A method for processing hot stock returns

Abstract

The present application relates to the technical field of inventory management, and specifically discloses a hot spot inventory return processing method, which includes: obtaining a set of commodity order operations, and extracting new scheduled orders and canceled orders from them to form subsets respectively, then setting permissions according to order status, and setting the canceled order status to "to be increased inventory"; in response to receiving an instruction to batch process order inventory, extracting a subset of canceled orders marked as "to be increased inventory"; based on the subset of canceled orders marked as "to be increased inventory", calculating the total inventory that needs to be increased, and performing an inventory increase operation on the inventory system; finally, marking the status of each canceled order in the subset of canceled orders as "inventory has been increased". In this way, the operation of releasing inventory is improved from one operation to a batch operation, so as to greatly reduce the lock preemption of hot spot inventory, thereby greatly improving the order processing speed in the multi-order concurrent scenario.

Description

Technical Field

[0001] The present application relates to the technical field of inventory management, and more specifically, to a method for processing hot spot inventory returns. Background Art

[0002] In e-commerce platforms, merchants often launch some hot-selling products. These products are highly sought after by consumers, which will lead to a large number of orders pouring into the system in a short period of time, and a large number of orders that have been successfully booked will be cancelled. This high-concurrency order processing demand brings great challenges to the inventory management system, especially for the management of hot-selling products (i.e. hot inventory).

[0003] In traditional hot inventory management, popular products (hot inventory) will face a large number of order operations, including booking new orders and canceling pre-booked orders. These operations require updating the inventory. Specifically, when an order needs to operate the inventory, the inventory record will be locked first, and other orders must wait for the lock to be released before continuing to operate.

[0004] However, although this mechanism can ensure data consistency, it can lead to serious lock contention problems in high-concurrency situations. Specifically, lock contention can lead to longer order processing time, especially when the order volume is very large, where a large number of orders may be waiting in line, thus affecting the system's response speed and performance. In addition, long-term lock holding will block other key operations, resulting in a waste of system resources and overall performance degradation. In extreme cases, the system may be paralyzed due to the inability to process the backlog of orders in a timely manner, seriously affecting the user experience and merchant operation efficiency.

[0005] Therefore, a hot stock return processing solution is expected Summary of the invention

[0006] The present application provides a hot spot inventory return processing method, which can monitor the changes of inventory and orders in real time, so as to quickly identify the changes of system status, generate processing prompts in time, and avoid the lag of manual judgment. At the same time, it can effectively respond to large-scale and complex changes, thereby realizing real-time monitoring and intelligent judgment of inventory and order changes.

[0007] In a first aspect, a hot spot inventory return processing method is provided, comprising: obtaining multiple commodity order operations issued by a platform at the same time to obtain a set of commodity order operations; extracting new scheduled orders and canceled orders from the set of commodity order operations to obtain a subset of new scheduled orders and a subset of canceled orders; setting the authority of the subset of canceled orders to not directly operate the inventory system, and each new scheduled order in the subset of new scheduled orders seizes a pessimistic lock and directly reserves successfully to complete the increase operation of the inventory system; marking the status of each canceled order in the subset of canceled orders as "inventory to be increased"; in response to receiving an instruction to batch process order inventory, extracting the subset of canceled orders marked as "inventory to be increased"; based on the subset of canceled orders marked as "inventory to be increased", calculating the total inventory that needs to be increased, and performing an inventory increase operation on the inventory system; marking the status of each canceled order in the subset of canceled orders as "inventory has been increased".

[0008] In one possible implementation, in response to receiving an instruction to process order inventory in batches, extracting a subset of the canceled orders marked as "inventory to be added", including: extracting a time queue of the remaining inventory of the inventory system; extracting a time queue of the total order volume of the subset of canceled orders; performing sequence encoding on the time queue of the remaining inventory and the time queue of the total order volume respectively to obtain a time series fluctuation characteristic of the remaining inventory and a time series fluctuation characteristic of the canceled order volume; performing interactive compensation of the time series common characteristics on the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order volume to obtain a time series collaborative interactive compensation representation of the remaining inventory-cancelled order volume; and obtaining a monitoring result based on the time series collaborative interactive compensation representation of the remaining inventory-cancelled order volume, wherein the monitoring result is used to indicate whether to generate a prompt for recommending "batch processing of order inventory".

[0009] In a possible implementation, the time queue of the remaining inventory and the time queue of the total order volume are respectively sequence-encoded to obtain the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order volume, including: inputting the time queue of the remaining inventory and the time queue of the total order volume into a sequence encoder based on a forward LSTM model respectively to obtain the time series fluctuation feature vector of the remaining inventory as the time series fluctuation feature of the remaining inventory and the time series fluctuation feature vector of the canceled order volume as the time series fluctuation feature of the canceled order volume.

[0010] In a possible implementation, the remaining inventory time series fluctuation characteristics and the canceled order quantity time series fluctuation characteristics are interactively compensated for their time series common features to obtain a remaining inventory-cancelled order quantity time series collaborative interactive compensation representation, including: inputting the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector into a common feature extraction network to obtain a common feature representation vector between the remaining inventory-cancelled order quantity time series features; based on the common feature representation vector between the remaining inventory-cancelled order quantity time series features, performing feature fine-grained compensation on the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector to obtain the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation.

[0011] In a possible implementation, the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector are input into a common feature extraction network to obtain a common feature representation vector between the remaining inventory and the canceled order time series features, including: adding the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector according to position points to obtain a remaining inventory-cancelled order time series point addition vector; multiplying the remaining inventory-cancelled order time series point addition vector with a weight matrix, adding the obtained remaining inventory-cancelled order time series weight vector and the bias vector according to position to obtain a remaining inventory-cancelled order time series commonality vector; inputting the remaining inventory-cancelled order time series commonality vector into a tanh function for processing to obtain a common feature representation vector between the remaining inventory-cancelled order time series features.

[0012] In a possible implementation, based on the common feature representation vector between the remaining inventory and the cancelled order quantity time series features, the remaining inventory time series fluctuation feature vector and the cancelled order quantity time series fluctuation feature vector are subjected to feature fine-grained compensation to obtain the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation, including: inputting the remaining inventory time series fluctuation feature vector, the cancelled order quantity time series fluctuation feature vector and the common feature representation vector between the remaining inventory-cancelled order quantity time series features into a vector probabilization unit based on the Sigmoid function to obtain a probabilistic remaining inventory time series fluctuation feature vector, a probabilistic cancelled order quantity time series fluctuation feature vector and a probabilistic remaining inventory-cancelled order quantity time series common feature representation vector; calculating the remaining inventory time series fine-grained compensation of the probabilistic remaining inventory time series fluctuation feature vector relative to the probabilistic remaining inventory-cancelled order quantity time series common feature representation vector. The method comprises the following steps: calculating a compensation vector for the time series of canceled orders of the probabilistic cancellation order quantity time series fluctuation feature vector relative to the probabilistic common feature representation vector between the time series features of the remaining inventory and the cancellation order quantity; inputting the time series fine-grained compensation vector of the remaining inventory and the time series fluctuation feature vector of the remaining inventory into the fine-grained compensation module to obtain the time series fine-grained feature compensation vector of the remaining inventory; inputting the time series fine-grained compensation vector of the canceled order quantity and the time series fluctuation feature vector of the canceled order quantity into the fine-grained compensation module to obtain the time series fine-grained feature compensation vector of the canceled order quantity; cascading the time series fine-grained feature compensation vector of the remaining inventory, the time series fine-grained feature compensation vector of the canceled order quantity and the common feature representation vector between the time series features of the remaining inventory and the cancellation order quantity to obtain the time series collaborative interaction compensation representation vector of the remaining inventory and the cancellation order quantity as the time series collaborative interaction compensation representation of the remaining inventory and the cancellation order quantity.

[0013] In a possible implementation, the remaining inventory time series fine-grained compensation vector of the probabilistic remaining inventory time series fluctuation feature vector relative to the probabilistic remaining inventory-cancelled order time series feature common feature representation vector is calculated, including: calculating the position point division of the probabilistic remaining inventory time series fluctuation feature vector and the probabilistic remaining inventory-cancelled order time series feature common feature representation vector to obtain the remaining inventory time series contribution vector; calculating the logarithmic function value with base two of the absolute value of each eigenvalue of the remaining inventory time series contribution vector to obtain the remaining inventory time series contribution logarithmic vector; calculating the position point multiplication of the probabilistic remaining inventory time series fluctuation feature vector and the remaining inventory time series contribution logarithmic vector to obtain a point product vector, and calculating the exponential function value of each point product value in the point product vector with base e as the natural constant to obtain the remaining inventory time series compensation vector; inputting the remaining inventory time series compensation vector into the softmax function to obtain the remaining inventory time series fine-grained compensation vector.

[0014] In a possible implementation, the remaining inventory time series fine-grained compensation vector and the remaining inventory time series fluctuation characteristic vector are input into a fine-grained compensation module to obtain the remaining inventory time series fine-grained characteristic compensation vector, including: calculating the remaining inventory time series fluctuation characteristic vector and the remaining inventory time series fluctuation characteristic vector by position point multiplication to obtain the remaining inventory time series fine-grained characteristic compensation vector.

[0015] In a possible implementation, a monitoring result is obtained based on the remaining inventory-cancelled order timing series collaborative interaction compensation representation, including: inputting the remaining inventory-cancelled order timing series collaborative interaction compensation representation vector into a classifier-based intelligent monitor to obtain a monitoring result.

[0016] The present application provides a method for processing the return of hot stock, which extracts the time queue of the remaining inventory of the inventory system, and extracts the time queue of the total order amount of the subset of canceled orders, and uses artificial intelligence-based data analysis and coding methods to perform time series coding of the remaining inventory and the total order amount, thereby intelligently judging whether to generate a prompt for recommending "batch processing of order inventory" based on the temporal commonality interactive compensation representation between the temporal characteristics of the remaining inventory and the temporal characteristics of the canceled orders. In this way, changes in inventory and orders can be monitored in real time to quickly identify changes in system status, generate processing prompts in a timely manner, and avoid delays in manual judgment. At the same time, it can effectively respond to large-scale and complex changes, thereby realizing real-time monitoring and intelligent judgment of changes in inventory and orders. 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 a hot spot inventory return processing method according to an embodiment of the present application.

[0019] Figure 2 A flow chart of a hot spot inventory return processing method according to an embodiment of the present application.

[0020] Figure 3 This is an order processing flow chart of the hot spot inventory return processing method according to an embodiment of the present application.

[0021] Figure 4 In the hot spot inventory return processing method of an embodiment of the present application, in response to receiving an instruction to batch process order inventory, a schematic flowchart of extracting a subset of the canceled orders marked as "inventory to be added" is provided.

[0022] Figure 5 In the hot spot inventory return processing method of an embodiment of the present application, in response to receiving an instruction to batch process order inventory, a data flow diagram of a subset of the canceled orders marked as "inventory to be added" is extracted.

[0023] Figure 6 In the hot stock return processing method of an embodiment of the present application, the timing common characteristics of the remaining inventory timing fluctuation characteristics and the canceled order quantity timing fluctuation characteristics are interactively compensated to obtain a schematic flowchart representing the remaining inventory-cancelled order quantity timing collaborative interactive compensation. DETAILED DESCRIPTION

[0024] Since the field of inventory management 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.

[0025] 1. Inventory: The virtual count of goods in the inventory system. Inventory control is used to prevent overselling of goods. Different goods have their own inventory. 2. Hot inventory: Some goods are hot-selling goods, and the corresponding inventory is hot inventory, which will increase and decrease frequently. 3. Order reservation: After the order enters the system, it is necessary to reserve the inventory of each item in the order first, that is, deduct the inventory of the corresponding goods. If the order reservation inventory fails, the fulfillment will not be completed. 4. Order cancellation: The reserved order can be canceled, and the reserved inventory of each item will be returned, that is, the inventory of the corresponding item will be increased. 5. Hot inventory: The inventory of hot-selling goods, because it is a hot-selling item, often has multiple orders booked at the same time, and multiple orders are canceled. 6. Pessimistic lock: A locking method of the database to ensure that there is no problem with the data. When an order operates an inventory, it first locks this inventory. If other orders also need to operate this inventory, they must wait. After the order operation is completed, the lock is released, and other orders compete for the lock again. Only the order that successfully seizes the lock can operate this inventory.

[0026] 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.

[0027] Based on this, this application proposes a hot stock return processing method, which improves the operation of releasing inventory from one operation to a batch operation, so as to greatly reduce the lock preemption of hot stock, thereby greatly improving the order processing speed in the multi-order concurrent scenario, such as Figure 1-Figure 3 As shown, the hot spot inventory return processing method includes: S1, obtaining multiple commodity order operations issued by the platform at the same time to obtain a set of commodity order operations; S2, extracting new scheduled orders and canceled orders from the set of commodity order operations to obtain a subset of new scheduled orders and a subset of canceled orders; S3, setting the authority of the subset of canceled orders to not directly operate the inventory system, and each new scheduled order in the subset of new scheduled orders seizes the pessimistic lock and directly reserves successfully to complete the increase operation of the inventory system; S4, marking the status of each canceled order in the subset of canceled orders as "inventory to be increased"; S5, in response to receiving an instruction to batch process order inventory, extracting the subset of canceled orders marked as "inventory to be increased"; S6, based on the subset of canceled orders marked as "inventory to be increased", calculating the total inventory that needs to be increased, and performing an inventory increase operation on the inventory system; S7, marking the status of each canceled order in the subset of canceled orders as "inventory has been increased".

[0028] In the above-mentioned hot stock return processing method, S1 obtains multiple commodity order operations issued by the platform at the same time to obtain a set of commodity order operations. It should be understood that on the e-commerce platform, especially during promotional activities, a large number of orders will flood into the system in a short period of time. These orders may include reservations for new orders and cancellations of scheduled orders. If these order operations are not processed centrally, each order needs to seize pessimistic locks separately, resulting in a large number of queuing phenomena, which seriously affects the response speed and performance of the system. By centrally processing multiple order operations, the lock competition of the database can be reduced and the concurrent processing capability of the system can be improved. Batch processing of orders can reduce frequent access to the database, reduce system load, and improve overall performance.

[0029] Specifically, after receiving the user's order request, the e-commerce platform (such as a shopping website, mobile application, etc.) will aggregate these requests and send them to the backend system. These order requests may include booking operations for new orders and cancellation operations for scheduled orders. The backend system (such as the order management system OMS) receives the order requests issued by the platform and stores these requests in a collection. The collection can be a memory queue, message queue, or other data structure for temporary storage and management of these order operations. For example, the platform sends multiple order operations for product A at the same time, including the entry of a new order for order 1, and the cancellation of scheduled orders for order 2, order 3, ..., and order X.

[0030] In the above-mentioned hot stock return processing method, S2 extracts new scheduled orders and canceled orders from the set of commodity order operations to obtain a subset of new scheduled orders and a subset of canceled orders. It should be understood that by dividing the order operations into two subsets, new scheduled orders and canceled orders, different types of operations can be processed separately to improve processing efficiency. For example, order 1 needs to be booked and needs to operate inventory a for deduction, and order 2, order 3, ..., order X need to cancel orders, and they all need to operate inventory a and perform an increase operation.

[0031] In the above hot stock return processing method, S3 and S4 set the permission of the subset of the canceled orders to not directly operate the inventory system, and each new scheduled order in the subset of the new scheduled orders seizes the pessimistic lock and directly reserves successfully to complete the increase operation of the inventory system, and marks the status of each canceled order in the subset of the canceled orders as "to be increased inventory". It should be understood that the new scheduled order needs to deduct the inventory immediately, so it is necessary to seize the pessimistic lock. If all orders seize the lock together, it will cause a large number of queues, affecting the processing speed. Cancelled orders can be recorded as "to be increased inventory", and there is no need to seize the lock immediately. They can be processed in batches, reduce lock competition, and improve the concurrent processing capability of the system. In addition, the operation of canceling an order is to increase inventory, which theoretically will not cause the problem of insufficient inventory. If each canceled order immediately seizes the pessimistic lock and updates the inventory, it will cause a large amount of lock competition, affecting the concurrent processing capability of the system. Marking the canceled order as "to be increased inventory" can delay the actual inventory increase operation, reduce lock competition, and improve the concurrent processing capability of the system. For example, order 1 starts to seize the pessimistic lock. At this time, order 2, order 3, ..., order X will not directly operate inventory a, so order 1 seizes the lock directly and successfully completes the increase operation of inventory a, and the reservation is successful. Order 2, order 3, ..., order X are all canceled orders. Canceling orders increases inventory. In theory, increasing inventory cannot be a negative number and will not fail. Therefore, only records of order 2, order 3, ..., order X are inserted at this time, and the record status is "to be increased inventory", and the quantity of inventory a operated by the corresponding order is recorded 2', 3', X'.

[0032] In the above-mentioned hot spot inventory return processing method, S5 and S6, in response to receiving the instruction of batch processing order inventory, extract the subset of the canceled orders marked as "inventory to be increased", and based on the subset of the canceled orders marked as "inventory to be increased", calculate the total inventory that needs to be increased, and perform an inventory increase operation on the inventory system. For example, the scheduling system triggers the batch processing of the inventory that needs orders, picks up the records of "inventory to be increased" Order 2, Order 3, ..., Order X, and calculates the total inventory that needs to be increased sum = 2'+3'+...+X'. And perform an inventory increase operation on inventory a, and the increase quantity is the result sum of the above calculation.

[0033] In the above hot spot inventory return processing method, S7 marks the status of each canceled order in the subset of canceled orders as "inventory has been added". For example, all the records of "to be added inventory" Order 2, Order 3, ..., Order X are marked as "inventory has been added". At this point, all order operations are successful. In this same scenario, the inventory operations of canceled orders are combined into one.

[0034] It should be understandable that when dealing with hot inventory, in the face of multiple orders, the inventory increase operation of canceling orders will be processed in batches to avoid each order from occupying the pessimistic lock separately, thereby greatly reducing lock contention and improving the concurrent processing capability of the system. Traditionally, the administrator manually triggers the "instruction for batch processing of order inventory" according to actual conditions. For example, when it is detected that the system load is high or there is a large backlog of orders, batch processing is manually triggered to relieve pressure. However, this manual method often has a lag in judgment and cannot respond to system changes in real time. Especially in high concurrency situations, the order backlog may increase rapidly, and the administrator may fail to detect and take action in time. In addition, in large-scale, high-concurrency systems, manual triggering methods are difficult to cope with complex changes in order inventory status, and cannot achieve efficient resource management and performance optimization.

[0035] Therefore, in response to receiving an instruction to process order inventory in batches, extracting the subset of canceled orders marked as "inventory to be added", the technical concept of the present application is to extract the time queue of the remaining inventory of the inventory system, and extract the time queue of the total order amount of the subset of canceled orders, and use artificial intelligence-based data analysis and encoding methods to perform time series encoding of the remaining inventory and the total order amount, so as to intelligently determine whether to generate a prompt to recommend "batch processing order inventory" based on the temporal commonality interactive compensation representation between the temporal characteristics of the remaining inventory and the temporal characteristics of the canceled order. In this way, the changes in inventory and orders can be monitored in real time to quickly identify changes in system status, generate processing prompts in a timely manner, and avoid the lag of manual judgment. At the same time, it can effectively respond to large-scale and complex changes, thereby realizing real-time monitoring and intelligent judgment of inventory and order changes.

[0036] Figure 4 In the hot spot inventory return processing method of an embodiment of the present application, in response to receiving an instruction to batch process order inventory, a schematic flowchart of extracting a subset of the canceled orders marked as "inventory to be added" is provided. Figure 5 In the hot spot inventory return processing method of the embodiment of the present application, in response to receiving an instruction for batch processing of order inventory, a data flow diagram of extracting a subset of the canceled orders marked as "inventory to be added" is provided. Figure 4 and Figure 5As shown, in response to receiving an instruction for batch processing of order inventory, extracting the subset of canceled orders marked as "inventory to be added", including: S51, extracting the time queue of the remaining inventory of the inventory system; S52, extracting the time queue of the total order volume of the subset of canceled orders; S53, respectively performing sequence encoding on the time queue of the remaining inventory and the time queue of the total order volume to obtain the timing fluctuation characteristics of the remaining inventory and the timing fluctuation characteristics of the canceled order volume; S54, performing interactive compensation of the timing common characteristics on the timing fluctuation characteristics of the remaining inventory and the timing fluctuation characteristics of the canceled order volume to obtain a timing collaborative interactive compensation representation of the remaining inventory-cancelled order volume; S55, obtaining a monitoring result based on the timing collaborative interactive compensation representation of the remaining inventory-cancelled order volume, and the monitoring result is used to indicate whether to generate a prompt for recommending "batch processing of order inventory".

[0037] In the above-mentioned hot spot inventory return processing method, the S51 extracts the time queue of the remaining inventory of the inventory system. It should be understood that by extracting the time queue of the remaining inventory of the inventory system, the changes in inventory can be monitored in real time, and abnormal fluctuations in inventory can be discovered in time. This helps system administrators or automated systems to adjust inventory strategies in a timely manner to avoid problems of insufficient or excessive inventory. At the same time, the time queue data can be used to predict future inventory demand. By analyzing the historical inventory change trend, the inventory demand in a certain period of time in the future can be predicted, so that inventory preparations can be made in advance.

[0038] In the above-mentioned hot stock return processing method, the S52 extracts the time queue of the total order amount of the subset of the canceled orders. It should be understood that by extracting the time queue of the total order amount of the subset of the canceled orders, the changes in the canceled orders can be monitored in real time, and abnormal fluctuations in orders can be discovered in time. This helps the system administrator or the automation system to adjust the order processing strategy in time to avoid the problem of order backlog or untimely processing. The time queue data can be used to predict future order trends. By analyzing the trend of historical order changes, the order volume in a certain time period in the future can be predicted, so as to prepare resources in advance.

[0039] In the above-mentioned hot stock return processing method, the S53, respectively, performs sequence encoding on the time queue of the remaining inventory and the time queue of the total order volume to obtain the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order volume. It should be understood that, considering that the remaining inventory and the total order volume are constantly updated and changed with the passage of time, and the mutual correlation in different time intervals is different, based on this, in order to better capture and mine the time series correlation changes and fluctuation patterns between different time spans, in the technical solution of the present application, the time queue of the remaining inventory and the time queue of the total order volume are respectively input into the sequence encoder based on the forward LSTM model to obtain the time series fluctuation feature vector of the remaining inventory and the time series fluctuation feature vector of the canceled order volume.

[0040] Optionally, in one embodiment of the present application, the time queue of the remaining inventory and the time queue of the total order volume are respectively sequence encoded to obtain the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order volume, including: inputting the time queue of the remaining inventory and the time queue of the total order volume into a sequence encoder based on a forward LSTM model respectively to obtain the time series fluctuation feature vector of the remaining inventory as the time series fluctuation feature of the remaining inventory and the time series fluctuation feature vector of the canceled order volume as the time series fluctuation feature of the canceled order volume.

[0041] In the above-mentioned hot stock return processing method, the S54 performs interactive compensation of the time series common characteristics on the remaining inventory time series fluctuation characteristics and the canceled order quantity time series fluctuation characteristics to obtain the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation. It should be understood that there is a mutual time series correlation relationship between the remaining inventory time series fluctuation characteristic vector and the canceled order quantity time series fluctuation characteristic vector. For example, a decrease in inventory in certain time periods may be accompanied by an increase in the number of canceled orders. In order to more accurately capture the temporal trends of the remaining inventory and the canceled order quantity, so as to better characterize the unique time series characteristics and interactive information between the two, in the technical solution of the present application, the remaining inventory time series fluctuation characteristic vector and the canceled order quantity time series fluctuation characteristic vector are interactively compensated for their time series common characteristics to obtain the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation vector as the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation.

[0042] Optionally, in one embodiment of the present application, Figure 6 In the hot stock return processing method of the embodiment of the present application, the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order quantity are interactively compensated for the time series common characteristics to obtain a schematic flow chart of the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation. Figure 6 As shown, interactive compensation of the common timing features of the remaining inventory time series fluctuation feature and the canceled order quantity time series fluctuation feature is performed to obtain a remaining inventory-cancelled order quantity time series collaborative interactive compensation representation, including: S541, inputting the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector into a common feature extraction network to obtain a common feature representation vector between the remaining inventory-cancelled order quantity time series features; S542, based on the common feature representation vector between the remaining inventory-cancelled order quantity time series features, performing feature fine-grained compensation on the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector to obtain a remaining inventory-cancelled order quantity time series collaborative interactive compensation representation.

[0043] Optionally, in one embodiment of the present application, Figure 6 In the hot stock return processing method of the embodiment of the present application, the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order quantity are interactively compensated for the time series common characteristics to obtain a schematic flow chart of the remaining inventory-cancelled order quantity time series collaborative interactive compensation representation. Figure 6 As shown, the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector are input into the common feature extraction network to obtain the common feature representation vector between the remaining inventory and the canceled order time series features, including: adding the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector according to the position point to obtain the remaining inventory-cancelled order time series point addition vector; multiplying the remaining inventory-cancelled order time series point addition vector with the weight matrix, adding the obtained remaining inventory-cancelled order time series weight vector and the bias vector according to the position to obtain the remaining inventory-cancelled order time series commonality vector; inputting the remaining inventory-cancelled order time series commonality vector into the tanh function for processing to obtain the common feature representation vector between the remaining inventory-cancelled order time series features.

[0044] Optionally, in one embodiment of the present application, based on the common feature representation vector between the remaining inventory and the time series features of the canceled order quantity, the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector are subjected to feature fine-grained compensation to obtain the remaining inventory-cancelled order quantity time series collaborative interaction compensation representation, including: inputting the remaining inventory time series fluctuation feature vector, the canceled order quantity time series fluctuation feature vector and the common feature representation vector between the remaining inventory-cancelled order quantity time series features into a vector probabilization unit based on the Sigmoid function to obtain a probabilistic remaining inventory time series fluctuation feature vector, a probabilistic canceled order quantity time series fluctuation feature vector and a probabilistic remaining inventory-cancelled order quantity time series common feature representation vector; calculating the remaining inventory time series fine-grainedness of the probabilistic remaining inventory time series fluctuation feature vector relative to the probabilistic remaining inventory-cancelled order quantity time series common feature representation vector. degree compensation vector; calculate the canceled order quantity time series fine-grained compensation vector of the probabilistic canceled order quantity time series fluctuation feature vector relative to the probabilistic remaining inventory-cancelled order quantity time series common feature representation vector; input the remaining inventory time series fine-grained compensation vector and the remaining inventory time series fluctuation feature vector into the fine-grained compensation module to obtain the remaining inventory time series fine-grained feature compensation vector; input the canceled order quantity time series fine-grained compensation vector and the canceled order quantity time series fluctuation feature vector into the fine-grained compensation module to obtain the canceled order quantity time series fine-grained feature compensation vector; cascade the remaining inventory time series fine-grained feature compensation vector, the canceled order quantity time series fine-grained feature compensation vector and the remaining inventory-cancelled order quantity time series common feature representation vector to obtain the remaining inventory-cancelled order quantity time series collaborative interaction compensation representation vector as the remaining inventory-cancelled order quantity time series collaborative interaction compensation representation.

[0045] Optionally, in one embodiment of the present application, the remaining inventory time series fine-grained compensation vector of the probabilistic remaining inventory time series fluctuation feature vector relative to the probabilistic remaining inventory-cancelled order time series feature common feature representation vector is calculated, including: calculating the position point division of the probabilistic remaining inventory time series fluctuation feature vector and the probabilistic remaining inventory-cancelled order time series feature common feature representation vector to obtain the remaining inventory time series contribution vector; calculating the logarithmic function value with base two of the absolute value of each eigenvalue of the remaining inventory time series contribution vector to obtain the remaining inventory time series contribution logarithmic vector; calculating the position point multiplication of the probabilistic remaining inventory time series fluctuation feature vector and the remaining inventory time series contribution logarithmic vector to obtain a point product vector, and calculating the exponential function value of each point product value in the point product vector with the natural constant e as the base to obtain the remaining inventory time series compensation vector; inputting the remaining inventory time series compensation vector into the softmax function to obtain the remaining inventory time series fine-grained compensation vector.

[0046] Optionally, in one embodiment of the present application, the remaining inventory time series fine-grained compensation vector and the remaining inventory time series fluctuation characteristic vector are input into a fine-grained compensation module to obtain a remaining inventory time series fine-grained feature compensation vector, including: calculating the remaining inventory time series fluctuation characteristic vector and the remaining inventory time series fluctuation characteristic vector by position point multiplication to obtain the remaining inventory time series fine-grained feature compensation vector.

[0047] In summary, in the embodiment of the present application, the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector are interactively compensated based on the common characteristics of the inventory-order time series, which can be expressed by the formula: ;in, and are respectively the remaining inventory time series fluctuation characteristic vector and the canceled order time series fluctuation characteristic vector, To add by location, and are the weight matrix and bias vector respectively, yes function, is the common feature representation vector between the remaining inventory and the cancelled order quantity time series features, yes function, is the probabilistic time series fluctuation characteristic vector of remaining inventory, is the time series fluctuation characteristic vector of the probabilistic order cancellation volume, is the common feature representation vector between the probabilistic remaining inventory-cancelled order quantity time series features, is the eigenvalue of each position in the probabilistic remaining inventory time series fluctuation eigenvector, is the eigenvalue of each position in the probabilistic order cancellation time series fluctuation eigenvector, is the feature value of each position in the common feature representation vector between the probabilistic remaining inventory-cancelled order quantity time series features, represents the logarithmic function value with base 2, represents the value of the natural exponential function with the natural constant e as the base, yes function, is the remaining inventory timing fine-grained compensation vector, is the fine-grained compensation vector of the order cancellation quantity timing, It is the point multiplication by position. is the fine-grained feature compensation vector of the remaining inventory quantity, is the fine-grained feature compensation vector of the order cancellation quantity time series, It is a cascade operation on vectors. is the representation vector of the remaining inventory-cancelled order timing cooperative interaction compensation.

[0048] Specifically, first, the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector are input into the common feature extraction network to highlight the common information between the two feature vectors, while retaining their own unique time series fluctuation features, and obtaining the common feature representation vector between the remaining inventory and the canceled order time series features. This ensures that the information of the two feature vectors is fully utilized during the analysis process. In particular, in this application, the two vectors can be jointly processed by using a fully connected layer. Then, a vector probabilization unit based on a Sigmoid function is used to normalize the remaining inventory time series fluctuation feature vector, the canceled order time series fluctuation feature vector, and the common feature representation vector so that it will be mapped to (0,1) between, and the eigenvalue is converted into a probability form to better measure the importance of the feature. Then, the fine-grained compensation vectors of the probabilistic remaining inventory time series fluctuation feature vector and the probabilistic canceled order time series fluctuation feature vector relative to the probabilistic remaining inventory-cancelled order time series feature common feature representation vector are calculated respectively, and the remaining inventory time series fine-grained compensation vector and the canceled order time series fine-grained compensation vector are obtained. In this way, it is possible to identify the unique information of the remaining inventory and the number of canceled orders in addition to the common features, so as to reveal the subtle information that is not reflected in the common features but is crucial to the model. Subsequently, the fine-grained compensation module is used to process the remaining inventory time series fine-grained compensation vector and the remaining inventory time series fluctuation feature vector. In other words, the fine-grained compensation module is used to enhance or supplement specific features. In this process, the module receives the original feature vector and the corresponding compensation vector, and then integrates the compensation vector into the original feature vector through a specific operation (for example, the dot product between the two vectors) to generate a remaining inventory time series fine-grained feature compensation vector that incorporates more detailed information. At the same time, the module is used to process the canceled order time series fine-grained compensation vector and the canceled order time series fluctuation feature vector to obtain the canceled order time series fine-grained feature compensation vector. Finally, the remaining inventory time series fine-grained feature compensation vector, the canceled order time series fine-grained feature compensation vector and the common feature representation vector between the remaining inventory and canceled order time series features are cascaded to obtain a remaining inventory-cancelled order time series collaborative interaction compensation representation vector that contains the common information between inventory and orders while retaining their respective unique time series fluctuation characteristics.

[0049] In the above-mentioned hot stock return processing method, the S55 obtains a monitoring result based on the remaining inventory-cancelled order time series collaborative interactive compensation representation, and the monitoring result is used to indicate whether to generate a prompt for recommending "batch processing order inventory". Optionally, in one embodiment of the present application, based on the remaining inventory-cancelled order time series collaborative interactive compensation representation, the monitoring result is obtained, including: inputting the remaining inventory-cancelled order time series collaborative interactive compensation representation vector into an intelligent monitor based on a classifier to obtain a monitoring result. That is, the remaining inventory-cancelled order time series collaborative interactive compensation representation obtained by performing common interactive compensation using the remaining inventory time series fluctuation feature vector and the canceled order time series fluctuation feature vector is classified and processed, so as to intelligently determine whether to generate a prompt for recommending "batch processing order inventory". In this way, the changes in inventory and orders can be monitored in real time to quickly identify changes in system status, generate processing prompts in a timely manner, and avoid the lag of manual judgment. At the same time, it can effectively cope with large-scale and complex changes, thereby realizing real-time monitoring and intelligent judgment of inventory and order changes.

[0050] Preferably, inputting the remaining inventory-cancelled order timing series collaborative interaction compensation representation vector into a classifier-based intelligent monitor to obtain a monitoring result includes: determining the number of zero eigenvalues ​​in the remaining inventory-cancelled order timing series collaborative interaction compensation representation 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 constants to obtain a first remaining inventory-cancelled order timing series collaborative interaction compensation representation value and a second remaining inventory-cancelled order timing series collaborative interaction compensation representation value; calculating a power function of each eigenvalue of the remaining inventory-cancelled order timing series collaborative interaction compensation representation vector with the reciprocal of the number of zero eigenvalues ​​as an exponent to obtain a remaining inventory-cancelled order timing series collaborative interaction compensation structure state vector; comparing the remaining inventory-cancelled order timing series collaborative interaction compensation structure state vector with the The first remaining inventory-cancelled order timing collaborative interaction compensation representation value is point multiplied to obtain the remaining inventory-cancelled order timing collaborative interaction compensation information representation vector; the autocorrelation matrix of the remaining inventory-cancelled order timing collaborative interaction compensation structure state vector is point multiplied with the second remaining inventory-cancelled order timing collaborative interaction compensation representation value and the weight hyperparameter respectively to obtain the remaining inventory-cancelled order timing collaborative interaction compensation regression understanding matrix; the remaining inventory-cancelled order timing collaborative interaction compensation information representation vector is matrix multiplied with the remaining inventory-cancelled order timing collaborative interaction compensation regression understanding matrix to obtain the optimized remaining inventory-cancelled order timing collaborative interaction compensation representation vector; and the optimized remaining inventory-cancelled order timing collaborative interaction compensation representation vector is input into the classifier-based intelligent monitor to obtain the monitoring result.

[0051] That is, the first remaining inventory-cancelled order timing sequence collaborative interaction compensation representation value and the second remaining inventory-cancelled order timing sequence collaborative interaction compensation representation value are expressed as: ;in, is the number of zero eigenvalues ​​in the remaining inventory-cancelled order timing cooperative interaction compensation representation vector, is the first remaining inventory quantity-cancelled order quantity timing collaborative interaction compensation representation value, is the second remaining inventory quantity-cancelled order quantity timing collaborative interaction compensation representation value, is an exponential function with a natural constant as the base; and the optimized remaining inventory-cancelled order quantity timing collaborative interaction compensation representation vector is expressed as: ;in, is the remaining inventory-cancelled order timing cooperative interaction compensation representation vector, is the weight hyperparameter, is the remaining inventory-cancelled order time series collaborative interaction compensation regression understanding matrix, is the remaining inventory-cancelled order time series collaborative interaction compensation information representation vector, is the number of zero eigenvalues ​​in the remaining inventory-cancelled order timing cooperative interaction compensation representation vector, is the first remaining inventory quantity-cancelled order quantity timing collaborative interaction compensation representation value, is the second remaining inventory quantity-cancelled order quantity timing collaborative interaction compensation representation value, represents a power function in which each eigenvalue of the remaining inventory-cancelled order timing series collaborative interaction compensation representation vector is calculated 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 representation vector of the optimized remaining inventory quantity-cancelled order quantity timing collaborative interaction compensation.

[0052] Here, in the case where the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector respectively represent the short-range-long-range bidirectional time series correlation characteristics of the remaining inventory and the total order quantity, when performing time series common feature interaction compensation, the common feature mask interaction compensation representation of the remaining inventory-cancelled order quantity time series collaborative interaction compensation representation vector will also have an interactive class regression understanding representation imbalance relative to the respective time series feature distribution field selective state space of the heterogeneous data, thereby deviating from the source data time series feature distribution representation class understanding, affecting the accuracy of the classification results.

[0053] Based on this, the applicant of the present application realizes the effective long-range modeling of the remaining inventory-cancelled order time series collaborative interaction compensation representation vector under the field linear complexity by modeling the zero dimension of the vector field of the remaining inventory-cancelled order time series collaborative interaction compensation representation vector as the global field redundant dependency, so as to perform the structured correlation self-distillation between information representation and class understanding in the selective state space based on feature distribution in the feature regression process of the remaining inventory-cancelled order time series collaborative interaction compensation representation vector, thereby maintaining the overall information complexity of the feature distribution of the remaining inventory-cancelled order time series collaborative interaction compensation representation vector, while promoting the remote dynamic class understanding balance of its feature convergence, so as to improve the class understanding consistency between the classification regression and feature extraction process, and improve the accuracy of the monitoring results obtained by the remaining inventory-cancelled order time series collaborative interaction compensation representation vector input based on the classifier intelligent monitor. In this way, the changes in inventory and orders can be monitored in real time to quickly identify changes in system status, generate processing prompts in time, and avoid the lag of manual judgment. At the same time, it can effectively cope with large-scale and complex changes, thereby realizing real-time monitoring and intelligent judgment of inventory and order changes.

[0054] In summary, according to the hot spot inventory return processing method of the embodiment of the present application, it is explained, which extracts the time queue of the remaining inventory of the inventory system, and extracts the time queue of the total order amount of the subset of canceled orders, and uses artificial intelligence-based data analysis and encoding methods to perform time series encoding of the remaining inventory and the total order amount, thereby intelligently judging whether to generate a prompt for recommending "batch processing of order inventory" based on the temporal commonality interactive compensation representation between the remaining inventory time series characteristics and the canceled order time series characteristics. In this way, the changes in inventory and orders can be monitored in real time to quickly identify changes in system status, generate processing prompts in a timely manner, and avoid the lag of manual judgment. At the same time, it can effectively respond to large-scale and complex changes, thereby realizing real-time monitoring and intelligent judgment of inventory and order changes.

[0055] 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.

[0056] 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.

[0057] 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.

[0058] 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.

[0059] 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.

[0060] 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 hot spot inventory return processing method, characterized in that: include: Obtain multiple product order operations issued by the platform at the same time to obtain a collection of product order operations; Extracting new scheduled orders and canceled orders from the set of commodity order operations to obtain a subset of new scheduled orders and a subset of canceled orders; The permission of the subset of canceled orders is set to not directly operate the inventory system, and each new scheduled order in the subset of new scheduled orders seizes the pessimistic lock and is directly booked successfully to complete the reduction operation of the inventory system; Marking the status of each canceled order in the subset of canceled orders as "to be increased inventory"; In response to receiving an instruction to process the order inventory in batches, extracting a subset of the canceled orders marked as "to be increased inventory"; Based on the subset of the canceled orders marked as "inventory to be added", the total inventory quantity to be added is calculated, and an inventory adding operation is performed on the inventory system; Marking the status of each canceled order in the subset of canceled orders as "inventory added"; In response to receiving an instruction for batch processing of order inventory, extracting a subset of the canceled orders marked as "inventory to be added" includes: Extracting a time queue of the remaining inventory of the inventory system; Extracting a time queue of the total order volume of the subset of canceled orders; Sequentially encode the time queue of the remaining inventory and the time queue of the total order volume to obtain the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the cancelled order volume; Performing interactive compensation of the time series common characteristics on the remaining inventory time series fluctuation characteristics and the canceled order time series fluctuation characteristics to obtain a remaining inventory-cancelled order time series collaborative interactive compensation representation; Based on the remaining inventory quantity-cancelled order quantity time series collaborative interactive compensation representation, a monitoring result is obtained, and the monitoring result is used to indicate whether to generate a prompt for recommending "batch processing order inventory"; The interactive compensation of the time series common characteristics of the remaining inventory quantity time series fluctuation characteristics and the time series fluctuation characteristics of the canceled order quantity is performed to obtain the remaining inventory quantity-cancelled order quantity time series collaborative interactive compensation representation, including: Inputting the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector into a common feature extraction network to obtain a common feature representation vector between the remaining inventory-cancelled order quantity time series features; Based on the common feature representation vector between the remaining inventory-cancelled order quantity time series features, the remaining inventory quantity time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector are subjected to feature fine-grained compensation to obtain the remaining inventory quantity-cancelled order quantity time series collaborative interactive compensation representation.

2. The hot spot inventory return processing method according to claim 1 is characterized in that: The time queue of the remaining inventory and the time queue of the total order volume are respectively sequence-encoded to obtain the time series fluctuation characteristics of the remaining inventory and the time series fluctuation characteristics of the canceled order volume, including: inputting the time queue of the remaining inventory and the time queue of the total order volume into a sequence encoder based on a forward LSTM model to obtain the time series fluctuation feature vector of the remaining inventory as the time series fluctuation feature of the remaining inventory and the time series fluctuation feature vector of the canceled order volume as the time series fluctuation feature of the canceled order volume.

3. The hot spot inventory return processing method according to claim 2 is characterized in that: Inputting the remaining inventory time series fluctuation feature vector and the canceled order quantity time series fluctuation feature vector into a common feature extraction network to obtain a common feature representation vector between the remaining inventory-cancelled order quantity time series features, including: The remaining inventory time series fluctuation characteristic vector and the canceled order time series fluctuation characteristic vector are added according to the position points to obtain the remaining inventory-cancelled order time series point addition vector; The remaining inventory-cancelled order quantity time series point addition vector is multiplied by the weight matrix, and the obtained remaining inventory-cancelled order quantity time series weight vector and the bias vector are added by position to obtain the remaining inventory-cancelled order quantity time series commonality vector; The remaining inventory-cancelled order quantity time series commonality vector is input into the tanh function for processing to obtain the commonality feature representation vector between the remaining inventory-cancelled order quantity time series features.

4. The hot spot inventory return processing method according to claim 3 is characterized in that: Based on the common feature representation vector between the remaining inventory and the cancelled order quantity time series features, the remaining inventory time series fluctuation feature vector and the cancelled order quantity time series fluctuation feature vector are subjected to feature fine-grained compensation to obtain the remaining inventory quantity-cancelled order quantity time series collaborative interactive compensation representation, including: Input the remaining inventory time series fluctuation feature vector, the canceled order time series fluctuation feature vector and the common feature representation vector between the remaining inventory and the canceled order time series features into a vector probabilization unit based on the Sigmoid function to obtain a probabilistic remaining inventory time series fluctuation feature vector, a probabilistic canceled order time series fluctuation feature vector and a probabilistic remaining inventory - common feature representation vector between the canceled order time series features; Calculate a remaining inventory time series fine-grained compensation vector of the probabilistic remaining inventory time series fluctuation feature vector relative to the common feature representation vector between the probabilistic remaining inventory-cancelled order time series features; Calculate a fine-grained compensation vector for the time series of canceled order quantity relative to the common feature representation vector between the probabilistic remaining inventory and the time series features of canceled order quantity; Inputting the remaining inventory time series fine-grained compensation vector and the remaining inventory time series fluctuation characteristic vector into a fine-grained compensation module to obtain a remaining inventory time series fine-grained characteristic compensation vector; Inputting the cancellation order quantity time series fine-grained compensation vector and the cancellation order quantity time series fluctuation characteristic vector into the fine-grained compensation module to obtain the cancellation order quantity time series fine-grained characteristic compensation vector; The remaining inventory quantity time series fine-grained feature compensation vector, the canceled order quantity time series fine-grained feature compensation vector and the remaining inventory-cancelled order quantity time series feature common feature representation vector are cascaded to obtain the remaining inventory quantity-cancelled order quantity time series collaborative interaction compensation representation vector as the remaining inventory quantity-cancelled order quantity time series collaborative interaction compensation representation.

5. The hot spot inventory return processing method according to claim 4 is characterized in that: Calculating a remaining inventory time series fine-grained compensation vector of the probabilistic remaining inventory time series fluctuation feature vector relative to the common feature representation vector between the probabilistic remaining inventory and the time series features of the canceled order quantity, including: Calculate the common feature representation vector between the probabilistic remaining inventory time series fluctuation feature vector and the probabilistic remaining inventory-cancelled order time series feature and divide it by the position point to obtain the remaining inventory time series contribution vector; Calculating a logarithmic function value with base 2 of the absolute value of each eigenvalue of the remaining inventory time series contribution vector to obtain a remaining inventory time series contribution logarithmic vector; Calculate the positional point multiplication of the probabilistic remaining inventory time series fluctuation characteristic vector and the remaining inventory time series contribution logarithm vector to obtain a point product vector, and calculate the exponential function value of each point product value in the point product vector with the natural constant e as the base to obtain the remaining inventory time series compensation vector; The remaining inventory time series compensation vector is input into a softmax function to obtain the remaining inventory time series fine-grained compensation vector.

6. The hot spot inventory return processing method according to claim 5 is characterized in that: The remaining inventory time series fine-grained compensation vector and the remaining inventory time series fluctuation characteristic vector are input into a fine-grained compensation module to obtain the remaining inventory time series fine-grained characteristic compensation vector, including: calculating the remaining inventory time series fluctuation characteristic vector and the remaining inventory time series fluctuation characteristic vector by position point multiplication to obtain the remaining inventory time series fine-grained characteristic compensation vector.

7. The hot spot inventory return processing method according to claim 6 is characterized in that: Based on the remaining inventory-cancelled order timing series collaborative interaction compensation representation, a monitoring result is obtained, including: inputting the remaining inventory-cancelled order timing series collaborative interaction compensation representation vector into a classifier-based intelligent monitor to obtain a monitoring result.

Images