A method for controlling inventory of returned orders

Abstract

The present application relates to the technical field of inventory control, and specifically discloses a method for controlling returned orders, which specifically includes: obtaining sales orders pushed down by the platform; transmitting the sales orders to the inventory system for reservation, wherein if the available inventory is sufficient, determining that the sales order reservation is successful; if the available inventory is insufficient, determining that the sales order reservation fails, and determining the sales order with failed reservation as an out-of-stock order; obtaining shortage orders created by warehouse personnel; transmitting the shortage orders to the inventory system for reservation and responding to the cancellation order operation of the sales order that has been successfully reserved, the inventory system returns the previously occupied available inventory; based on the priority of the out-of-stock order, determining the priority ownership of the occupied available inventory. In this way, the time and energy consumption of manual review is reduced, the review speed is accelerated, and the processing efficiency is improved. And the priority can be adjusted in real time to ensure the timeliness and accuracy of inventory management.

Description

Technical Field

[0001] The present application relates to the technical field of inventory control, and more specifically, to a method for controlling order return inventory. Background Art

[0002] E-commerce platforms and retail companies are usually faced with a large number of orders, especially during promotional activities, when the order volume surges. Due to the fluctuation of order volume, inventory is often in dynamic change. Inventory management is the core link of these companies. By accurately tracking and managing the inventory of various commodities, it can ensure the smooth fulfillment of orders.

[0003] When managing inventory, each product has its own available inventory. When a sales order enters the system, the system will try to reserve the inventory. If there are too many orders and the available inventory is insufficient, the reservation will fail and the order will be marked as out of stock. In addition, warehouse staff will also conduct inventory counts regularly to identify and handle inventory losses caused by damage, expiration, etc., and create loss orders.

[0004] When a user cancels an order, the inventory occupied by the order will be returned to the system, and the system will wake up the out-of-stock order and try to re-book it. However, in the traditional inventory management and control process, these returned inventories will be immediately put back into the bookable inventory pool, triggering random competition between multiple out-of-stock orders and new orders. And because new orders and out-of-stock orders compete at the same time, the allocation of inventory is completely random, and it is impossible to guarantee that high-priority orders (such as shortage orders) will get priority in inventory. This not only leads to confusion in inventory management, but may also cause delays in high-priority orders, affecting the accuracy of inventory data and the stability of the supply chain. In addition, this random allocation mechanism may lead to a waste of inventory resources, because high-priority orders are not processed in time, while low-priority new orders may take priority in occupying inventory, further exacerbating the complexity and uncertainty of inventory management, affecting customer satisfaction and the operational efficiency of the enterprise.

[0005] Therefore, a return order inventory control solution is desired. Summary of the invention

[0006] The present application provides a method for controlling inventory of returned orders, which can reduce the time and energy consumption of manual review, speed up the review process, and improve processing efficiency. It can also adjust the priority in real time to ensure the timeliness and accuracy of inventory management.

[0007] In a first aspect, a method for controlling order return and inventory control is provided, comprising: obtaining a sales order pushed down by a platform; transmitting the sales order to an inventory system for reservation, wherein if the reservable inventory is sufficient, determining that the sales order has been successfully reserved; if the reservable inventory is insufficient, determining that the sales order has failed to be reserved, and determining the sales order with a failed reservation as an out-of-stock order; obtaining a shortage order created by a warehouse staff; transmitting the shortage order to an inventory system for reservation, wherein if the reservable inventory is insufficient, determining that the shortage order has failed to be reserved, and confirming the shortage order with a failed reservation as an out-of-stock order; in response to receiving a cancel order operation for a successfully reserved sales order, the inventory system returns the previously occupied reservable inventory; and determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order.

[0008] In a possible implementation, in determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order, the priority of the out-of-stock order is greater than the priority of the sales order.

[0009] In one possible implementation, based on the priority of the out-of-stock order, the priority ownership of the occupied reservable inventory is determined, including: obtaining the set of shortage orders; semantically encoding each shortage order in the set of shortage orders to obtain a set of semantic encoding feature vectors of shortage orders; performing dynamic feature kernel clustering analysis on the set of semantic encoding feature vectors of shortage orders to obtain a semantic kernel clustering representation of shortage orders; calculating the semantic priority offset between each semantic encoding feature vector of shortage orders in the set of semantic encoding feature vectors of shortage orders and the semantic kernel clustering representation of shortage orders to obtain a set of semantic priority offsets; and determining the priority ranking of the set of shortage orders based on the ranking of the set of semantic priority offsets.

[0010] In a possible implementation, a dynamic feature kernel clustering analysis is performed on the set of semantically encoded feature vectors of shortage orders to obtain a semantic kernel clustering representation of shortage orders, including: calculating the positional mean vector of the set of semantically encoded feature vectors of shortage orders to obtain an initial aggregate representation vector of shortage orders; calculating the static energy potential factor of each semantically encoded feature vector of shortage orders in the set of semantically encoded feature vectors of shortage orders to obtain a set of static energy potential factors of shortage orders; calculating the dynamic energy potential factor of each semantically encoded feature vector of shortage orders in the set of semantically encoded feature vectors of shortage orders relative to the initial aggregate representation vector of shortage orders to obtain a set of dynamic energy potential factors of shortage orders; based on the The static potential factors and dynamic potential factors of each loss-making order semantic encoding feature vector in the set of loss-making order semantic encoding feature vectors are used to determine the superposition state potential factors of each loss-making order semantic encoding feature vector in the set of loss-making order semantic encoding feature vectors to obtain a set of superposition state potential factors of loss-making orders; the set of superposition state potential factors of loss-making orders is input into the potential factor gated weighted network to obtain a set of superposition potential weight coefficients of loss-making order semantic features; based on the set of superposition potential weight coefficients of loss-making order semantic features, the position-weighted sum of the set of loss-making order semantic encoding feature vectors is calculated to obtain a loss-making order semantic kernel cluster representation vector as the loss-making order semantic kernel cluster representation.

[0011] In a possible implementation, the static energy potential factor of each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors is calculated to obtain a set of shortage order static energy potential factors, including: calculating the mean and standard deviation of the shortage order semantic encoding feature vector to obtain the shortage order semantic mean and shortage order semantic standard deviation; calculating the positional subtraction of the shortage order semantic encoding feature vector and the shortage order semantic mean, dividing the shortage order semantic offset vector after the subtraction by the shortage order semantic standard deviation, and calculating the fourth power of the eigenvalue of each position in the vector after the division to obtain the shortage order semantic static energy potential vector; calculating the positional mean of the shortage order semantic static energy potential vector to obtain the shortage order static energy potential factor.

[0012] In a possible implementation, the dynamic potential factor of each loss-order semantic encoding feature vector in the set of loss-order semantic encoding feature vectors is calculated relative to the loss-order initial aggregate representation vector to obtain a set of loss-order dynamic potential factors, including: matrix multiplying the loss-order aggregation weight matrix with the loss-order initial aggregate representation vector to obtain the loss-order initial aggregate weight representation vector; matrix multiplying the loss-order weight matrix with the loss-order semantic encoding feature vector to obtain the loss-order semantic encoding weight vector; performing positional dot multiplication of the loss-order initial aggregate weight representation vector with the loss-order semantic encoding weight vector, and then adding the obtained loss-order semantic encoding fusion weight vector to the loss-order bias vector to obtain the loss-order semantic dynamic potential vector; calculating the square of the Frobenius norm of the loss-order semantic dynamic potential vector to obtain the loss-order dynamic potential factor.

[0013] In a possible implementation, based on the static potential factor and dynamic potential factor of each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors, the superposition state potential factor of each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors is determined to obtain a set of superposition state potential factors of shortage orders, including: calculating an exponential function with the natural constant e as the base and the static potential factor of the shortage order semantic encoding feature vector as the exponent to obtain the shortage order semantic bias static potential factor; calculating an exponential function with the natural constant e as the base and the square of the dynamic potential factor of the shortage order semantic encoding feature vector as the exponent to obtain the shortage order semantic bias dynamic potential factor; adding the shortage order semantic bias static potential factor and the shortage order semantic bias dynamic potential factor to obtain the superposition state potential factor of the shortage order semantic encoding feature vector.

[0014] In one possible implementation, the set of the short-order superposition state energy potential factors is input into the energy potential factor gated weighting network to obtain a set of short-order semantic feature superposition energy potential weight coefficients, including: normalizing the set of short-order superposition state energy potential factors to obtain a set of normalized short-order superposition state energy potential factors; inputting the set of normalized short-order superposition state energy potential factors into a mask function for gated mask screening to obtain a set of short-order semantic feature superposition energy potential weight coefficients, wherein, in response to each normalized short-order superposition state energy potential factor in the set of normalized short-order superposition state energy potential factors being greater than a preset threshold, the normalized short-order superposition state energy potential factor is set to the original value, and the rest are set to zero, so as to obtain the set of short-order semantic feature superposition energy potential weight coefficients.

[0015] In a possible implementation, the set of the deficit order superposition state energy potential factors is normalized to obtain a set of normalized deficit order superposition state energy potential factors, including: calculating an exponential function value with the natural constant e as the base and the negative of each deficit order superposition state energy potential factor in the set of deficit order superposition state energy potential factors as the exponent to obtain a set of deficit order superposition state energy potential exponential factors; adding the set of deficit order superposition state energy potential exponential factors to the constant one by position, and then calculating the reciprocal of each deficit order superposition state energy potential exponential modulation factor in the obtained set of deficit order superposition state energy potential exponential modulation factors to obtain the set of normalized deficit order superposition state energy potential factors.

[0016] The present application provides a method for controlling returned order inventory. In determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order, the technical concept of the present application is to obtain the set of shortage orders, and use data analysis and coding technology based on deep learning to semantically encode each shortage order, and then perform dynamic feature kernel clustering analysis on the semantic coding features of each shortage order, so as to intelligently determine the priority ranking of the set of shortage orders based on each semantic priority offset measure between the semantic coding features of each shortage order and the semantic kernel clustering representation of the shortage order. In this way, the time and energy consumption of manual review is reduced, the review speed is accelerated, and the processing efficiency is improved. And the priority can be adjusted in real time to ensure the timeliness and accuracy of inventory management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0018] Figure 1 This is a schematic flowchart of the order return inventory control method according to an embodiment of the present application.

[0019] Figure 2 This is an order processing flow chart of the order return inventory control method according to an embodiment of the present application.

[0020] Figure 3 This is a schematic flowchart of determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order in the return order return inventory control method of an embodiment of the present application.

[0021] Figure 4 This is a data flow diagram for determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order in the order return inventory control method of an embodiment of the present application.

[0022] Figure 5 In the order return inventory control method of an embodiment of the present application, a dynamic feature kernel clustering analysis is performed on the set of semantic encoding feature vectors of the shortage order to obtain a schematic flowchart of the semantic kernel clustering representation of the shortage order. DETAILED DESCRIPTION

[0023] Since the field of inventory control is involved here, the relevant terms and concepts in the field of inventory management that may be involved in the embodiments of the present application are first introduced below.

[0024] 1. Reservable inventory: The number of reservable inventory for a single item is a virtual inventory count. When the number of reservable inventory is 0, all new orders will not be able to be booked and will be out of stock. 2. Reservation order: After entering the system, you need to reserve the inventory of each item in the order first. After the reservation is successful, the reservable inventory will be deducted accordingly, that is, the reserved inventory cannot be reserved by other orders at the same time. If the order fails to reserve inventory, the fulfillment will not be completed. 3. Cancel order: After the user places an order, the order enters the OMS system to reserve inventory. After that, the user can cancel the order. After the order is cancelled, the inventory occupied by the order needs to be returned to the reservable inventory. 4. Re-booking when out of stock: When there is new reservable inventory, the out-of-stock order will be awakened and asked to try to make a reservation. 5. Sales order: For orders used for sales, inventory reservation is required. After success, the subsequent process will be carried out to complete the order fulfillment. 6. Shortage order: Warehouse staff will regularly count the inventory and may create a shortage order due to damage, expiration and other reasons. Shortage orders need to reserve inventory first to prevent this part of the inventory from being sold. 7. Returned Inventory: After an order is canceled, according to the solution of this application, this part of the inventory is placed in the returned inventory pool and does not directly enter the reservable inventory.

[0025] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0026] Based on this, this application proposes a method for controlling inventory returned by return order, which solves the problems of random inventory allocation and insufficient priority management in the prior art, such as Figure 1 and Figure 2As shown, the order return and inventory control method includes: S1, obtaining a sales order pushed down by a platform; S2, transmitting the sales order to an inventory system for reservation, wherein if the reservable inventory is sufficient, determining that the sales order has been successfully reserved; if the reservable inventory is insufficient, determining that the sales order has failed to be reserved, and determining the sales order with a failed reservation as an out-of-stock order; S3, obtaining a shortage order created by a warehouse staff; S4, transmitting the shortage order to an inventory system for reservation, wherein if the reservable inventory is insufficient, determining that the shortage order has failed to be reserved, and determining the shortage order with a failed reservation as an out-of-stock order; S5, in response to receiving a cancel order operation for a successfully reserved sales order, the inventory system returns the previously occupied reservable inventory; S6, determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order.

[0027] In the above-mentioned order return inventory control method, S1 obtains the sales order pushed by the platform. It should be understood that e-commerce platforms and retail companies need to accurately track and manage the inventory of various commodities to ensure the smooth fulfillment of orders. Obtaining sales orders is the first step in inventory management, which can help the system understand the upcoming order requirements. By automatically obtaining sales orders, manual intervention can be reduced, the speed and accuracy of order processing can be improved, and operating costs can be reduced. The order volume of e-commerce platforms fluctuates greatly, especially during promotional activities. Real-time acquisition of sales orders can adjust inventory in time to avoid the situation where orders cannot be fulfilled due to insufficient inventory. By obtaining sales orders, the system can reasonably allocate inventory according to the priority and inventory status of the orders to ensure that high-priority orders (such as inventory shortage orders) get inventory first.

[0028] In the above-mentioned order return and inventory control method, S2, the sales order is transmitted to the inventory system for reservation, wherein if the available inventory is sufficient, it is determined that the sales order is successfully reserved; if the available inventory is insufficient, it is determined that the sales order reservation fails, and the sales order that fails to be reserved is determined as an out-of-stock order. It should be understood that the purpose of transmitting the sales order to the inventory system for reservation is to ensure that the order can be fulfilled smoothly. The inventory system will check the current available inventory quantity to ensure that there is enough inventory to meet the order requirements. Through the reservation operation, the system can lock the inventory required for the order to prevent other orders from reserving the same inventory at the same time, thereby avoiding the situation of overselling inventory. The reservation operation can update the inventory status in real time to ensure the accuracy of inventory data. This is very important for inventory management and order processing, especially when the inventory is tight, it can avoid order fulfillment failure caused by insufficient inventory. For example, the available inventory is split into two inventory pools, one is the original available inventory pool, and the other is the return inventory pool. The available inventory pool is only used for orders newly entering the system, and the inventory in the return inventory pool is released from the cancellation order operation of other orders. The return inventory pool is only used for out-of-stock orders. When the available inventory in the system is not sufficient, the platform pushes order a, and order a attempts to deduct inventory from the available inventory pool. The deduction is successful and the reservation is successful. After that, the platform pushes order b, and order b attempts to make a reservation and deducts inventory from the available inventory pool. At this time, the available inventory is insufficient, and the deduction fails and the reservation fails. Order b is marked as out of stock.

[0029] In the above-mentioned return order inventory control method, S3 obtains the inventory shortage order created by the warehouse staff. It should be understood that the inventory shortage order reflects the actual inventory loss in the warehouse, such as damage, expiration, loss, etc. Obtaining the inventory shortage order helps to update the inventory records in a timely manner and ensure the accuracy of the inventory data. By obtaining the inventory shortage order, these problem inventories can be handled in a timely manner to prevent these inventories from continuing to be sold, affecting the customer's shopping experience and the company's reputation. Timely processing of inventory shortage orders can help companies reduce inventory losses and reduce operating costs. With accurate inventory shortage records, companies can better analyze problems in inventory management and take measures to improve them. For example, the warehouse staff performed an inventory count and created an inventory shortage order c.

[0030] In the above-mentioned return order and inventory control method, S4, the shortage order is transmitted to the inventory system for reservation, wherein, if the available inventory is insufficient, it is determined that the shortage order has failed to be reserved, and the shortage order that failed to be reserved is confirmed as an out-of-stock order. It should be understood that the shortage order reflects the actual inventory loss in the warehouse, such as damage, expiration, loss, etc. Transmitting the shortage order to the inventory system for reservation can ensure that these lost inventories are recorded and processed in a timely manner, thereby maintaining the accuracy of inventory data. Through the reservation operation, it can be ensured that the inventory required for the shortage order is locked, preventing other orders from reserving the same inventory at the same time, thereby avoiding the situation of overselling inventory. This helps to maintain the orderliness and accuracy of inventory management. For example, shortage order c attempts to make a reservation operation and deducts inventory from the available inventory pool. As above, the available inventory is insufficient, and the failure of the deduction leads to a reservation failure. Shortage order c is marked as out of stock.

[0031] In the above-mentioned order return and inventory control method, S5, in response to receiving the cancellation operation of the sales order that has been successfully booked, the inventory system returns the previously occupied bookable inventory. It should be understood that when the user cancels the order, the inventory system needs to release the previously booked inventory in a timely manner to ensure the accuracy of the inventory data. This helps to avoid the accumulation of errors in inventory data and ensure the transparency and reliability of inventory management. Releasing the inventory occupied by the canceled order can increase the amount of bookable inventory, so that other orders have the opportunity to obtain these stocks. This helps to improve the utilization of inventory and reduce inventory waste. Releasing inventory can provide opportunities for other out-of-stock orders, especially high-priority shortage orders, to ensure that these orders can obtain inventory first and improve the fulfillment rate of orders. For example, a sales order a has a cancellation operation, and the previously booked inventory of a needs to be returned, and the returned inventory enters the return inventory pool.

[0032] In the above-mentioned method for controlling the returned order and returned inventory, S6 determines the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order. It should be understood that by determining the ownership of inventory based on priority, it can be ensured that high-priority orders (such as inventory shortage orders) have priority in obtaining inventory. This helps to improve the efficiency of inventory management and reduce order delays or cancellations caused by insufficient inventory. High-priority orders are usually urgent or important orders. For example, inventory shortage orders involve inventory losses and need to be processed as soon as possible. Prioritizing these orders can reduce customer waiting time and improve customer satisfaction. Inventory shortage orders reflect the actual inventory losses in the warehouse. Timely processing of these orders can ensure the accuracy of inventory data. This helps enterprises better understand the inventory status and make more reasonable inventory management decisions. For example, the platform then pushes down sales order d, and sales order d tries to make inventory reservations. At this time, there is still no inventory in the reservable inventory pool, and the sales order d reservation fails and is marked as out of stock. At the same time, there is new inventory in the returned inventory pool, which triggers the previous out-of-stock high-priority order c to make an out-of-stock reservation. Order c deducts inventory from the returned inventory pool, and the deduction is successful and the reservation is successful.

[0033] Optionally, in one embodiment of the present application, in determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order, the priority of the out-of-stock order is greater than the priority of the sales order.

[0034] It should be understood that if there is only one shortage list and one sales order in the out-of-stock order, then the shortage order will be selected; if there are multiple shortage orders, then the multiple shortage orders need to be prioritized. The traditional priority sorting of multiple shortage orders relies on warehouse managers or financial personnel to manually review the shortage orders and adjust the priority according to the specific situation. However, manually reviewing each shortage order takes a lot of time and effort, especially when the order volume is large, the review process may be very slow. In addition, manual review cannot adjust the priority in real time, especially when the inventory changes frequently, it may not be able to respond to the new inventory situation in time.

[0035] Therefore, in determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order, the technical concept of the present application is to obtain the set of the shortage orders, and use data analysis and coding technology based on deep learning to semantically encode each shortage order, and then perform dynamic feature kernel clustering analysis on the semantic coding features of each shortage order, so as to intelligently determine the priority ranking of the set of shortage orders based on each semantic priority offset metric between the semantic coding features of each shortage order and the semantic kernel clustering representation of the shortage order. In this way, the time and energy consumption of manual review is reduced, the review speed is accelerated, and the processing efficiency is improved. And the priority can be adjusted in real time to ensure the timeliness and accuracy of inventory management.

[0036] Figure 3 This is a schematic flowchart of determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order in the return order return inventory control method of an embodiment of the present application. Figure 4 This is a schematic diagram of data flow for determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order in the return order return inventory control method of the embodiment of the present application. Figure 3 and Figure 4 As shown, based on the priority of the out-of-stock order, the priority ownership of the occupied reservable inventory is determined, including: S61, obtaining the set of the shortage orders; S62, semantically encoding each shortage order in the set of shortage orders to obtain a set of semantic encoding feature vectors of shortage orders; S63, performing dynamic feature kernel clustering analysis on the set of semantic encoding feature vectors of shortage orders to obtain a semantic kernel clustering representation of shortage orders; S64, calculating the semantic priority offset between each semantic encoding feature vector of shortage orders in the set of semantic encoding feature vectors of shortage orders and the semantic kernel clustering representation of shortage orders to obtain a set of semantic priority offsets; S65, determining the priority ranking of the set of shortage orders based on the ranking of the set of semantic priority offsets.

[0037] In the above-mentioned return order inventory control method, the S61 obtains the set of the shortage orders. It should be understood that the shortage orders reflect the actual inventory losses in the warehouse, such as damage, expiration, loss, etc. It should be understood that obtaining the set of shortage orders can ensure that these lost inventories are recorded and processed in a timely manner, thereby maintaining the accuracy of inventory data. This is very important for inventory management and order processing. Shortage orders usually have a higher priority and need to be processed first.

[0038] In the above-mentioned return order inventory control method, the S62 semantically encodes each shortage order in the set of shortage orders to obtain a set of semantic encoding feature vectors of shortage orders. It should be understood that, considering that the set of shortage orders obtained contains key semantic information such as product name and shortage reason, based on this, in order to be able to more clearly understand and analyze the semantic information contained in each shortage order, so as to more accurately classify and prioritize the shortage orders, in the technical solution of the present application, each shortage order in the set of shortage orders is semantically encoded to capture and extract the key semantic information about each order, and obtain a set of semantic encoding feature vectors of shortage orders. Optionally, in one embodiment of the present application, it is assumed that there is a shortage order containing the product name, customer message and loss amount of the shortage order. The BERT model can be used to encode the product name and customer message to obtain the text feature vector of each order. The loss amount is directly used as a numerical feature. If there are other category features, such as product type, One-Hot encoding can be used. Finally, all the above feature vectors are concatenated in order to form the semantic encoding feature vector of the shortage order.

[0039] Optionally, in one embodiment of the present application, semantic encoding is performed on each shortage order in the set of shortage orders to obtain a set of semantic encoding feature vectors of shortage orders, including: data preprocessing is performed on the shortage orders, and then embedding encoding is performed on the text of the shortage orders to convert it into a sequence of word vectors. Contextual semantic encoding is performed on the sequence of word vectors using a converter-based context encoder to obtain a semantic encoding feature vector of the shortage order.

[0040] In the above-mentioned order return inventory control method, the S63 performs dynamic feature kernel clustering analysis on the set of semantic encoding feature vectors of shortage orders to obtain a semantic kernel clustering representation of shortage orders. It should be understood that, considering that the set of semantic encoding feature vectors of shortage orders expresses the key semantic information and attribute characteristics of each shortage order, and there is a semantic association between the semantic encoding features of each shortage order, for example, orders with similar problem type characteristics (such as damage, expiration, and loss) often have similar processing methods and priorities, based on this, in order to be able to find a representative and information-rich kernel representation from this set, in the technical solution of the present application, a dynamic feature kernel clustering analysis is performed on the set of semantic encoding feature vectors of shortage orders to obtain a semantic kernel clustering representation vector of shortage orders as a semantic kernel clustering representation of shortage orders.

[0041] Optionally, in one embodiment of the present application, Figure 5In the method for controlling returned order inventory in the embodiment of the present application, a dynamic feature kernel clustering analysis is performed on the set of semantic encoding feature vectors of the shortage order to obtain a schematic flow chart of the semantic kernel clustering representation of the shortage order. Figure 5 As shown, a dynamic feature kernel clustering analysis is performed on the set of semantically encoded feature vectors of shortage orders to obtain a semantic kernel clustering representation of shortage orders, including: S631, calculating the positional mean vector of the set of semantically encoded feature vectors of shortage orders to obtain an initial aggregate representation vector of shortage orders; S632, calculating the static energy potential factor of each semantically encoded feature vector of shortage orders in the set of semantically encoded feature vectors of shortage orders to obtain a set of static energy potential factors of shortage orders; S633, calculating the dynamic energy potential factor of each semantically encoded feature vector of shortage orders in the set of semantically encoded feature vectors of shortage orders relative to the initial aggregate representation vector of shortage orders to obtain a set of dynamic energy potential factors of shortage orders; S634, based on the The static potential factors and dynamic potential factors of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors are used to determine the superposition state potential factors of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors to obtain a set of superposition state potential factors of missing orders; S635, inputting the set of superposition state potential factors of missing orders into the potential factor gated weighting network to obtain a set of superposition potential weight coefficients of missing order semantic features; S636, based on the set of superposition potential weight coefficients of missing order semantic features, calculating the position-weighted sum of the set of missing order semantic encoding feature vectors to obtain a missing order semantic kernel cluster representation vector as the missing order semantic kernel cluster representation.

[0042] Optionally, in one embodiment of the present application, the static energy potential factors of each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors are calculated to obtain a set of shortage order static energy potential factors, including: calculating the mean and standard deviation of the shortage order semantic encoding feature vectors to obtain the shortage order semantic mean and shortage order semantic standard deviation; calculating the positional subtraction of the shortage order semantic encoding feature vector and the shortage order semantic mean, dividing the shortage order semantic offset vector after the subtraction by the shortage order semantic standard deviation, and calculating the fourth power of the eigenvalue of each position in the vector after the division to obtain the shortage order semantic static energy potential vector; calculating the positional mean of the shortage order semantic static energy potential vector to obtain the shortage order static energy potential factor.

[0043] Optionally, in one embodiment of the present application, the dynamic energy potential factor of each loss-making order semantic encoding feature vector in the set of loss-making order semantic encoding feature vectors is calculated relative to the loss-making order initial aggregation representation vector to obtain a set of loss-making order dynamic energy potential factors, including: matrix multiplying the loss-making order aggregation weight matrix with the loss-making order initial aggregation representation vector to obtain the loss-making order initial aggregation weight representation vector; matrix multiplying the loss-making order weight matrix with the loss-making order semantic encoding feature vector to obtain the loss-making order semantic encoding weight vector; performing positional dot multiplication of the loss-making order initial aggregation weight representation vector with the loss-making order semantic encoding weight vector, and then adding the obtained loss-making order semantic encoding fusion weight vector to the loss-making order bias vector by position to obtain the loss-making order semantic dynamic energy potential vector; calculating the square of the Frobenius norm of the loss-making order semantic dynamic energy potential vector to obtain the loss-making order dynamic energy potential factor.

[0044] Optionally, in one embodiment of the present application, based on the static energy potential factor and dynamic energy potential factor of each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors, the superposition state energy potential factor of each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors is determined to obtain a set of shortage order superposition state energy potential factors, including: calculating an exponential function with the natural constant e as the base and the static energy potential factor of the shortage order semantic encoding feature vector as the exponent to obtain the shortage order semantic bias static energy potential factor; calculating an exponential function with the natural constant e as the base and the square of the dynamic energy potential factor of the shortage order semantic encoding feature vector as the exponent to obtain the shortage order semantic bias dynamic energy potential factor; adding the shortage order semantic bias static energy potential factor and the shortage order semantic bias dynamic energy potential factor to obtain the superposition state energy potential factor of the shortage order semantic encoding feature vector.

[0045] Optionally, in one embodiment of the present application, the set of the superimposed state energy potential factors of the shortage order is input into the energy potential factor gated weighting network to obtain a set of superimposed energy potential weight coefficients of the semantic features of the shortage order, including: normalizing the set of superimposed state energy potential factors of the shortage order to obtain a set of normalized superimposed state energy potential factors of the shortage order; inputting the set of normalized superimposed state energy potential factors of the shortage order into a mask function for gated mask screening to obtain a set of superimposed energy potential weight coefficients of the semantic features of the shortage order, wherein, in response to each normalized superimposed state energy potential factor of the shortage order in the set of normalized superimposed state energy potential factors of the shortage order is greater than a preset threshold, the normalized superimposed state energy potential factor of the shortage order is set to the original value, and the rest are set to zero, so as to obtain the set of superimposed energy potential weight coefficients of the semantic features of the shortage order.

[0046] Optionally, in one embodiment of the present application, the set of the loss order superposition state energy potential factors is normalized to obtain a set of normalized loss order superposition state energy potential factors, including: calculating an exponential function value with the natural constant e as the base and the negative number of each loss order superposition state energy potential factor in the set of loss order superposition state energy potential factors as the exponent to obtain a set of loss order superposition state energy potential exponential factors; adding the set of loss order superposition state energy potential exponential factors to the constant one by position, and then calculating the reciprocal of each loss order superposition state energy potential exponential modulation factor in the obtained set of loss order superposition state energy potential exponential modulation factors to obtain the set of normalized loss order superposition state energy potential factors.

[0047] In summary, in this application, a dynamic feature kernel representation is performed on the set of semantically encoded feature vectors of the shortage order, which can be expressed by the formula: ;in, is a set of semantic encoding feature vectors of the shortage order, , , and are the first, second, and third in the set of semantic encoding feature vectors of the shortage order. and The semantic encoding feature vector of the loss order, is the number of vectors in the set of semantic encoding feature vectors of the missing orders, is the initial aggregate representation vector of the loss order, yes The eigenvalues ​​at each position in , and They are The mean and standard deviation of yes The number of eigenvalues ​​in , yes The corresponding static energy potential factor of the loss order is: is the loss order aggregation weight matrix, is the loss order weight matrix, It is the point multiplication by position. is the short-selling order bias vector, is the square of the Frobenius norm of the calculated vector, yes The corresponding dynamic potential factor of the loss order, yes The corresponding loss order superposition state energy potential factor, yes The corresponding normalized shortfall order superposition state energy potential factor, For gated mask operation, is the preset threshold, yes The corresponding semantic features of the missing orders are superimposed with the potential weight coefficient. is the semantic kernel clustering representation vector of the loss order.

[0048] In detail, first, the positional mean vector of the set of semantic encoding feature vectors of missing orders is calculated to determine a comprehensive feature that can summarize the semantic encoding features of all missing orders, that is, to find a core representative point or benchmark vector, while filtering out the noise in the data and making the data pattern more obvious, to obtain the initial aggregate representation vector of the missing orders. Next, in order to evaluate the importance and inherent characteristic attributes of each missing order to highlight the order semantic information that is crucial to priority evaluation, the static energy potential factor of each semantic encoding feature vector of the missing order is calculated to obtain a set of static energy potential factors of the missing orders. Then, in order to evaluate the semantic connection and interaction between each independent semantic encoding feature vector of the missing order and the initial aggregation point, the dynamic energy potential factor of each semantic encoding feature vector of the missing order relative to the initial aggregate representation vector of the missing order is calculated to obtain a set of dynamic energy potential factors of the missing orders, which can measure the difference between each order and the overall average feature, so as to better capture and understand the relative relationship between the data. Afterwards, based on the static potential factor and dynamic potential factor of each missing order semantic encoding feature vector, the superposition potential factor of each missing order semantic encoding feature vector is calculated to obtain a set of superposition potential factors of missing orders. In other words, the static potential factor provides a global perspective, reflecting the importance of orders in multiple dimensions; the dynamic potential factor provides a local perspective, reflecting the relative importance of orders under the overall order, and by calculating the superposition potential factor, the influence of a single order semantic feature in the entire context can be more comprehensively reflected. Next, in order to dynamically adjust the weight of each missing order, so as to more accurately reflect its importance and urgency in actual processing, the set of superposition potential factors of the missing order is input into the potential factor gated weighted network to dynamically adjust its weight according to the superposition potential factor of each missing order, and obtain a set of superposition potential weight coefficients of semantic features of missing orders. Finally, based on the set of superposition potential weight coefficients of semantic features of missing orders, the position-weighted sum of the set of semantic encoding feature vectors of missing orders is calculated to obtain the semantic kernel clustering representation vector of missing orders. In this way, the characteristics of all the missing orders are combined to more comprehensively reflect the feature distribution of the entire missing order set, and the order vector can be used as a representative vector of the missing order set, laying a data foundation for the subsequent semantic priority offset calculation.

[0049] In the above-mentioned order return inventory control method, the S64 calculates the semantic priority offset between each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors and the missing order semantic kernel clustering representation to obtain a set of semantic priority offsets. It should be understood that in order to more accurately evaluate the priority of each missing order and ensure that high-priority orders can be processed in a timely manner, in the technical solution of the present application, the semantic priority offset between each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors and the missing order semantic kernel clustering representation vector is calculated to obtain a set of semantic priority offsets. In this way, the difference between each order and the overall order characteristics can be quantified, so as to help managers intuitively understand the priority distribution of missing orders, thereby better allocating resources and scheduling tasks. Optionally, in one embodiment of the present application, the semantic priority offset between each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors and the missing order semantic kernel clustering representation is calculated to obtain a set of semantic priority offsets, including: calculating the cosine similarity between each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors and the missing order semantic kernel clustering representation, defining the negative value of the cosine similarity as the semantic priority offset to obtain a set of semantic priority offsets.

[0050] In the above-mentioned order return inventory control method, the S65 determines the priority ranking of the set of shortage orders based on the ranking of the set of semantic priority deviations. That is, the priority ranking of the set of shortage orders is intelligently determined by measuring the semantic priority deviation between each shortage order semantic encoding feature vector in the set of shortage order semantic encoding feature vectors and the shortage order semantic kernel clustering representation vector. In this way, the time and energy consumption of manual review is reduced, the review speed is accelerated, and the processing efficiency is improved. And the priority can be adjusted in real time to ensure the timeliness and accuracy of inventory management. Optionally, in one embodiment of the present application, the larger the semantic priority deviation, the higher the priority. It should be understood that the larger the semantic priority deviation, the greater the difference between the characteristics of the shortage order and the characteristics of the overall shortage order set. In inventory management and order processing, orders with greater differences are generally considered to be higher priority orders. This is because these orders may involve special circumstances or emergencies and need to be handled with priority.

[0051] Preferably, calculating the semantic priority offset between each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors and the missing order semantic kernel clustering representation vector to obtain a set of semantic priority offsets includes: cascading each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors into a missing order semantic encoding cascade feature vector; determining the number of zero eigenvalues ​​in the missing order semantic encoding cascade feature vector, and respectively calculating the reciprocal of the logarithm of the number of zero eigenvalues ​​with base two and the exponential value of the reciprocal of the number of zero eigenvalues ​​with base natural constant to obtain a first missing order semantic encoding cascade representation value and a second missing order semantic encoding cascade representation value; calculating the power function of each eigenvalue of the missing order semantic encoding cascade feature vector with the reciprocal of the number of zero eigenvalues ​​as the exponent to obtain a missing order semantic encoding cascade structure state vector ; Perform point multiplication on the state vector of the missing order semantic encoding cascade structure and the first missing order semantic encoding cascade representation value to obtain the missing order semantic encoding cascade information representation vector; perform point multiplication on the autocorrelation matrix of the missing order semantic encoding cascade structure state vector and the second missing order semantic encoding cascade representation value and weight hyperparameter respectively to obtain the missing order semantic encoding cascade regression understanding matrix; perform matrix multiplication on the missing order semantic encoding cascade information representation vector and the missing order semantic encoding cascade regression understanding matrix to obtain the optimized missing order semantic encoding cascade feature vector; split the optimized missing order semantic encoding cascade feature vector into individual optimized missing order semantic encoding feature vectors; and calculate the semantic priority offset between each optimized missing order semantic encoding feature vector and the missing order semantic kernel clustering representation vector to obtain a set of semantic priority offsets.

[0052] That is, the first loss order semantic encoding cascade representation value and the second loss order semantic encoding cascade representation value are expressed as: ;in, is the number of zero eigenvalues ​​in the concatenated feature vector of the semantic encoding of the shortage order, is the semantic encoding cascade representation value of the first loss order, is the second disk loss order semantic encoding cascade representation value, is an exponential function with a natural constant as the base; and the optimized semantic encoding cascade feature vector of the loss order is expressed as: ;in, is the semantic encoding cascade feature vector of the shortage order, is the weight hyperparameter, is the semantic encoding cascade regression understanding matrix of the loss order, is the semantic encoding cascade information representation vector of the loss order, is the number of zero eigenvalues ​​in the concatenated feature vector of the semantic encoding of the shortage order, is the semantic encoding cascade representation value of the first loss order, is the second disk loss order semantic encoding cascade representation value, represents a power function that calculates each eigenvalue of the cascade eigenvector of the semantic encoding of the short-selling order with the inverse of the number of zero eigenvalues ​​as the exponent, represents the transpose of a vector, It is the point multiplication by position. is matrix multiplication, It is the optimized semantic encoding cascade feature vector of the shortage order.

[0053] Here, in the case where each of the semantic encoding feature vectors of the shortage order in the set of semantic encoding feature vectors of shortage order respectively represents the encoded semantic features of the set of shortage orders in the local sample domain, when performing dynamic feature kernel clustering analysis and semantic priority offset calculation, each semantic priority offset between the semantic encoding feature vectors of shortage order and the semantic kernel clustering representation vector of shortage order will also have a remote global sample domain regression understanding representation imbalance relative to the feature distribution field selective state space of the local sample domain, thereby deviating from the semantic offset mapping comprehension of the source semantic distribution representation, affecting the calculation accuracy of the set of semantic priority offsets.

[0054] Based on this, the applicant of the present application realizes effective long-range modeling of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors under the field linear complexity by modeling the vector field zero dimension of the missing order semantic encoding cascade feature vector obtained by cascading each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors as a global field redundant dependency, so as to perform structured correlation self-distillation between information representation and regression understanding in the selective state space based on feature distribution in the feature mapping regression process of the missing order semantic encoding cascade feature vector, thereby maintaining the overall information complexity of the feature distribution of the missing order semantic encoding cascade feature vector, while promoting the remote dynamic regression understanding balance of its feature convergence, so as to improve the understanding consistency between the semantic offset mapping regression and the feature extraction process, and improve the calculation accuracy of the set of semantic priority offsets. In this way, the time and energy consumption of manual review is reduced, the review speed is accelerated, and the processing efficiency is improved. And the priority can be adjusted in real time to ensure the timeliness and accuracy of inventory management.

[0055] In summary, according to the returned order inventory control method of the embodiment of the present application, it is explained that in determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order, the technical concept of the present application is to obtain the set of the shortage orders, and use data analysis and coding technology based on deep learning to semantically encode each shortage order, and then perform dynamic feature kernel clustering analysis on the semantic coding features of each shortage order, so as to intelligently determine the priority ranking of the set of shortage orders according to each semantic priority offset measure between the semantic coding features of each shortage order and the semantic kernel clustering representation of the shortage order. In this way, the time and energy consumption of manual review is reduced, the review speed is accelerated, and the processing efficiency is improved. And the priority can be adjusted in real time to ensure the timeliness and accuracy of inventory management.

[0056] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0057] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0058] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0059] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0060] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the device claim can also be implemented by one unit through software or hardware.

[0061] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for controlling inventory returned by order return, characterized in that: include: Get sales orders pushed by the platform; The sales order is transmitted to the inventory system for reservation, wherein if the available inventory is sufficient, the sales order is determined to be successfully reserved; if the available inventory is insufficient, the sales order is determined to have failed to be reserved, and the sales order that has failed to be reserved is determined as an out-of-stock order; Get the shortage orders created by warehouse staff; The shortage order is transmitted to the inventory system for reservation, wherein if the available inventory is insufficient, it is determined that the shortage order has failed to be reserved, and the shortage order that has failed to be reserved is confirmed as a stock-out order; In response to receiving a cancel order operation of a successfully reserved sales order, the inventory system returns the previously occupied reservable inventory; Based on the priority of the out-of-stock order, determine the priority ownership of the occupied reservable inventory; Wherein, based on the priority of the out-of-stock order, determining the priority ownership of the occupied reservable inventory includes: Obtaining a collection of the shortage orders; Performing semantic coding on each shortage order in the set of shortage orders to obtain a set of semantic coding feature vectors of shortage orders; Performing dynamic feature kernel clustering analysis on the set of semantic encoding feature vectors of the shortage orders to obtain a semantic kernel clustering representation of the shortage orders; Calculating the semantic priority offset between each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors and the missing order semantic kernel clustering representation to obtain a set of semantic priority offsets; Determining the priority ranking of the set of shortage orders based on the ranking of the set of semantic priority deviations; Among them, the semantic priority offset between each semantic encoding feature vector of the shortage order in the set of the semantic encoding feature vectors of the shortage order and the semantic kernel clustering representation of the shortage order is calculated to obtain a set of semantic priority offsets, including: calculating the cosine similarity between each semantic encoding feature vector of the shortage order in the set of the semantic encoding feature vectors of the shortage order and the semantic kernel clustering representation of the shortage order, defining the negative value of the cosine similarity as the semantic priority offset to obtain a set of semantic priority offsets.

2. The method for controlling inventory returned by returned orders according to claim 1, characterized in that: In determining the priority ownership of the occupied reservable inventory based on the priority of the out-of-stock order, the priority of the out-of-stock order is greater than the priority of the sales order.

3. The method for controlling inventory returned by order return according to claim 2, characterized in that: Performing dynamic feature kernel clustering analysis on the set of semantic encoding feature vectors of the missing orders to obtain a semantic kernel clustering representation of the missing orders, including: The positional mean vector of the set of semantic encoding feature vectors of the missing orders is calculated to obtain the initial aggregate representation vector of the missing orders, which can be expressed by the formula: ;in, is a set of semantic encoding feature vectors of the shortage order, and are the first, second, and third in the set of semantic encoding feature vectors of the shortage order. and The semantic encoding feature vector of the loss order, is the number of vectors in the set of semantic encoding feature vectors of the missing orders, is the initial aggregate representation vector of the loss orders; Calculating the static energy potential factor of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors to obtain a set of missing order static energy potential factors; Calculating the dynamic energy potential factor of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors relative to the missing order initial aggregate representation vector to obtain a set of missing order dynamic energy potential factors; Based on the static energy potential factor and the dynamic energy potential factor of each semantic encoding feature vector of the shortage order in the set of semantic encoding feature vectors of the shortage order, determine the superposition state energy potential factor of each semantic encoding feature vector of the shortage order in the set of semantic encoding feature vectors of the shortage order to obtain a set of superposition state energy potential factors of the shortage order; Inputting the set of the superposition state energy potential factors of the shortage order into the energy potential factor gated weighting network to obtain a set of superposition energy potential weight coefficients of the semantic features of the shortage order; Based on the set of potential weight coefficients of the semantic features of the shortage orders, a position-weighted sum of the set of semantic encoding feature vectors of the shortage orders is calculated to obtain a shortage order semantic kernel cluster representation vector as the shortage order semantic kernel cluster representation; The step of calculating the static potential factor of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors to obtain a set of missing order static potential factors includes: Calculating the mean and standard deviation of the semantic encoding feature vector of the missing order to obtain the semantic mean of the missing order and the semantic standard deviation of the missing order; Calculate the positional subtraction between the semantic encoding feature vector of the missing order and the semantic mean of the missing order, divide the semantic offset vector of the missing order after the subtraction by the semantic standard deviation of the missing order, and calculate the fourth power of the eigenvalue of each position in the vector after the division to obtain the static energy potential vector of the semantic missing order; Calculate the positional mean of the semantic static energy potential vector of the shortage order to obtain the static energy potential factor of the shortage order; The step of calculating the dynamic potential factor of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors relative to the missing order initial aggregate representation vector to obtain a set of missing order dynamic potential factors includes: Performing matrix multiplication of the loss order aggregation weight matrix and the loss order initial aggregation representation vector to obtain the loss order initial aggregation weight representation vector; Performing matrix multiplication of the missing order weight matrix and the missing order semantic encoding feature vector to obtain the missing order semantic encoding weight vector; After performing positional dot multiplication of the initial aggregated weight representation vector of the shortage order and the semantic coding weight vector of the shortage order, the obtained semantic coding fusion weight vector of the shortage order and the shortage order bias vector are added by position to obtain the semantic dynamic energy potential vector of the shortage order; Calculating the square of the Frobenius norm of the semantic dynamic energy potential vector of the deficit order to obtain the dynamic energy potential factor of the deficit order; Wherein, based on the static energy potential factor and the dynamic energy potential factor of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors, determining the superposition state energy potential factor of each missing order semantic encoding feature vector in the set of missing order semantic encoding feature vectors to obtain a set of missing order superposition state energy potential factors, including: Calculate an exponential function with the natural constant e as the base and the static energy potential factor of the semantic encoding feature vector of the order deficit order as the exponent to obtain the static energy potential factor of the semantic bias of the order deficit order; Calculate an exponential function with the natural constant e as the base and the square of the dynamic energy potential factor of the semantic encoding feature vector of the order deficit order as the exponent to obtain the dynamic energy potential factor of the semantic bias of the order deficit order; The static potential factor of the semantic bias of the shortage order and the dynamic potential factor of the semantic bias of the shortage order are added to obtain the superposition potential factor of the semantic encoding feature vector of the shortage order.

4. The method for controlling inventory returned by order return according to claim 3, characterized in that: The set of the superposition state energy potential factors of the shortage order is input into the energy potential factor gated weighting network to obtain a set of superposition energy potential weight coefficients of the semantic features of the shortage order, including: Normalizing the set of the short-order superposition state energy potential factors to obtain a set of normalized short-order superposition state energy potential factors; The set of normalized short-order superposition state energy potential factors is input into the mask function for gated mask screening to obtain a set of short-order semantic feature superposition energy potential weight coefficients, wherein, in response to each normalized short-order superposition state energy potential factor in the set of normalized short-order superposition state energy potential factors being greater than a preset threshold, the normalized short-order superposition state energy potential factor is set to the original value, and the rest are set to zero to obtain the set of short-order semantic feature superposition energy potential weight coefficients.

5. The method for controlling inventory of returned orders according to claim 4, characterized in that: The set of the short-order superposition state energy potential factors is normalized to obtain a set of normalized short-order superposition state energy potential factors, including: Calculate the exponential function value with the natural constant e as the base and the negative number of each loss order superposition state energy potential factor in the set of loss order superposition state energy potential factors as the exponent to obtain a set of loss order superposition state energy potential exponential factors; After adding the set of the deficit order superposition state energy potential index factors and the constant one by position, the inverse of each deficit order superposition state energy potential index modulation factor in the set of deficit order superposition state energy potential index modulation factors is calculated to obtain the set of normalized deficit order superposition state energy potential factors.

Images