Commodity management and control method and device based on large model decision, equipment and medium

Abstract

The application discloses a commodity management method and device based on a large model decision, equipment and a medium, relates to the technical field of commodity management, and comprises the following steps: constructing a commodity state feature vector for the same batch of commodities in the vending machine based on the latest on-sale commodity information and the full-link inventory state information of the vending machine in a preset area; the commodity state feature vector comprises the on-sale inventory of the same batch of commodities in a single vending machine, the standard shelf life, the production date and the remaining shelf life; inputting the commodity state feature vector into a commodity decision model obtained by training a decision large model based on a Transformer architecture by using historical on-sale commodity information and historical full-link inventory state information of different vending machines, outputting a commodity management decision based on the feature vector, and performing a corresponding management operation based on the commodity management decision. The application can effectively manage commodities in different inventory states and on-sale states.

Description

Technical Field

[0001] This application relates to the field of commodity management technology, and in particular to commodity control methods, devices, equipment and media based on large model decision-making. Background Technology

[0002] Currently, vending machines are commonly deployed in shopping malls, convenience stores, train stations, gyms, swimming pools, stadiums, and other venues to meet customers' needs for convenience and emergency consumption in different scenarios. However, controlling the shelf life of short-shelf-life products (such as yogurt, fresh milk, bread, and snacks) in vending machines currently faces the following technical challenges:

[0003] 1. Unreliable data source and broken supply chain: Automated decision-making relies on the "product production date," which may be inaccurate at the time of procurement and warehousing, and may be subject to errors, loss, or tampering in subsequent warehousing, distribution, and other stages. Furthermore, traditional shelf-life control methods cannot guarantee data consistency across the entire chain from "warehouse inventory" to "in-transit inventory" and then to "vending machine inventory," easily leading to a complete disconnect between "system accounts (i.e., accounting records)" and "physical goods (i.e., actual inventory)."

[0004] 2. Rigid decision-making logic and disconnect from the scenario: Current automated management methods (such as pre-setting the number of days before the expiration date for product delisting) use static rules. Their decision-making logic cannot make differentiated decisions based on the sales characteristics of different products, and it cannot respond in a timely manner to dynamic business needs such as weekend promotions and holiday stocking. This leads to a disconnect between business decisions and actual sales, which can easily cause product / inventory losses or missed sales opportunities.

[0005] 3. Insufficient level of intelligence and reliance on human experience: For complex near-expiry product handling strategies, such as when to promote, what discount to offer, and whether to remove products from shelves, the strategies rely heavily on the experience of operations personnel, making it difficult to achieve standardization and automation. Summary of the Invention

[0006] In view of this, the purpose of this application is to provide a product management method, device, equipment, and medium based on large-scale model decision-making, which can more accurately and effectively manage products in different inventory and sales states, while reducing labor and time costs, making operations management easier and more efficient. The specific solution is as follows:

[0007] Firstly, this application discloses a commodity control method based on large-scale model decision-making, applied to a cloud server, including:

[0008] Obtain the latest information on products available for sale and the current end-to-end inventory status of vending machines within a preset area; the latest information on products available for sale includes batch information of the products in the vending machines, and the batch information includes the production date of the products and the channel identifier associated with the production date;

[0009] Based on the latest information on products for sale and the end-to-end inventory status information, a product status feature vector is constructed for the same batch of products in the vending machine; the product status feature vector includes the single-machine inventory, standard shelf life, production date, and remaining shelf life of the same batch of products in the vending machine.

[0010] The product status feature vector is input into the trained product decision model; the product decision model is a model obtained by training a large decision model based on the Transformer architecture using historical product information and historical full-link inventory status information of different vending machines.

[0011] Obtain the target product control decision output by the product decision model based on the product state feature vector, and execute the control operation corresponding to the product in the vending machine based on the target product control decision.

[0012] Optionally, before obtaining the latest information on products available for sale and the current end-to-end inventory status information of vending machines within a preset area, the method further includes:

[0013] When it is detected that goods have been procured and entered the central warehouse, the physical packaging information of the goods is obtained;

[0014] Based on the physical packaging information, the batch information of the goods in the same batch is entered into the purchase receipt in the warehouse management system; the batch information includes the production date of the goods.

[0015] Optionally, the commodity control method based on large model decision-making further includes:

[0016] When the warehouse manager sends a pick-up and outbound receipt after picking is completed, the current warehouse inventory information and the product inventory information in the current in-transit inventory information are updated atomically through database transactions to obtain the latest warehouse inventory information and the updated in-transit inventory information; the product inventory information includes batch information of the same batch of products;

[0017] When a delivery receipt is received from the delivery person after delivery is completed, the current inventory information for sale and the product inventory information in the updated in-transit inventory information of the vending machine are atomically updated through a database transaction to obtain the latest inventory information for sale and the latest inventory information in transit; the latest inventory information for sale includes the product identifier and the corresponding channel identifier;

[0018] The latest available inventory information is sent to the vending machine to bind the production date of the products on different aisles of the vending machine with the aisle of the product based on the latest available inventory information, so as to obtain the latest available product information of the vending machine.

[0019] Optionally, the end-to-end inventory status information includes the latest warehouse inventory information, the latest in-transit inventory information, and the latest on-sale inventory information, and the end-to-end inventory status information is located in a preset database;

[0020] Accordingly, constructing a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information includes:

[0021] Based on the latest available inventory information, obtain the single-machine available inventory of the same batch of goods in a single vending machine, and statistically analyze the single-machine available inventory of all vending machines in the preset area to obtain the total available inventory of the same batch.

[0022] Construct a product status feature vector for the same batch of products in the vending machine, which includes the single machine's inventory, the standard shelf life, the production date, the remaining shelf life, and the total inventory of the same batch.

[0023] Optionally, the target product control decision is output based on the product status feature vector, including:

[0024] Determine whether the stand-alone inventory in the product status feature vector is less than a first preset inventory threshold. If the stand-alone inventory is less than the first preset inventory threshold, generate a target product management decision to replenish the product at the target channel corresponding to the channel identifier.

[0025] If the inventory of a single unit is not less than the first preset inventory threshold, then calculate the ratio of the remaining shelf life to the standard shelf life, and determine whether the ratio is less than the preset product health threshold.

[0026] If the ratio is less than the preset product health threshold, a target product control decision is generated to remove the product from the target channel corresponding to the channel identifier.

[0027] Accordingly, the execution of control operations corresponding to the goods in the vending machine based on the target goods control decision includes:

[0028] The target product control decision is sent to the corresponding delivery terminal and / or the vending machine, so as to replenish or remove the product in the corresponding aisle of the vending machine according to the target product control decision.

[0029] Optionally, the target product control decision is output based on the product status feature vector, including:

[0030] If the inventory of a single unit for sale is not less than the first preset inventory threshold, then it is determined whether the total inventory of the same batch for sale exceeds the second preset inventory threshold; the second preset inventory threshold is greater than the first preset inventory threshold.

[0031] If the total inventory of the same batch exceeds the second preset inventory threshold, a target product control decision is generated to discount the current batch of products according to the first discount rate.

[0032] Optionally, constructing a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information includes:

[0033] The number of days that the same batch of products are available on the shelves is determined based on the product listing time in the latest product information.

[0034] Construct a product status feature vector for the same batch of products in the vending machine, including the single machine's inventory, standard shelf life, production date, remaining shelf life, and the number of days on the shelf;

[0035] Accordingly, based on the product status feature vector, the target product control decision is output, including:

[0036] If the ratio is not less than the preset product health threshold, then it is determined whether the number of days on the shelf exceeds the preset unsold period.

[0037] If the number of days on the shelf exceeds the preset slow-moving period, a target product management decision is generated to discount the current batch of products according to the second discount rate.

[0038] Secondly, this application discloses a commodity control device based on large-scale model decision-making, applied to a cloud server, comprising:

[0039] The acquisition module is used to acquire the latest information on products available for sale and the current end-to-end inventory status information of vending machines within a preset area; the latest information on products available for sale includes batch information of products in the vending machine, and the batch information includes the production date of the product and the channel identifier bound to the production date;

[0040] The vector creation module is used to construct a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information; the product status feature vector includes the single-machine on-sale inventory, standard shelf life, production date and remaining shelf life of the same batch of products in the vending machine.

[0041] The input module is used to input the product state feature vector into the trained product decision model; the product decision model is a model obtained by training a large decision model based on the Transformer architecture using historical product information and historical full-link inventory status information of different vending machines.

[0042] The product control module is used to obtain the target product control decision output by the product decision model based on the product state feature vector, and to execute the control operation corresponding to the product in the vending machine based on the target product control decision.

[0043] Thirdly, this application discloses an electronic device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the aforementioned commodity control method based on large model decision-making.

[0044] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned commodity control method based on large model decision-making.

[0045] As can be seen, this application is applied to a cloud server. First, it obtains the latest information on products for sale and the current end-to-end inventory status information of vending machines within a preset area. Then, based on the latest information on products for sale and the end-to-end inventory status information, it constructs a product status feature vector for the same batch of products in the vending machine (including the single-machine inventory, standard shelf life, production date, and remaining shelf life of the same batch of products in a single vending machine). The product status feature vector is then input into a product decision model based on the Transformer architecture. The target product control decision output by the product decision model based on the product status feature vector is then obtained, and control operations corresponding to the products in the vending machine are executed based on the target product control decision. This application achieves automated management and control of goods in vending machines through a cloud server. First, based on the latest available product information and current end-to-end inventory status information of the vending machine, a product status feature vector is constructed for each batch of products. Then, a product decision model based on the Transformer architecture is used to comprehensively analyze the product status feature vectors, thereby obtaining corresponding product decision results. Based on these product decision results, corresponding management and control operations are executed. By considering multi-dimensional information (such as the latest available product information and end-to-end inventory status information) and making comprehensive decisions through a decision model, more accurate and effective management and control of products in different inventory and available-to-sale states can be achieved. Furthermore, automated product management and control is realized through a cloud server, reducing labor and time costs and making operation and management easier and more efficient. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0047] Figure 1 This application discloses a flowchart of a commodity control method based on large model decision-making.

[0048] Figure 2 This is a schematic diagram of a commodity control device based on large model decision-making disclosed in this application;

[0049] Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0050] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0051] This application discloses a commodity management method based on large model decision-making, applied to a cloud server. See [link to relevant documentation]. Figure 1 As shown, the method includes:

[0052] Step S11: Obtain the latest information on products available for sale and the current end-to-end inventory status information of vending machines within the preset area; the latest information on products available for sale includes the batch information of the products in the vending machine, and the batch information includes the production date of the products and the channel identifier bound to the production date.

[0053] It should be noted that the commodity control scheme based on large model decision proposed in this application is specifically applied to a cloud server to achieve safe and efficient management and control of the commodities in the vending machine.

[0054] In this embodiment, the cloud server can collect the latest information on the products on sale and the current full-chain inventory status of each vending machine in a preset area (such as areas in shopping malls, schools, stations, gyms, swimming pools, stadiums, etc.) according to a preset period (such as early morning every day). The preset area can be an area managed by a supplier, an area delivered by a delivery person, or an area divided according to the type of place (such as all school areas within a certain range). The latest information on the products on sale specifically includes the batch information of the products in each vending machine. The batch information includes, but is not limited to, the production date of the product (such as 20240520), the channel identifier pre-bound to the production date, and the product attribute information (such as yogurt, bread).

[0055] It is understandable that the goods in the aisle are located on individual pallets, so the unique aisle markings can also reflect the specific pallet location of a particular product.

[0056] Specifically, the current end-to-end inventory status information refers to the end-to-end inventory data of the same batch of goods in the vending machine, from warehouse inventory (in the central warehouse) to in-transit inventory (in the delivery vehicle group) and then to vending machine inventory (in the vending machine's product aisle).

[0057] It should be noted that before obtaining the latest information on products for sale and the current end-to-end inventory status of vending machines within the preset area, the process specifically includes: when it is detected that a product has been purchased and entered into the central warehouse, obtaining the physical packaging information of the product; based on the physical packaging information, entering the batch information of the same batch of products in the purchase receipt in the warehouse management system; the batch information includes the production date of the product. In this embodiment, before managing the products in each vending machine, it is necessary to obtain the end-to-end inventory status information of the products. Specifically, it is possible to monitor whether the products purchased based on the purchase order have arrived and entered the central warehouse through acquisition devices such as cameras and image sensors. If they have entered the central warehouse, the physical packaging information of the current batch of products is obtained (such as basic elements such as trademarks, markings, and prices, as well as basic product information, packaging specifications, batch number, and expiration date). Then, based on the physical packaging information, the batch information of the same batch of products is entered into the purchase receipt in the warehouse management system (WMS), specifically including the production date, brand, and product name of the product. For example, the batch information includes the production date of "XX brand yogurt" as "20240520".

[0058] Furthermore, after entering the batch information of the current batch of goods, the process further includes: upon receiving a pick-up receipt from the warehouse manager after picking, atomically updating the product inventory information in the current warehouse inventory information and the current in-transit inventory information through a database transaction to obtain the latest warehouse inventory information and the updated in-transit inventory information; the product inventory information includes the batch information of the same batch of goods; upon receiving a delivery receipt from the delivery person after delivery, atomically updating the current on-sale inventory information of the vending machine and the product inventory information in the updated in-transit inventory information through a database transaction to obtain the latest on-sale inventory information and the latest in-transit inventory information; the latest on-sale inventory information includes the product identifier and the corresponding channel identifier; sending the latest on-sale inventory information to the vending machine to bind the production date of the goods on different channels in the vending machine to the channel where the goods are located based on the latest on-sale inventory information, thereby obtaining the latest on-sale product information of the vending machine. Understandably, before goods are delivered, a corresponding pick-up slip needs to be generated based on the delivery plan. This slip guides warehouse staff on which location to retrieve goods, which batch to retrieve, and how much to retrieve, converting the quantity into easily operable units. The pick-up slip can list information such as the name, quantity, and production date of the goods to be picked up. This allows warehouse staff to directly retrieve the corresponding goods based on the pick-up slip, enabling delivery personnel to complete the delivery task. It should be noted that after the warehouse staff completes the picking operation from the central warehouse, they can first verify whether the retrieved physical goods match the information on the pick-up slip. If they match, a pick-up and delivery receipt is sent via appropriate device (such as a mobile terminal). Once the cloud server receives the delivery receipt, it can perform an atomic update on the current warehouse inventory information and the product inventory information (including batch information of the same batch of products) in the current in-transit inventory information through a database transaction. This will yield the latest warehouse inventory information and the updated in-transit inventory information. For example, in a database transaction, the following atomic update operation can be performed: deduct the inventory quantity of the corresponding product (including the production date) from the record with stock_type=1 (warehouse) in the jxc_stock table, and simultaneously add the same amount of inventory quantity to the record with stock_type=2 (in-transit / vehicle group). This step can realize the transfer of inventory status from warehouse to in-transit. Furthermore, the atomic update method can ensure that when the product leaves the central warehouse, its batch information is synchronously transferred to the delivery stage, ensuring data consistency during the inventory status transfer process.

[0059] Furthermore, after the delivery person completes the corresponding delivery / replenishment task according to the delivery note (which specifies the target vending machine, the product aisle, and the product's production date, etc.), they can send a delivery receipt to the cloud server via a handheld terminal (such as a mobile phone). Specifically, this can be done by clicking on the interface containing "Delivery Receipt." Once the server receives the delivery receipt from the delivery person, it can invoke the delivery receipt service to perform a second atomic update on the current in-sale inventory information and the updated in-transit inventory information of the vending machine through database transactions. This results in the latest in-sale inventory information (including product identifiers and corresponding product aisle identifiers) and the latest in-transit inventory information. For example, within a database transaction, the following atomic update operation is performed: The inventory quantity of the corresponding product (including the production date) is deducted from the record with stock_type=2 (in transit / vehicle group) (i.e., in-transit inventory information). Simultaneously, a new record with stock_type=3 (available for sale / pallet) (i.e., available for sale inventory information) is added to the jxc_stock table. This record specifically includes information such as stock_id (a unique physical identifier for the vending machine), commodity_id (a unique identifier for the product), and create_date (the production date from the delivery note). Through these two atomic update operations, the inventory status transfer from warehouse to delivery vehicle to vending machine is achieved. This ensures that the production date of the product is ultimately and accurately bound to the corresponding physical vending machine's vending machine, guarantees information synchronization, avoids the risk of information loss, and effectively prevents problems such as duplicate deductions, overselling, and unnecessary inventory.

[0060] Finally, the latest available inventory information is sent to the corresponding vending machine (i.e., the target vending machine currently being delivered by the delivery person). After receiving the latest available inventory information, the target vending machine can bind the production date of the products on different channels within the target vending machine to the channel where the products are located (e.g., binding create_date to stock_id), thereby obtaining the latest available product information of the target vending machine.

[0061] In one specific implementation, the end-to-end inventory status information specifically includes the latest warehouse inventory information, the latest in-transit inventory information, and the latest available inventory information located in a preset database. This end-to-end inventory status information spans the entire supply chain and provides a traceable, uninterrupted, high-fidelity data chain, which can solve the pain points of information "disruption" and "distortion" in the flow of goods.

[0062] In addition to pick-up and delivery receipts, RFID (Radio Frequency Identification) / UWB (Ultra-Wideband) devices can be installed on goods or vehicles to automatically sense changes in location and quickly trigger state transitions.

[0063] Step S12: Based on the latest on-sale product information and the full-chain inventory status information, construct a product status feature vector for the same batch of products in the vending machine; the product status feature vector includes the single-machine on-sale inventory, standard shelf life, production date, and remaining shelf life of the same batch of products in a single vending machine.

[0064] In this embodiment, a product status feature vector can be constructed for the same batch of products in each vending machine based on the latest information on products on sale and the full-chain inventory status information collected above. The product status feature vector can specifically include the single-machine inventory of the same batch of products in the current vending machine, the standard shelf life (i.e., shelf life duration), the production date, and the remaining shelf life (which can be obtained based on the production date and the standard shelf life).

[0065] In one specific implementation, constructing a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information may specifically include: obtaining the single-machine on-sale inventory of the same batch of products in a single vending machine based on the latest on-sale inventory information, and statistically analyzing the single-machine on-sale inventory of all vending machines within the preset area to obtain the total on-sale inventory of the same batch; constructing a product status feature vector for the same batch of products in the vending machine that includes the single-machine on-sale inventory, the standard shelf life, the production date, the remaining shelf life, and the total on-sale inventory of the same batch. In this embodiment, firstly, the quantity of the same batch of products currently on sale in a certain vending machine is obtained based on the latest on-sale inventory information to obtain the single-machine on-sale inventory; then, the single-machine on-sale inventory of all vending machines within the preset area is statistically analyzed, i.e., the total quantity of the same batch is statistically analyzed to obtain the total on-sale inventory of the same batch; finally, a product status feature vector is constructed based on the single-machine on-sale inventory, the standard shelf life, the production date, the remaining shelf life, and the total on-sale inventory of the same batch.

[0066] Step S13: Input the product state feature vector into the trained product decision model; the product decision model is a model obtained by training a large decision model based on the Transformer architecture using historical product information and historical full-link inventory status information of different vending machines.

[0067] In this embodiment, after creating the product status feature vector, the vector can be input into the product decision model obtained by training a large decision model based on the Transformer architecture using historical product information and historical end-to-end inventory status information from different vending machines, so as to make comprehensive decisions through the model.

[0068] The decision-making model can be an artificial intelligence model, such as DeepMind's Gato model. This model learns and outputs a series of decision results (such as normal sales, discounts, and removal from shelves) by observing environmental state factors (such as inventory, sales volume, and time).

[0069] In addition, to improve decision-making accuracy and meet more diverse needs, besides historical information on products on sale and historical inventory status information throughout the entire supply chain, other factors that influence product decisions can be added to the training dataset, such as historical weather information and corresponding sales information, historical holiday information and corresponding sales information, etc., to train the model, thereby enabling the model to make differentiated decisions under different weather and holiday conditions.

[0070] Step S14: Obtain the target product control decision output by the product decision model based on the product state feature vector, and execute the control operation corresponding to the product in the vending machine based on the target product control decision.

[0071] In this embodiment, the target product control decision output by the product decision model is obtained. The target product control decision is the result of a comprehensive analysis of multiple factors (such as inventory, sales volume, time, etc.) in the product state feature vector, such as normal sales, discounted sales, immediate removal from shelves, and replenishment. Then, based on the target product control decision, the corresponding control operations (such as removing the product from shelves) are performed on the products in the vending machine.

[0072] In one specific implementation, the target product control decision is output based on the product status feature vector. Specifically, it may include: determining whether the single-unit inventory in the product status feature vector is less than a first preset inventory threshold; if the single-unit inventory is less than the first preset inventory threshold, a target product control decision is generated to replenish the product at the target channel corresponding to the channel identifier; if the single-unit inventory is not less than the first preset inventory threshold, the ratio of the remaining shelf life to the standard shelf life is calculated, and it is determined whether the ratio is less than a preset product health threshold; if the ratio is less than the preset product health threshold, a target product control decision is generated to remove the product at the target channel corresponding to the channel identifier from the shelves. In this embodiment, it is first determined whether the single-machine inventory in the product status feature vector (i.e., the quantity of a certain batch of products in a single vending machine) is less than a first preset inventory threshold. If it is less, it indicates that the quantity of products is too small. At this time, a replenishment notice for the current batch of products can be generated based on the identification information of the target channel where the current batch of products is located (i.e., the target channel identifier), thus obtaining a target product control decision. If it is not less than the first preset inventory threshold, it means that the quantity in stock can normally meet the needs of customers. At this time, the ratio of the remaining shelf life (e.g., the remaining shelf life is 12 days) to the standard shelf life (e.g., the shelf life is 30 days) (i.e., 12 / 30) can be calculated. Then, it is determined whether the ratio (i.e., 12 / 30) is less than a preset product health threshold (which can be selected according to actual application needs, such as 1 / 4). If the ratio is less than the preset product health threshold, it indicates that the current product has certain health risks and may face the risk of expiring soon. At this time, a notice to remove the product at the target channel corresponding to the target channel identifier can be generated, thus obtaining a target product control decision. If the ratio is not less than the preset product health threshold, it indicates that there are no health risks and the product can be sold normally.

[0073] Accordingly, the execution of control operations corresponding to the products in the vending machine based on the target product control decision may specifically include: sending the target product control decision to the corresponding delivery terminal and / or the vending machine, so as to replenish or remove the products in the corresponding aisle of the vending machine according to the target product control decision. For example, sending the target product control decision containing replenishment or removal notifications to the delivery person's mobile terminal allows for timely replenishment or removal based on the received notification, thereby improving inventory turnover efficiency. Alternatively, sending the target product control decision containing removal notifications to the vending machine allows the vending machine to lock the corresponding tray upon receiving the decision result, thereby avoiding the risk of near-expiry / expired products entering the market (e.g., food safety risks caused by the sale of expired products due to human delays), which is particularly suitable for scenarios with strict regulations, such as dairy products and baked goods in schools.

[0074] In addition, the decision-making process for controlling target products can also include information such as the type, batch, and location of the products to be removed from shelves, so that delivery personnel can remove near-expiry / expired products in a timely manner, saving labor and time costs and improving the work efficiency of delivery personnel.

[0075] In another specific implementation, the target product control decision is output based on the product status feature vector. Specifically, this may include: if the single-unit inventory is not less than a first preset inventory threshold, then determining whether the total inventory of the same batch exceeds a second preset inventory threshold; the second preset inventory threshold is greater than the first preset inventory threshold; if the total inventory of the same batch exceeds the second preset inventory threshold, then generating a target product control decision to discount the current batch of products according to a first discount rate. That is, when the single-unit inventory is not less than the first preset inventory threshold (i.e., sufficient product inventory), it is further determined whether the total inventory of the same batch exceeds the second preset inventory threshold (this threshold is greater than the first preset inventory threshold). If it does, it indicates that the current batch of products has a large remaining inventory. In this case, to accelerate turnover, a notification to discount the current batch of products (which may include the amounts before and after the discount) can be generated, thus obtaining the target product control decision. It should be noted that the specific discount rate can be selected according to the actual application scenario of different types of products.

[0076] Specifically, constructing a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information may include: determining the number of days the current batch of products is on the shelf based on the product listing time in the latest on-sale product information; and constructing a product status feature vector for the same batch of products in the vending machine that includes the single-machine inventory, standard shelf life, production date, remaining shelf life, and the number of days on the shelf. That is, when constructing the product status feature vector, the number of days the current batch of products is on the shelf (i.e., the number of unsold days) is taken into account.

[0077] Accordingly, the target product management decision is output based on the product status feature vector. Specifically, this may include: if the ratio is not less than a preset product health threshold, determining whether the number of days on the shelf exceeds a preset slow-moving period; if the number of days on the shelf exceeds the preset slow-moving period, generating a target product management decision to discount the current batch of products according to a second discount rate. In this embodiment, if the ratio is not less than the preset product health threshold, it is further determined whether the number of days on the shelf of the current batch of products exceeds the preset slow-moving period. If it does, it indicates that the recent sales rate of the product on the current vending machine is slow. At this time, a discount notification for the current batch of products can be generated and sent to the corresponding vending machine to display information before and after the discount on the vending machine to attract customers' purchasing intentions.

[0078] Additionally, other information can be added to the product status feature vector, such as in-transit inventory of the same batch, historical sales volume, whether it is slow-moving, sales trends, date, weather, and holidays, to make more reasonable decisions based on more factors. For example, if a product in a certain region has low historical sales volume and slow-moving inventory, a pre-sale discount notification can be generated for products at a certain location to quickly clear inventory and avoid inventory backlog in that region. Similarly, during holidays, discount promotion notifications can be generated for products with large inventories to reduce the overall loss rate and improve inventory turnover efficiency.

[0079] Specifically, in addition to directly modifying the price, promotional instructions can also be sent out by displaying dynamic discount codes on the vending machine screen or nearby advertising screens, as a flexible and low-cost marketing intervention method.

[0080] Furthermore, after obtaining the target product control decision, the decision can be analyzed to generate a corresponding analysis report (containing the target product control decision). This report is then sent to the relevant managers for manual judgment / decision-making, further verifying and confirming the decision results obtained from the model. Alternatively, anomalies / errors in the target product control decision can be corrected based on the analysis report. The confirmed / corrected decision results are then sent to the corresponding terminals, such as vending machines and delivery terminals, to execute the corresponding product control operations based on the confirmed / corrected decision results. In addition, a procurement forecast list (i.e., a procurement plan) can be generated based on the decision results (such as the target product control decision and the confirmed / corrected decision results) to guide the subsequent procurement of goods by the central warehouse.

[0081] As can be seen, this application embodiment achieves automated management and control of goods in vending machines through a cloud server. First, based on the latest information on goods on sale in the vending machine and the current end-to-end inventory status information, a product status feature vector is constructed for the same batch of goods. Then, a product decision model based on the Transformer architecture is used to comprehensively analyze the product status feature vector to obtain the corresponding product decision results. Based on the product decision results, corresponding management and control operations are executed. Through the above method, multi-dimensional information (such as the latest information on goods on sale and end-to-end inventory status information) is integrated and comprehensive decision is made through a decision model. This allows for more accurate and effective management and control of goods in different inventory and on-sale states. Furthermore, the automated product management and control achieved through the cloud server can reduce labor and time costs, making operation and management easier and more efficient.

[0082] Accordingly, this application also discloses a commodity control device based on large model decision-making, applied to a cloud server, see [link to relevant documentation]. Figure 2 As shown, the device includes:

[0083] The acquisition module 11 is used to acquire the latest information on products available for sale and the current full-chain inventory status information of vending machines within a preset area; the latest information on products available for sale includes batch information of products in the vending machine, and the batch information includes the production date of the product and the channel identifier bound to the production date;

[0084] The vector creation module 12 is used to construct a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the full-chain inventory status information; the product status feature vector includes the single-machine on-sale inventory, standard shelf life, production date and remaining shelf life of the same batch of products in the vending machine.

[0085] Input module 13 is used to input the product state feature vector into the trained product decision model; the product decision model is a model obtained by training a large decision model based on the Transformer architecture using historical product information and historical full-link inventory status information of different vending machines.

[0086] The product control module 14 is used to obtain the target product control decision output by the product decision model based on the product state feature vector, and to execute the control operation corresponding to the product in the vending machine based on the target product control decision.

[0087] The specific workflow of each of the above modules can be found in the relevant content disclosed in the foregoing embodiments, and will not be repeated here.

[0088] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0089] Figure 3 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the commodity control method based on large model decision-making disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0090] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0091] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0092] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the large-model-based commodity control method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0093] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned commodity control method based on large model decision-making. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0094] Furthermore, embodiments of this application also disclose a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the commodity control method based on large model decision-making as disclosed above.

[0095] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0096] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0097] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0098] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0099] The above provides a detailed description of the commodity control method, apparatus, equipment, and medium based on large model decision-making provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A commodity control method based on large-scale model decision-making, characterized in that, Applications in cloud servers, including: Obtain the latest information on products available for sale and the current end-to-end inventory status of vending machines within a preset area; the latest information on products available for sale includes batch information of the products in the vending machines, and the batch information includes the production date of the products and the channel identifier associated with the production date; Based on the latest information on products for sale and the end-to-end inventory status information, a product status feature vector is constructed for the same batch of products in the vending machine; the product status feature vector includes the single-machine inventory, standard shelf life, production date, and remaining shelf life of the same batch of products in the vending machine. The product status feature vector is input into the trained product decision model; the product decision model is a model obtained by training a large decision model based on the Transformer architecture using historical product information and historical full-link inventory status information of different vending machines. Obtain the target product control decision output by the product decision model based on the product state feature vector, and execute the control operation corresponding to the product in the vending machine based on the target product control decision.

2. The commodity control method based on large model decision-making according to claim 1, characterized in that, Before obtaining the latest information on available products and current end-to-end inventory status of vending machines within the preset area, the process also includes: When it is detected that goods have been procured and entered the central warehouse, the physical packaging information of the goods is obtained; Based on the physical packaging information, the batch information of the goods in the same batch is entered into the purchase receipt in the warehouse management system; the batch information includes the production date of the goods.

3. The commodity control method based on large model decision-making according to claim 2, characterized in that, Also includes: When the warehouse manager sends a pick-up and outbound receipt after picking is completed, the current warehouse inventory information and the current in-transit inventory information are updated atomically through database transactions to obtain the latest warehouse inventory information and the updated in-transit inventory information. The product inventory information includes batch information for products from the same batch. When a delivery receipt is received from the delivery person after delivery is completed, the current inventory information for sale and the inventory information in the updated in-transit inventory information of the vending machine are atomically updated through a database transaction to obtain the latest inventory information for sale and the latest inventory information in transit. The latest inventory information for sale includes product identification and corresponding channel identification; The latest available inventory information is sent to the vending machine to bind the production date of the products on different aisles of the vending machine with the aisle of the product based on the latest available inventory information, so as to obtain the latest available product information of the vending machine.

4. The commodity control method based on large model decision-making according to claim 3, characterized in that, The end-to-end inventory status information includes the latest warehouse inventory information, the latest in-transit inventory information, and the latest inventory for sale information, and the end-to-end inventory status information is located in a preset database; Accordingly, constructing a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information includes: Based on the latest available inventory information, obtain the single-machine available inventory of the same batch of goods in a single vending machine, and statistically analyze the single-machine available inventory of all vending machines in the preset area to obtain the total available inventory of the same batch. Construct a product status feature vector for the same batch of products in the vending machine, which includes the single machine's inventory, the standard shelf life, the production date, the remaining shelf life, and the total inventory of the same batch.

5. The commodity control method based on large model decision-making according to claim 4, characterized in that, Based on the product status feature vector, the target product control decision is output, including: Determine whether the stand-alone inventory in the product status feature vector is less than a first preset inventory threshold. If the stand-alone inventory is less than the first preset inventory threshold, generate a target product management decision to replenish the product at the target channel corresponding to the channel identifier. If the inventory of a single unit is not less than the first preset inventory threshold, then calculate the ratio of the remaining shelf life to the standard shelf life, and determine whether the ratio is less than the preset product health threshold. If the ratio is less than the preset product health threshold, a target product control decision is generated to remove the product from the target channel corresponding to the channel identifier. Accordingly, the execution of control operations corresponding to the goods in the vending machine based on the target goods control decision includes: The target product control decision is sent to the corresponding delivery terminal and / or the vending machine, so as to replenish or remove the product in the corresponding aisle of the vending machine according to the target product control decision.

6. The commodity control method based on large model decision-making according to claim 4, characterized in that, Based on the product status feature vector, the target product control decision is output, including: If the inventory of a single unit for sale is not less than the first preset inventory threshold, then it is determined whether the total inventory of the same batch for sale exceeds the second preset inventory threshold; the second preset inventory threshold is greater than the first preset inventory threshold. If the total inventory of the same batch exceeds the second preset inventory threshold, a target product control decision is generated to discount the current batch of products according to the first discount rate.

7. The commodity control method based on large model decision-making according to claim 5, characterized in that, The step of constructing a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information includes: The number of days that the same batch of products are available on the shelves is determined based on the product listing time in the latest product information. Construct a product status feature vector for the same batch of products in the vending machine, including the single machine's inventory, standard shelf life, production date, remaining shelf life, and the number of days on the shelf; Accordingly, based on the product status feature vector, the target product control decision is output, including: If the ratio is not less than the preset product health threshold, then it is determined whether the number of days on the shelf exceeds the preset unsold period. If the number of days on the shelf exceeds the preset slow-moving period, a target product management decision is generated to discount the current batch of products according to the second discount rate.

8. A commodity control device based on large-scale model decision-making, characterized in that, Applications in cloud servers, including: The acquisition module is used to acquire the latest information on products available for sale and the current end-to-end inventory status information of vending machines within a preset area; the latest information on products available for sale includes batch information of products in the vending machine, and the batch information includes the production date of the product and the channel identifier bound to the production date; The vector creation module is used to construct a product status feature vector for the same batch of products in the vending machine based on the latest on-sale product information and the end-to-end inventory status information; the product status feature vector includes the single-machine on-sale inventory, standard shelf life, production date and remaining shelf life of the same batch of products in the vending machine. The input module is used to input the product state feature vector into the trained product decision model; the product decision model is a model obtained by training a large decision model based on the Transformer architecture using historical product information and historical full-link inventory status information of different vending machines. The product control module is used to obtain the target product control decision output by the product decision model based on the product state feature vector, and to execute the control operation corresponding to the product in the vending machine based on the target product control decision.

9. An electronic device, characterized in that, It includes a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the commodity control method based on large model decision-making as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer programs; wherein, when the computer programs are executed by a processor, they implement the commodity control method based on large model decision-making as described in any one of claims 1 to 7.

Images