 quantile: 16 days", "0.7 quantile: 17 days", "0.8 quantile: 19 days", "0.9 quantile: 21 days". Furthermore, the electronic device 101 may generate the historical replenishment optimal decision information 105 according to the above historical inventory decision feedback information 102. In this application scenario, the historical replenishment optimal decision information 105 may be "14 days". Finally, based on the historical item turnover prediction 103 at each quantile, the inventory replenishment information 104 at each quantile, and the historical replenishment optimal decision information 105, the initial item prediction turnover generation model 106 and the initial replenishment strategy information generation model 107 are trained to generate an item prediction turnover generation model 108 and a replenishment strategy information generation model 109.

[0047] It should be noted that the electronic device 101 can be hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules for providing distributed services, for example, or it can be implemented as a single software or software module. No specific limitation is made here.

[0048] It should be understood that Figure 1 The number of electronic devices in the embodiment is only for illustration. Any number of electronic devices may be provided according to implementation requirements.

[0049] Continue to refer Figure 2 , shows a process 200 of some embodiments of the model training method according to the present disclosure. The model training method comprises the following steps:

[0050] Step 201, obtaining first historical inventory item data and historical inventory decision feedback information.

[0051] In some embodiments, the execution entity of the above model training method (for example Figure 1The electronic device 101 shown can obtain the first historical inventory item data and historical inventory decision feedback information through a wired connection or a wireless connection. Among them, the first historical inventory item data can be the inventory item data under the first historical time period. The inventory item data can be data related to each inventory item in the warehouse. In practice, the inventory item data can include but is not limited to at least one of the following: basic information of the item, item circulation data, item delivery time, and inventory related information. The basic information of the item can include: basic attribute information and feature information of the item. In practice, the basic information of the item can include but is not limited to at least one of the following: item name, item brand. The basic information of the item can better represent the information of the SKU, and this type of information has an intuitive and important impact on the item circulation and inventory of the SKU. In practice, for the logistics field, the corresponding value circulation can be item sales. The corresponding value circulation data can be item sales data. In practice, the value circulation data can include but is not limited to at least one of the following: SKU, date, sales volume, PV (Page View). The delivery time of the item can be the delivery time of the supplier, which is the key information affecting the amount of replenishment decision. The historical inventory decision feedback information may be decision feedback information generated by executing the inventory replenishment decision in the second historical time period. The second historical time period may be the same time period as the first historical time period, or may be a different historical time period. The second historical time period may be the execution time period of at least one inventory replenishment decision before the third historical time period. The third historical time period may be the historical time period corresponding to the historical item turnover prediction amount at each quantile. The third historical time period may be the historical time period after the first historical time period and the second historical time period. The historical item turnover prediction amount at each quantile is the item turnover prediction amount for the third historical time period. That is, the historical item turnover prediction amount at each quantile is the item turnover prediction amount at each quantile under the total item turnover prediction amount corresponding to the third historical time period. The historical inventory decision feedback information may characterize the strategy execution problem of at least one inventory replenishment strategy. In practice, the historical inventory decision feedback information may include: whether there is a manual order within 3 days after the execution of the model automatic replenishment order, and whether there is a shortage in the previous replenishment order execution cycle. The model automatic replenishment order may be a list output by the model that characterizes how to replenish inventory items. The manual order may be a list generated manually that characterizes how to replenish inventory items. Manual orders can reflect the attitude of business personnel towards the model's automatic replenishment orders, and can effectively reflect whether the replenishment quantity in the model's automatic replenishment orders is too much or too little. Out-of-stock in the previous replenishment order execution cycle indicates that the historical replenishment decision was too low, and the model needs to make decision corrections.

[0052] Step 202: Based on the first historical inventory item data, the initial item predicted turnover volume generation model is used to generate the historical item turnover prediction volume at each quantile.

[0053] In some embodiments, the execution subject may generate the historical predicted item turnover at each quantile based on the first historical inventory item data using the initial item predicted turnover generation model. The initial item predicted turnover generation model may be an item predicted turnover generation model that has not yet been trained. The item predicted turnover generation model may be a neural network model that generates the predicted item turnover at each quantile in the target time period. In practice, the historical predicted item turnover at each quantile corresponds to the third historical time period. Each quantile may be obtained through network learning or may be a plurality of pre-set quantiles. In practice, each quantile may be: [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.85, 0.9, 0.95, 0.975]. For example, the item predicted turnover generation model may be a Transformer model. For example, the historical item circulation prediction amount at each quantile can be: [the historical item circulation prediction amount at the 0.1 quantile, the historical item circulation prediction amount at the 0.2 quantile, the historical item circulation prediction amount at the 0.3 quantile, the historical item circulation prediction amount at the 0.4 quantile, the historical item circulation prediction amount at the 0.5 quantile, the historical item circulation prediction amount at the 0.6 quantile, the historical item circulation prediction amount at the 0.7 quantile, the historical item circulation prediction amount at the 0.8 quantile, the historical item circulation prediction amount at the 0.85 quantile, the historical item circulation prediction amount at the 0.9 quantile, the historical item circulation prediction amount at the 0.95 quantile, the historical item circulation prediction amount at the 0.975 quantile].

[0054] As an example, first, the execution subject may perform data preprocessing on the first historical inventory item data to generate preprocessed data, and then input the preprocessed data into the initial item forecast turnover generation model to generate historical item turnover forecasts at each quantile.

[0055] Step 203: Generate inventory replenishment information at each of the quantiles according to the historical item turnover forecast at each of the quantiles.

[0056] In some embodiments, the execution entity may generate the inventory replenishment information at each quantile according to the historical item turnover prediction at each quantile. The inventory replenishment information may be information related to replenishment of inventory items. For example, the inventory replenishment information may include the inventory replenishment quantity. The inventory replenishment information at each quantile may be the inventory replenishment information predicted at each quantile. For example, for each quantile: [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.85, 0.9, 0.95, 0.975], the inventory replenishment information corresponding to each quantile may be: [inventory replenishment information corresponding to the 0.1 quantile, inventory replenishment information corresponding to the 0.2 quantile, inventory replenishment information corresponding to the 0.3 quantile, inventory replenishment information corresponding to the 0.4 quantile, inventory replenishment information corresponding to the 0.5 quantile, inventory replenishment information corresponding to the 0.6 quantile, inventory replenishment information corresponding to the 0.7 quantile, inventory replenishment information corresponding to the 0.8 quantile, inventory replenishment information corresponding to the 0.85 quantile, inventory replenishment information corresponding to the 0.9 quantile, inventory replenishment information corresponding to the 0.95 quantile, inventory replenishment information corresponding to the 0.975 quantile].

[0057] As an example, the execution entity may directly determine the historical item flow forecast quantity at each quantile as the inventory replenishment quantity at each quantile, as the inventory replenishment information at each quantile.

[0058] In some optional implementations of some embodiments, the above-mentioned inventory replenishment information is inventory support days. The inventory support days may be the number of days that the supply of goods can be maintained through inventory replenishment.

[0059] Optionally, the execution entity may determine and generate the inventory support days at each of the quantiles based on the historical item turnover forecast at each of the quantiles.

[0060] As an example, for each quantile in each quantile, first, determine the historical item turnover prediction amount corresponding to the quantile as the target historical item turnover prediction amount. Then, determine the number of days of item turnover corresponding to the prediction under the target historical item turnover prediction amount as the inventory support days corresponding to the quantile.

[0061] For example, the historical predicted quantity of goods turnover at the 0.1 quantile is 100 pieces, and the corresponding historical predicted quantity of goods turnover is 100 pieces, which represents the predicted quantity of goods turnover within 10 days. Therefore, the inventory support days at the 0.1 quantile is 10 days.

[0062] Step 204: Generate historical replenishment optimal decision information based on the above historical inventory decision feedback information.

[0063] In some embodiments, the execution entity may generate historical replenishment optimal decision information based on the historical inventory decision feedback information. The historical replenishment optimal decision information may be the optimal decision information related to replenishment at a historical time. For example, the historical replenishment optimal decision information may be optimal replenishment quantity information.

[0064] As an example, the historical replenishment optimal decision information may be the historical replenishment optimal replenishment quantity.

[0065] Exemplarily, the execution entity may generate historical replenishment optimal decision information based on the historical inventory decision feedback information, including:

[0066] The first step is to determine whether the inventory decision corresponding to the above historical inventory decision feedback information includes manual decision.

[0067] The second step, in response to the determination, includes determining the replenishment quantity corresponding to the manual decision as the manual decision replenishment quantity, wherein the replenishment quantity corresponding to the manual decision represents the increase in replenishment caused by the manual decision.

[0068] The third step is to determine the replenishment quantity under the model decision in the above inventory decision as the model decision replenishment quantity.

[0069] The fourth step is to add the manually decided replenishment quantity and the above model decision replenishment quantity to generate the historical replenishment optimal replenishment as the historical replenishment optimal decision information.

[0070] In some optional implementations of some embodiments, the above-mentioned historical replenishment optimal decision information is the historical replenishment optimal decision days. The historical replenishment optimal decision days may be the optimal days supporting replenishment processing under historical time. For example, the replenishment of 20 days of goods under historical time is supported. Here, 20 days is the historical replenishment optimal decision days.

[0071] Optionally, the execution entity may generate historical replenishment optimal decision information according to the historical inventory decision feedback information, including:

[0072] The first step is to determine whether the inventory decision corresponding to the above historical inventory decision feedback information includes a manual decision. The inventory decision may be a decision strategy for inventory adjustment in historical time. Manual decision is a decision strategy for inventory adjustment that is characterized by mandatory human participation. The inventory decision may include but is not limited to at least one of the following: model decision, manual decision. Model decision may be a decision strategy for inventory adjustment based on the output of a relevant model in historical time.

[0073] In a specific scenario, during the replenishment process, first, the inventory replenishment process is carried out according to the output of the relevant model (such as replenishment quantity, replenishment time, etc.) (i.e., corresponding to the model decision). Then, for emergencies such as sudden activities, the replenishment process carried out according to the model cannot effectively meet the current demand for goods, and human intervention is required to carry out the replenishment process (i.e., manual decision).

[0074] The second step, in response to the determination, includes determining the replenishment quantity corresponding to the manual decision, wherein the replenishment quantity corresponding to the manual decision represents the increase in replenishment caused by the manual decision.

[0075] The third step is to generate the inventory support days corresponding to the above replenishment quantity as the first inventory support days.

[0076] As an example, first, the execution entity may determine an average item turnover volume within 60 days before the historical time, and then divide the replenishment quantity by the average item turnover volume to generate the first inventory support days.

[0077] The fourth step is to determine the inventory support days under the model decision in the above inventory decision as the second inventory support days.

[0078] As an example, first, the execution subject may determine the replenishment quantity corresponding to the model dinner as the model decision replenishment quantity. Then, the model decision replenishment quantity is divided by the average item turnover volume within 60 days before the historical time to generate the second inventory support days.

[0079] Step 5: Generate the above-mentioned historical replenishment optimal decision days based on the above-mentioned first inventory support days and the above-mentioned second inventory support days.

[0080] As an example, the execution entity may add the first inventory support days and the second inventory support days to generate the historical replenishment optimal decision days.

[0081] As another example, first, the execution subject may determine a third inventory support days set in the same historical time period. Then, the first inventory support days and the second inventory support days are added to generate the added days. Finally, the added days are weighted summed with each third inventory support days in the third inventory support days set to generate the historical replenishment optimal decision days.

[0082] Step 205: Based on the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, the initial item turnover prediction generation model and the initial replenishment strategy information generation model are trained to generate an item turnover prediction generation model and a replenishment strategy information generation model.

[0083] In some embodiments, the execution entity may perform model training on the initial item predicted turnover generation model and the initial replenishment strategy information generation model based on the historical item turnover predictions at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, so as to generate an item predicted turnover generation model and a replenishment strategy information generation model.

[0084] In some optional implementations of some embodiments, the execution entity may perform model training on the initial item predicted turnover generation model and the initial replenishment strategy information generation model according to the historical item turnover prediction quantities at the respective quantiles, the inventory replenishment information at the respective quantiles, and the historical replenishment optimal decision information, so as to generate the item predicted turnover generation model and the replenishment strategy information generation model, including:

[0085] The first step is to match the inventory support days under the above-mentioned quantiles with the above-mentioned historical replenishment optimal decision days to generate the target prediction quantile.

[0086] As an example, first, the execution entity can select the inventory support days closest to the historical replenishment optimal decision days from the inventory support days under each quantile as the target inventory support days. Then, the quantile corresponding to the target inventory support days is used as the target prediction quantile.

[0087] The second step is to obtain the real inventory support days under the above-mentioned quantiles. The real inventory support days can be the real inventory support days in historical time.

[0088] The third step is to match the actual inventory support days under each of the above quantiles with the above historical replenishment optimal decision days to generate the target real quantile. The implementation method of the target real quantile can refer to the generation of the target prediction quantile.

[0089] The fourth step is to generate quantile loss information for the above target predicted quantile and the above target true quantile.

[0090] As an example, the execution entity may generate quantile loss information for the target predicted quantile and the target true quantile through a cross entropy loss function.

[0091] The fifth step is to generate item circulation loss information based on the actual amount of historical item circulation at the above-mentioned quantiles and the predicted amount of historical item circulation at the above-mentioned quantiles.

[0092] As an example, the execution entity may use a cross entropy loss function to generate item circulation loss information according to the actual amount of historical item circulation at each of the quantiles and the predicted amount of historical item circulation at each of the quantiles.

[0093] In the sixth step, based on the above-mentioned quantile loss information and the above-mentioned item circulation loss information, the above-mentioned initial item prediction circulation volume generation model and the initial replenishment strategy information generation model are trained to generate the item prediction circulation volume generation model and the replenishment strategy information generation model.

[0094] As an example, first, the weighted summation of the loss information of the quantile loss information and the item circulation loss information is performed to generate the first weighted sum loss information. Then, based on the first weighted sum loss information, the model training of the initial item predicted circulation volume generation model and the initial replenishment strategy information generation model is performed by back propagation to generate the item predicted circulation volume generation model and the replenishment strategy information generation model. Optionally, the execution subject may perform model training of the initial item predicted circulation volume generation model and the initial replenishment strategy information generation model based on the quantile loss information and the item circulation loss information to generate the item predicted circulation volume generation model and the replenishment strategy information generation model, including:

[0095] The first step is to obtain replenishment parameter data. The replenishment parameter data may be parameter data related to the replenishment operation. In practice, the replenishment parameter data may include but is not limited to at least one of the following: inventory constraint days, order cycle, supplier delivery time, and historical inventory turnover data for the same period. For example, the historical inventory turnover data for the same period may be the turnover days for the same period. The inventory constraint days represent the maximum and minimum values ​​of the total inventory support days.

[0096] The second step is to generate the total inventory forecast based on the above replenishment parameter data and the historical item turnover forecast at each of the above quantiles.

[0097] As an example, first, the execution entity may multiply the order cycle by a value of 0.5 to generate first multiplication information. Then, the historical inventory turnover data for the same period is added to the first multiplication information to generate first addition information. Next, the supplier delivery time and the order cycle are added to generate second addition information. The maximum value of the first addition information and the second addition information is determined as the total inventory support days. Finally, the item turnover volume with the same item turnover time as the above-mentioned total inventory support days is selected from the historical item turnover volume at each percentile as the total inventory forecast amount.

[0098] The third step is to generate inventory loss information for the above total inventory forecast.

[0099] As an example, first, the actual total inventory amount is obtained, and then the target loss function is used to determine the inventory loss information for the above-mentioned predicted total inventory amount and the actual total inventory amount.

[0100] The fourth step is to train the initial item predicted turnover generation model and the initial replenishment strategy information generation model based on the above-mentioned percentile loss information, the above-mentioned item circulation loss information and the above-mentioned inventory loss information to generate an item predicted turnover generation model and a replenishment strategy information generation model.

[0101] As an example, first, the execution subject may perform weighted summation of the quantile loss information, the item circulation loss information, and the inventory loss information to generate second weighted sum loss information. Then, based on the second weighted sum loss information, the initial item forecast circulation volume generation model and the initial replenishment strategy information generation model are trained by back propagation to generate an item forecast circulation volume generation model and a replenishment strategy information generation model.

[0102] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, using the item prediction turnover generation model and the replenishment strategy information generation model, the item prediction turnover and replenishment strategy information can be accurately generated to achieve accurate replenishment of inventory items. Specifically, the reason why the relevant replenishment strategy information is not accurate enough is that there is uncertainty in the replenishment strategy generation model, and there are various uncertainties in demand, and turnover deterioration or spot rate deterioration often occurs. Replenishment processing is carried out according to the replenishment strategy output by the replenishment strategy generation model, resulting in inaccurate replenishment. Based on this, the model training method of some embodiments of the present disclosure first obtains the first historical inventory item data and the historical inventory decision feedback information. Here, the first historical inventory item data obtained is the inventory data related to the inventory item, which is used as the data basis to facilitate the generation of the historical item turnover prediction at each subsequent quantile. The historical inventory decision feedback information obtained can fully reflect the specific inventory feedback situation under the execution of the historical inventory decision. Through the inventory feedback, we can fully understand whether the replenishment quantity generated by the model is insufficient or excessive in the historical time, so as to adjust and solve the problem in the next inventory decision, so as to continuously improve the accuracy of the replenishment quantity output by the model, so that the subsequent feedback information of the historical inventory decision can be based on the training process of the initial item prediction turnover generation model and the initial replenishment strategy information generation model, and fully consider the problems existing in the historical decision, so as to obtain a more accurate model for subsequent output. Then, according to the above-mentioned first historical inventory item data, the initial item prediction turnover generation model can be used to accurately generate the historical item turnover prediction quantity at each quantile. Here, by determining the historical item turnover prediction quantity at each quantile, the circulation distribution of the item turnover prediction quantity at the predetermined historical time can be more fully reflected. In addition, it is convenient to determine the inventory replenishment information at each quantile in the future. Furthermore, based on the historical predicted item turnover at each quantile, the inventory replenishment information at each quantile can be accurately generated, so as to accurately generate model training data for the initial item predicted turnover generation model and the initial replenishment strategy information generation model training with the historical predicted item turnover at each quantile. Next, based on the historical inventory decision feedback information, the historical replenishment optimal decision information is generated as the strategy label of the optimal strategy for replenishment in historical time, which is convenient for subsequent model training. Finally, based on the historical predicted item turnover at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, the initial item predicted turnover generation model and the initial replenishment strategy information generation model are trained to generate the item predicted turnover generation model and the replenishment strategy information generation model.In summary, based on the historical inventory decision feedback information, as well as based on the generated historical item turnover predictions at each quantile and the inventory replenishment information at each quantile, the initial item prediction turnover generation model and the initial replenishment strategy information generation model can be effectively and accurately trained to obtain a more accurate item prediction turnover generation model and a replenishment strategy information generation model.

[0103] Further references Figure 3 , shows a process 300 of other embodiments of the model training method according to the present disclosure. The model training method comprises the following steps:

[0104] Step 301, obtaining first historical inventory item data and historical inventory decision feedback information.

[0105] Step 302: preprocess the first historical inventory item data to generate preprocessed data.

[0106] In some embodiments, an execution entity (e.g. Figure 1 The electronic device 101 shown can perform data preprocessing on the above-mentioned first historical inventory item data to generate preprocessed data.

[0107] Among them, data preprocessing may include but is not limited to at least one of the following: data integration at SKU granularity based on the use of data association technology, aggregation technology, etc., data outlier cleaning, data feature engineering, converting data to a specified range to transform the overall mean and standard deviation, normal standardization, Robust standardization, and data standardization based on quartiles.

[0108] Step 303: input the preprocessed data into the initial coding layer to generate coding information.

[0109] In some embodiments, the execution entity may input the preprocessed data into the initial encoding layer to generate encoding information. The initial item predicted turnover generation model includes: an initial encoding layer and an initial decoding layer. The initial encoding layer may be an encoding layer for data encoding that has not yet been trained. The initial decoding layer may be a decoding layer for vector decoding that has not yet been trained. For example, the encoding layer may be an encoding network based on LSTM. The decoding layer may be a decoding network based on LSTM.

[0110] In some optional implementations of some embodiments, the initial coding layer includes: an initial recent item turnover coding layer, an initial flow coding layer, an initial value information coding layer, and an initial concurrent item turnover coding layer. The initial recent item turnover coding layer may be a network layer that has not yet been trained and is used to encode and process recent item turnover related data. The recent item turnover related data may be data related to item turnover in a recent period. In the field of logistics, the recent item turnover related data may be data related to item sales in a recent period. In practice, the recent item turnover related data may include: recent sales, time information, and item information. The initial flow coding layer may be a network layer that has not yet been trained and is used to encode and process flow data. Flow data may be data related to page clicks. The initial value information coding layer may be a network layer that has not yet been trained and is used to encode and process value information. In practice, in the field of logistics, value information may be price. The initial concurrent item turnover coding layer may be a network layer that has not yet been trained and is used to encode and process concurrent item turnover related data. The data related to the item turnover volume in the same period may be data related to the item turnover volume in the same period. The same period may be within the same date. For example, the current time period is from January 1, 2011 to January 31, 2011. The corresponding time period in the same period may be from January 1, 2012 to January 31, 2012. The corresponding time period in the same period may also be from January 1, 2013 to January 31, 2013. The initial recent item turnover volume encoding layer, the initial flow rate encoding layer, the initial value information encoding layer, and the initial item turnover encoding layer in the same period may all be network layers of the Transformer structure.

[0111] Optionally, the execution subject may input the preprocessed data into the initial coding layer to generate coding information, which may include the following steps:

[0112] In the first step, the recent item circulation data included in the preprocessed data is input into the initial recent item circulation amount coding layer to generate recent item circulation coding information, wherein the recent item circulation coding information may be data in vector form and of a predetermined length.

[0113] The second step is to input the traffic data included in the preprocessed data into the initial traffic coding layer to generate traffic coding information, wherein the traffic coding information may be data in a vector form with a predetermined length.

[0114] The third step is to input the value data included in the preprocessed data into the initial value information encoding layer to generate value encoding information, wherein the value encoding information can be data in vector form with a predetermined length.

[0115] The fourth step is to input the contemporaneous item flow data included in the preprocessed data into the initial contemporaneous item flow coding layer to generate contemporaneous item flow coding information, wherein the contemporaneous item flow coding information can be data in vector form with a predetermined length.

[0116] The fifth step is to generate the above coding information based on the above recent item circulation coding information, the above flow coding information, the above value coding information and the above contemporaneous item circulation coding information.

[0117] As an example, the execution entity may directly perform vector concatenation on the recent item circulation coding information, the flow coding information, the value coding information and the item circulation coding information of the same period to generate coding information.

[0118] As another example, the execution subject may directly input the recent item circulation coding information, the flow coding information, the value coding information, and the item circulation coding information of the same period into the item coding layer to generate coding information. The item coding layer may be a network layer that aggregates and codes item-related data. The item coding layer may be a network layer of a Transformer structure.

[0119] Step 304: input the above-mentioned coded information into the above-mentioned initial decoding layer to generate the historical item flow prediction quantity at the above-mentioned each quantile.

[0120] In some embodiments, the execution entity may input the encoded information into the initial decoding layer to generate the historical item flow prediction at each quantile.

[0121] Step 305: Generate inventory replenishment information at each of the above quantiles based on the historical item turnover forecast at each of the above quantiles.

[0122] Step 306: Generate historical replenishment optimal decision information based on the above historical inventory decision feedback information.

[0123] Step 307: Based on the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, the initial item turnover prediction generation model and the initial replenishment strategy information generation model are trained to generate an item turnover prediction generation model and a replenishment strategy information generation model.

[0124] In some embodiments, the specific implementation of steps 301, 305-307 and the technical effects thereof can be referred to in Figure 2 The steps 201, 203-205 in the corresponding embodiment are not described in detail here.

[0125] from Figure 3 It can be seen that Figure 2 Compared with the description of some corresponding embodiments, Figure 3 The process 300 of the model training method in some corresponding embodiments can accurately generate the predicted amount of item circulation at each quantile through the encoding layer and the decoding layer.

[0126] Continue to refer Figure 4 , shows a process 400 of some embodiments of the inventory item replenishment method according to the present disclosure. The inventory item replenishment method includes the following steps:

[0127] Step 401, obtaining second historical inventory item data.

[0128] In some embodiments, the execution subject (e.g., electronic device) of the above inventory item replenishment method may obtain the second historical inventory item data by wired or wireless means. The second historical inventory item data may be inventory item data within a predetermined historical time period before the current time. For example, the current time is July 1. The corresponding second historical inventory item data may be inventory item data within June 1 to June 30.

[0129] Step 402: Based on the second historical inventory item data, a pre-trained item predicted turnover volume generation model is used to generate the predicted future item turnover volume at each quantile.

[0130] In some embodiments, the execution subject may generate the predicted future item turnover at each quantile based on the second historical inventory item data using a pre-trained item prediction turnover generation model. The predicted item turnover generation model is generated based on a model training method. The predicted future item turnover may be the predicted item turnover in a future time period. In practice, for logistics scenarios, the predicted future item turnover may be the future sales volume at each quantile.

[0131] As an example, first, the execution subject may perform data preprocessing on the second historical inventory item data to obtain preprocessed data, and then input the preprocessed data into a pre-trained item forecast turnover generation model to generate future item turnover forecasts at each quantile.

[0132] Step 403: Generate future replenishment strategy information based on the predicted future item turnover at each of the above quantiles using a pre-trained replenishment strategy information generation model.

[0133] In some embodiments, the execution subject may generate future replenishment strategy information based on the predicted future item turnover under each quantile, using a pre-trained replenishment strategy information generation model. The future replenishment strategy information includes: future optimal quantile. The item predicted turnover generation model is generated based on a model training method. The future replenishment strategy information may be strategy information that characterizes the optimal strategy for replenishment in a future time period. The future optimal quantile may be the optimal quantile for replenishment processing in a future time period. Based on the future optimal quantile, accurate and effective control of inventory can be achieved.

[0134] Specifically, the future replenishment strategy information also includes: future optimal item turnover.

[0135] As an example, the above-mentioned execution entity can directly input the predicted future item turnover quantity at each quantile into the replenishment strategy information generation model to generate future replenishment strategy information.

[0136] Step 404: Replenishment processing is performed on the inventory items according to the above future optimal quantile.

[0137] In some embodiments, the execution entity may execute replenishment processing for inventory items according to the future optimal quantile.

[0138] In some optional implementations of some embodiments, the execution entity may perform replenishment processing for inventory items according to the future optimal quantile, including:

[0139] The first step is to determine the future optimal replenishment quantity corresponding to the above future optimal quantile.

[0140] As an example, first, the execution subject may determine the future optimal item turnover corresponding to the future optimal quantile. Then, based on the future optimal item turnover, determine the future optimal replenishment quantity. Second, based on the future optimal replenishment quantity, perform replenishment processing for the inventory items.

[0141] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: through the inventory item replenishment method of some embodiments of the present disclosure, using the item forecast turnover generation model and the replenishment strategy information generation model, the item forecast turnover and replenishment strategy information can be accurately generated to utilize the future optimal quantile point to achieve accurate replenishment of inventory items.

[0142] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a model training device, and these device embodiments are Figure 2 Corresponding to the method embodiments shown, the model training device can be specifically applied to various electronic devices.

[0143] like Figure 5 As shown, a model training device 500 includes: a first acquisition unit 501, a first generation unit 502, a second generation unit 503, a third generation unit 504 and a training unit 505. The first acquisition unit 501 is configured to acquire the first historical inventory item data and the historical inventory decision feedback information; the first generation unit 502 is configured to generate the historical item turnover prediction quantity at each quantile according to the first historical inventory item data and using the initial item prediction turnover quantity generation model; the second generation unit 503 is configured to generate the inventory replenishment information at each quantile according to the historical item turnover prediction quantity at each quantile; the third generation unit 504 is configured to generate the historical replenishment optimal decision information according to the historical inventory decision feedback information; the training unit 505 is configured to perform model training on the initial item prediction turnover quantity generation model and the initial replenishment strategy information generation model according to the historical item turnover prediction quantity at each quantile, the inventory replenishment information at each quantile and the historical replenishment optimal decision information, so as to generate the item prediction turnover quantity generation model and the replenishment strategy information generation model.

[0144] In some optional implementations of some embodiments, the above-mentioned initial item predicted turnover generation model includes: an initial encoding layer and an initial decoding layer; and the first generation unit 502 can be further configured to: perform data preprocessing on the above-mentioned first historical inventory item data to generate preprocessed data; input the above-mentioned preprocessed data into the above-mentioned initial encoding layer to generate encoding information; input the above-mentioned encoding information into the above-mentioned initial decoding layer to generate the historical item turnover prediction quantity under the above-mentioned each quantile.

[0145] In some optional implementations of some embodiments, the above-mentioned initial coding layer includes: an initial recent item turnover volume coding layer, an initial flow coding layer, an initial value information coding layer and an initial concurrent item turnover coding layer; and the first generating unit 502 can be further configured to: input the recent item turnover data included in the above-mentioned preprocessing data into the above-mentioned initial recent item turnover volume coding layer to generate recent item turnover coding information; input the flow data included in the above-mentioned preprocessing data into the above-mentioned initial flow coding layer to generate flow coding information; input the value data included in the above-mentioned preprocessing data into the above-mentioned initial value information coding layer to generate value coding information; input the concurrent item turnover data included in the above-mentioned preprocessing data into the above-mentioned initial concurrent item turnover coding layer to generate concurrent item turnover coding information; generate the above-mentioned coding information according to the above-mentioned recent item turnover coding information, the above-mentioned flow coding information, the above-mentioned value coding information and the above-mentioned concurrent item turnover coding information.

[0146] In some optional implementations of some embodiments, the above-mentioned inventory replenishment information is the inventory support days; and the second generation unit 503 can be further configured to: determine and generate the inventory support days at the above-mentioned each quantile according to the historical item turnover forecast quantity at the above-mentioned each quantile.

[0147] In some optional implementations of some embodiments, the above-mentioned historical replenishment optimal decision information is the historical replenishment optimal decision days; and the third generation unit 504 can be further configured to: determine whether the inventory decision corresponding to the above-mentioned historical inventory decision feedback information includes a manual decision; in response to the determination that it includes, determine the replenishment quantity corresponding to the above-mentioned manual decision; generate the inventory support days corresponding to the above-mentioned replenishment quantity as the first inventory support days; determine the inventory support days under the model decision in the above-mentioned inventory decision as the second inventory support days; generate the above-mentioned historical replenishment optimal decision days according to the above-mentioned first inventory support days and the above-mentioned second inventory support days.

[0148] In some optional implementations of some embodiments, the training unit 505 may be further configured to: match the inventory support days under the above-mentioned each quantile with the above-mentioned historical replenishment optimal decision days in terms of days to generate a target prediction quantile; obtain the actual inventory support days under the above-mentioned each quantile; match the actual inventory support days under the above-mentioned each quantile with the above-mentioned historical replenishment optimal decision days in terms of days to generate a target real quantile; generate quantile loss information for the above-mentioned target prediction quantile and the above-mentioned target real quantile; generate item circulation loss information according to the historical item circulation actual quantity under the above-mentioned each quantile and the historical item circulation predicted quantity under the above-mentioned each quantile; perform model training on the above-mentioned initial item prediction circulation quantity generation model and the initial replenishment strategy information generation model according to the above-mentioned quantile loss information and the above-mentioned item circulation loss information to generate an item prediction circulation quantity generation model and a replenishment strategy information generation model.

[0149] In some optional implementations of some embodiments, the training unit 505 may be further configured to: obtain replenishment parameter data; generate a total inventory forecast based on the replenishment parameter data and the historical item turnover forecasts at each quantile; generate inventory loss information for the total inventory forecast; perform model training on the initial item forecast turnover generation model and the initial replenishment strategy information generation model based on the quantile loss information, the item turnover loss information, and the inventory loss information, so as to generate an item forecast turnover generation model and a replenishment strategy information generation model.

[0150] It is understandable that the units recorded in the model training device 500 are similar to the reference Figure 2Therefore, the operations, features and beneficial effects described above for the method are also applicable to the model training device 500 and the units contained therein, and will not be repeated here.

[0151] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a device for replenishing inventory items. These device embodiments are similar to Figure 4 Corresponding to the method embodiments shown, the inventory item replenishment device can be specifically applied to various electronic devices.

[0152] like Figure 6 As shown, a stock item replenishment device 600 includes: a second acquisition unit 601, a fourth generation unit 602, a fifth generation unit 603 and an execution unit 604. The second acquisition unit 601 is configured to acquire second historical stock item data; the fourth generation unit 602 is configured to generate future item turnover predictions at each quantile using a pre-trained item turnover prediction generation model based on the second historical stock item data, wherein the item turnover prediction generation model is generated based on a model training method; the fifth generation unit 603 is configured to generate future replenishment strategy information based on the future item turnover predictions at each quantile using a pre-trained replenishment strategy information generation model, wherein the future replenishment strategy information includes: a future optimal quantile, wherein the item turnover prediction generation model is generated based on a model training method; the execution unit 604 is configured to perform replenishment processing for the stock items according to the future optimal quantile.

[0153] It is understood that the units described in the inventory replenishment device 600 are similar to those described in the reference Figure 4 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the inventory replenishment device 600 and the units included therein, and will not be described in detail here.

[0154] Reference below Figure 7 , which shows an electronic device (eg, Figure 1 Schematic diagram of the structure of the electronic device 101)700. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0155] like Figure 7As shown, the electronic device 700 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory 702 or a program loaded from a storage device 708 into a random access memory 703. Various programs and data required for the operation of the electronic device 700 are also stored in the random access memory 703. The processing device 701, the read-only memory 702, and the random access memory 703 are connected to each other via a bus 704. An input / output interface 705 is also connected to the bus 704.

[0156] Typically, the following devices may be connected to the input / output interface 705: an input device 706 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 707 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 708 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 709. The communication device 709 may allow the electronic device 700 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 7 The electronic device 700 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 7 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0157] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from a network through a communication device 709, or installed from a storage device 708, or installed from a read-only memory 702. When the computer program is executed by the processing device 701, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0158] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0159] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0160] The computer-readable medium may be included in the electronic device; or it may exist independently without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains the first historical inventory item data and the historical inventory decision feedback information; generates the historical item turnover prediction quantity at each quantile using the initial item prediction turnover quantity generation model according to the first historical inventory item data; generates the inventory replenishment information at each quantile according to the historical item turnover prediction quantity at each quantile; generates the historical replenishment optimal decision information according to the historical inventory decision feedback information; performs model training on the initial item prediction turnover quantity generation model and the initial replenishment strategy information generation model according to the historical item turnover prediction quantity at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, so as to generate the item prediction turnover quantity generation model and the replenishment strategy information generation model. Acquire second historical inventory item data; generate future item turnover predictions at each quantile using a pre-trained item turnover prediction generation model based on the second historical inventory item data, wherein the item turnover prediction generation model is generated based on a model training method; generate future replenishment strategy information using a pre-trained replenishment strategy information generation model based on the future item turnover predictions at each quantile, wherein the future replenishment strategy information includes: a future optimal quantile, wherein the item turnover prediction generation model is generated based on a model training method; perform replenishment processing on the inventory items based on the future optimal quantile.

[0161] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0162] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0163] The units described in some embodiments of the present disclosure may be implemented by software or hardware. The units described may also be set in a processor, for example, it may be described as: a processor includes a first acquisition unit, a first generation unit, a second generation unit, a third generation unit and a training unit. The names of these units do not constitute a limitation on the units themselves in some cases. For example, the first acquisition unit may also be described as a "unit for acquiring first historical inventory item data and historical inventory decision feedback information".

[0164] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0165] Some embodiments of the present disclosure also provide a computer program product, including a computer program, which implements any of the above-mentioned model training methods or inventory item replenishment methods when executed by a processor.

[0166] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A model training method, comprising: Obtaining first historical inventory item data and historical inventory decision feedback information; Based on the first historical inventory item data, using the initial item predicted turnover volume generation model, generate the historical item turnover prediction volume at each quantile; Generate inventory replenishment information at each quantile according to the historical item turnover forecast at each quantile; Generate historical replenishment optimal decision information based on the historical inventory decision feedback information; According to the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile and the historical replenishment optimal decision information, model training is performed on the initial item prediction turnover quantity generation model and the initial replenishment strategy information generation model to generate an item prediction turnover quantity generation model and a replenishment strategy information generation model.

2. The method according to claim 1, wherein: The initial item predicted turnover generation model includes: an initial encoding layer and an initial decoding layer; and The method of generating the historical item turnover prediction amount at each quantile by using the initial item turnover prediction model according to the first historical inventory item data includes: performing data preprocessing on the first historical inventory item data to generate preprocessed data; Inputting the preprocessed data into the initial coding layer to generate coding information; The encoded information is input into the initial decoding layer to generate the historical item flow prediction quantity at each quantile.

3. The method according to claim 2, wherein: The initial coding layer includes: an initial recent item circulation volume coding layer, an initial flow coding layer, an initial value information coding layer and an initial item circulation coding layer in the same period; and The step of inputting the preprocessed data into the initial coding layer to generate coding information comprises: Inputting the recent item circulation data included in the preprocessing data into the initial recent item circulation amount coding layer to generate recent item circulation coding information; Inputting the flow data included in the preprocessing data into the initial flow coding layer to generate flow coding information; Inputting the value data included in the preprocessed data into the initial value information encoding layer to generate value encoding information; Inputting the contemporaneous item circulation data included in the preprocessing data into the initial contemporaneous item circulation coding layer to generate contemporaneous item circulation coding information; The coding information is generated according to the recent item circulation coding information, the flow coding information, the value coding information and the item circulation coding information of the same period.

4. The method according to claim 1, wherein: The inventory replenishment information is the inventory support days; as well as Generating the inventory replenishment information at each quantile according to the historical item turnover forecast at each quantile includes: According to the historical predicted quantity of item turnover at each quantile, the number of days of inventory support at each quantile is determined and generated.

5. The method according to claim 4, wherein: The historical replenishment optimal decision information is the historical replenishment optimal decision days; as well as The generating of historical replenishment optimal decision information according to the historical inventory decision feedback information includes: Determining whether the inventory decision corresponding to the historical inventory decision feedback information includes a manual decision; In response to the determination, the method includes determining a replenishment quantity corresponding to the manual decision; Generate the inventory support days corresponding to the replenishment quantity as the first inventory support days; Determine the inventory support days under the model decision in the inventory decision as the second inventory support days; The historical replenishment optimal decision days are generated according to the first inventory support days and the second inventory support days.

6. The method according to claim 5, wherein: The performing of model training on the initial item predicted turnover generation model and the initial replenishment strategy information generation model according to the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information to generate the item predicted turnover generation model and the replenishment strategy information generation model, including: Matching the inventory support days under each quantile with the historical replenishment optimal decision days in terms of days to generate a target forecast quantile; Get the actual inventory support days at each quantile; Matching the actual inventory support days under each quantile with the historical replenishment optimal decision days in terms of days to generate a target actual quantile; Generating quantile loss information for the target predicted quantile and the target true quantile; Generate item circulation loss information according to the actual amount of historical item circulation at each quantile and the predicted amount of historical item circulation at each quantile; According to the quantile loss information and the item circulation loss information, the initial item predicted circulation volume generation model and the initial replenishment strategy information generation model are trained to generate an item predicted circulation volume generation model and a replenishment strategy information generation model.

7. The method according to claim 6, wherein: The performing of model training on the initial item predicted circulation volume generation model and the initial replenishment strategy information generation model according to the quantile loss information and the item circulation loss information to generate the item predicted circulation volume generation model and the replenishment strategy information generation model includes: Get replenishment parameter data; Generate a total inventory forecast based on the replenishment parameter data and the historical item turnover forecast at each quantile; generating inventory loss information for the total inventory forecast; According to the quantile loss information, the item circulation loss information and the inventory loss information, the initial item forecast circulation volume generation model and the initial replenishment strategy information generation model are trained to generate an item forecast circulation volume generation model and a replenishment strategy information generation model.

8. A method for replenishing inventory items, comprising: Obtaining second historical inventory item data; According to the second historical inventory item data, using a pre-trained item prediction turnover generation model, the future item turnover prediction at each quantile is generated, wherein the item prediction turnover generation model is generated based on the method according to any one of claims 1 to 7; According to the predicted future item turnover at each quantile, using a pre-trained replenishment strategy information generation model, future replenishment strategy information is generated, wherein the future replenishment strategy information includes: future optimal quantiles, and the item predicted turnover generation model is generated based on the method of any one of claims 1 to 7; A replenishment process for the inventory items is performed according to the future optimal quantile.

9. The method according to claim 8, wherein: The performing replenishment processing for the inventory items according to the future optimal quantile includes: Determine the future optimal replenishment quantity corresponding to the future optimal quantile; A replenishment process for the inventory items is performed according to the future optimal replenishment quantity.

10. A model training device, comprising: A first acquisition unit is configured to acquire first historical inventory item data and historical inventory decision feedback information; A first generating unit is configured to generate a historical item turnover prediction amount at each quantile based on the first historical inventory item data and using an initial item turnover prediction generation model; A second generating unit is configured to generate inventory replenishment information at each quantile according to the historical item flow forecast quantity at each quantile; A third generating unit is configured to generate historical replenishment optimal decision information according to the historical inventory decision feedback information; The training unit is configured to perform model training on the initial item predicted turnover generation model and the initial replenishment strategy information generation model according to the historical item turnover prediction quantities at each quantile, the inventory replenishment information at each quantile, and the historical replenishment optimal decision information, so as to generate the item predicted turnover generation model and the replenishment strategy information generation model.

11. A device for replenishing inventory items, comprising: a second acquisition unit configured to acquire second historical inventory item data; a fourth generating unit configured to generate the predicted future item turnover at each quantile based on the second historical inventory item data and using a pre-trained item turnover prediction generation model, wherein the item turnover prediction generation model is generated based on the method according to any one of claims 1 to 7; a fifth generating unit, configured to generate future replenishment strategy information according to the predicted future item turnover at each quantile using a pre-trained replenishment strategy information generation model, wherein the future replenishment strategy information includes: a future optimal quantile, and the item predicted turnover generation model is generated based on the method of any one of claims 1 to 7; The execution unit is configured to execute a replenishment process for the inventory items according to the future optimal quantile.

12. An electronic device comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 9.

13. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

14. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images