Transport capacity demand regulation and control reference data obtaining method, electronic equipment and storage equipment

Abstract

The invention discloses a transport capacity demand regulation and control reference data obtaining method, electronic equipment and storage equipment. The method comprises the following steps: obtaining production order prediction data and regulation and control benefit prediction data of a first distribution starting point and a first distribution ending point in a transport capacity demand regulation and control area under different transport capacity demand regulation and control feature data; the waybill bearing capacity prediction data, meeting the waybill performance requirement, of a second delivery starting and ending point in the transport capacity requirement regulation and control area are obtained, the starting point position of the second delivery starting and ending point comprises the starting point position of the first delivery starting and ending point, and the ending point position of the second delivery starting and ending point comprises the ending point position of the first delivery starting and ending point; and according to the waybill bearing capacity prediction data, the production bill prediction data and the regulation and control benefit prediction data, obtaining the transport capacity demand regulation and control reference data of the transport capacity demand regulation and control region. In this way, the transport capacity demand and the transport capacity are balanced, and the benefits obtained after regulation and control do not influence the waybill distribution ecological environment of the transport capacity demand regulation and control area as much as possible or slightly influence the waybill distribution ecological environment of the transport capacity demand regulation and control area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computers, in particular to a method for obtaining reference data for regulating transport demand, an electronic device and a storage device. BACKGROUND

[0002] With the acceleration of people's life pace, more and more people choose delivery to meet the needs of life without going out, which means a substantial increase in transport demand. If the transport demand is significantly higher than the transport capacity, the transport capacity cannot meet the transport demand, and a series of problems such as transport delay and transport error will occur. If the transport demand and the transport capacity can be balanced, or in other words, if the transport demand and the transport capacity can be matched as much as possible, then these problems can be avoided as much as possible. Therefore, how to balance the transport demand and the transport capacity is a problem to be solved. SUMMARY

[0003] The embodiment of the present application provides a method for obtaining reference data of demand regulation of transport capacity, so as to balance the demand of transport capacity and the transport capacity as much as possible. The method comprises the following steps: obtaining order production prediction data and regulation benefit prediction data of a first distribution starting and ending point in a demand regulation area of transport capacity under different demand regulation characteristics of transport capacity; obtaining long-period characteristic data of the first distribution starting and ending point, wherein the long-period characteristic data of the first distribution starting and ending point comprises long-period supply data of a starting point position of the first distribution starting and ending point, long-period supply data sequence of order entities around the starting point position of the first distribution starting and ending point, and long-period supply data sequence of order entities around a terminal point position of the first distribution starting and ending point; obtaining short-period order operation prediction data of the first distribution starting and ending point according to the long-period supply data of the starting point position of the first distribution starting and ending point, the long-period supply data sequence of order entities around the starting point position of the first distribution starting and ending point, and the long-period supply data sequence of order entities around the terminal point position of the first distribution starting and ending point, wherein the time length of the short period in the short-period order operation prediction data is shorter than the time length of the long period in the long-period characteristic data; obtaining real-time characteristic data of the first distribution starting and ending point, and obtaining a plurality of demand regulation characteristics of transport capacity for the first distribution starting and ending point; obtaining order production prediction data and regulation benefit prediction data of the first distribution starting and ending point according to the short-period order operation prediction data of the first distribution starting and ending point, the real-time characteristic data of the first distribution starting and ending point, and the plurality of demand regulation characteristics of transport capacity for the first distribution starting and ending point; obtaining order acceptance capacity prediction data of a second distribution starting and ending point in the demand regulation area of transport capacity, which meets the order fulfillment requirements, wherein the starting point position of the second distribution starting and ending point comprises the starting point position of the first distribution starting and ending point, and the terminal point position of the second distribution starting and ending point comprises the terminal point position of the first distribution starting and ending point; and obtaining reference data of demand regulation of transport capacity of the demand regulation area of transport capacity according to the order acceptance capacity prediction data, the order production prediction data and the regulation benefit prediction data, wherein the reference data of demand regulation of transport capacity comprises reference data of demand regulation of transport capacity determined for at least one first distribution starting and ending point.

[0004] In the embodiment of the present application, both order production data, i.e., order production prediction data of a first distribution starting and ending point in a demand regulation area of transport capacity under different demand regulation characteristics of transport capacity, and transport capacity data, i.e., order acceptance capacity prediction data of a second distribution starting and ending point in the demand regulation area of transport capacity, which meets the order fulfillment requirements, are considered. According to the order production data (i.e., order production prediction data) and the transport capacity data (i.e., order acceptance capacity prediction data), reference data of demand regulation of transport capacity required to be executed for regulating the demand of transport capacity is determined, so that the determined reference data of demand regulation of transport capacity can be used to regulate the demand of transport capacity. Since the order production data, which represents the demand of transport capacity to a certain extent, and the transport capacity data are considered when the reference data of demand regulation of transport capacity is determined, the demand of transport capacity and the transport capacity can be matched as much as possible after the demand of transport capacity is regulated by using the determined reference data of demand regulation of transport capacity, so that the demand of transport capacity and the transport capacity can be balanced as much as possible.

[0005] In addition, in the embodiment of the present application, the transport demand regulation reference data includes transport demand regulation reference data determined for at least one first distribution origin-destination point. In other words, the embodiment of the present application proposes to determine transport demand regulation reference data for distribution origin-destination points (OD, Origin Destination) in the transport demand regulation region. Since transport demand regulation at the granularity of distribution origin-destination points is more accurate and precise than unified transport demand regulation for the entire transport demand regulation region, the regulation result obtained by transport demand regulation at the granularity of distribution origin-destination points is also more ideal. Therefore, the embodiment of the present application can determine transport demand regulation reference data for a more fine-grained distribution range, and regulate transport demand for a more fine-grained distribution range, which can not only balance the transport demand and transport capacity of the more fine-grained distribution range, but also further balance the transport demand and transport capacity of the entire transport demand regulation region.

[0006] In addition, if only the balance between transport demand and transport capacity is considered, there can be multiple regulation reference data to match transport demand and transport capacity. If the regulation reference data selected from the multiple regulation reference data severely suppresses transport demand, the order volume will be significantly reduced, which may affect the transport order distribution ecological environment of the transport demand regulation region, for example, severely affecting the user's delivery to home service experience, causing serious loss of order volume and transport order volume of merchants, delivery personnel, online to offline (O2O, Online to Offline) service system, and affecting the income of merchants, delivery personnel, and online to offline service platform. In view of this, in the embodiment of the present application, in addition to obtaining the order generation prediction data of the first distribution origin-destination point in the transport demand regulation region under different transport demand regulation characteristic data, the regulation benefit prediction data of the first distribution origin-destination point in the transport demand regulation region under different transport demand regulation characteristic data is also obtained, and when obtaining the transport demand regulation reference data, not only the order acceptance prediction data and the order generation prediction data are considered, but also the regulation benefit prediction data representing the regulation benefit is considered, so that the transport demand regulation reference data obtained not only balances the transport demand and the transport capacity, but also makes the benefit obtained after regulation as little as possible or does not affect the transport order distribution ecological environment of the transport demand regulation region.

[0007] It's also worth noting that in practical applications, if the order volume in a certain region is large, causing the number of waybills in that region to significantly exceed its transportation capacity, then delivery resources will be overloaded with many waybills simultaneously. This means delivery resources will have a heavy workload, resulting in longer average delivery times per order. In this situation, it's inevitable that users will request order reminders. User reminders generate information, which inevitably creates signaling overhead in the information transmission system, consuming transmission resources. User reminders also require delivery resources or online-to-offline service systems to respond, which in turn increases the workload of these systems, consuming computing and storage resources. Therefore, when transportation capacity and demand are mismatched, it can lead to increased signaling overhead, excessive consumption of transmission, computing, and storage resources, and increased workload for delivery resources or online-to-offline service systems. Conversely, if appropriate reference data for transportation capacity demand can be determined and used to adjust transportation demand, thereby achieving a balance between transportation capacity and demand, then the above problems can be avoided. Therefore, the embodiments of this application can avoid the above-mentioned problems caused by the imbalance between transportation capacity demand and transportation capacity as much as possible. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the logical structure of the capacity demand control system provided in the embodiments of this application.

[0009] Figure 2 This is a schematic diagram of the logistics capacity model provided in the embodiments of this application.

[0010] Figure 3 This is a schematic diagram of the structure of the commercial order model provided in the embodiments of this application.

[0011] Figure 4 Another logical structure diagram of the capacity demand control system provided in the embodiments of this application.

[0012] Figure 5 A flowchart illustrating the method for obtaining reference data for capacity demand regulation provided in this application embodiment.

[0013] Figure 6 A schematic diagram of the device for obtaining reference data for capacity demand regulation provided in the embodiments of this application.

[0014] Figure 7 A schematic diagram of an electronic device provided in the first embodiment of this application. Detailed Implementation

[0015] This application's embodiments can be applied to delivery service scenarios. In the delivery service scenario applied in this application embodiment, when a capacity-demanding party needs delivery services, they can send a delivery service request to a capacity provider, which will then provide the delivery service. A capacity-demanding party can refer to an entity with capacity needs, such as a user, merchant, or an order service system that provides order services to users and merchants. For a user, when a user needs to deliver goods to other entities (such as other users, units, etc.) through the delivery service provided by a capacity provider, the user is a capacity-demanding party. For a merchant, when a merchant needs to deliver goods to a user through the delivery service provided by a capacity provider, the merchant is a capacity-demanding party. For an order service system, when an order service system needs to deliver goods from a merchant to a user through the delivery service provided by a capacity provider, the order service system is a capacity-demanding party. A capacity provider can refer to an entity providing delivery services, such as a delivery service unit or delivery person. Delivery people may also be referred to as riders, food delivery workers, couriers, etc. in some cases. The delivery service provider can refer to the company where the delivery person works or the delivery station to which the delivery person belongs.

[0016] In practical applications, transportation capacity and demand are not always matched. This can lead to a series of problems such as delayed and incorrect delivery orders. Taking the online-to-offline service system mentioned above as an example, if a region experiences a large volume of orders during a certain period, but the number of orders that delivery personnel in that region can handle during that period is significantly less than that volume, then the transportation capacity will be insufficient to meet the demand, resulting in delivery delays and errors. If transportation capacity and demand can be balanced as much as possible, these problems can be avoided as much as possible.

[0017] First Embodiment Please refer to Figure 1This application proposes a capacity demand control system, which includes a commercial order distribution service subsystem 101 (hereinafter referred to as "System 101"), a logistics capacity distribution service subsystem 102 (hereinafter referred to as "System 102"), a real-time handshake service subsystem 103 (hereinafter referred to as "System 103"), a control center 104, a logistics capacity distribution auxiliary service subsystem 105 (hereinafter referred to as "System 105"), and a logistics delivery time assessment service subsystem 106 (hereinafter referred to as "System 106"). The control center 104, System 105, and System 106 are optional. System 101 is used to obtain the order production forecast data and regulation benefit forecast data of the first delivery origin and destination in the capacity demand regulation area under different capacity demand regulation characteristic data; System 102 is used to obtain the order acceptance capacity forecast data of the second delivery origin and destination in the capacity demand regulation area that meets the order fulfillment requirements; System 103 is used to obtain the capacity demand regulation reference data of the capacity demand regulation area based on the order acceptance capacity forecast data, order production forecast data and regulation benefit forecast data; Regulation center 104 is used to execute the capacity demand regulation reference data; System 105 is used to determine whether the order production forecast data of the second delivery origin and destination corresponding to the capacity demand regulation reference data meets the order fulfillment requirements, and feeds back the determination result to System 103; System 106 is used to determine whether the order production forecast data of the second delivery origin and destination meets the order fulfillment requirements in a specified time period. If it is determined that the order production forecast data of the second delivery origin and destination does not meet the order fulfillment requirements in a specified time period, then System 103 is triggered to obtain the capacity demand regulation reference data of the capacity demand regulation area based on the order acceptance capacity forecast data, order production forecast data and regulation benefit forecast data.

[0018] The capacity demand control area can be an area where capacity demand needs to be controlled, such as a business district. The first delivery origin and destination can include the start and end points of the delivery task. In the instant delivery field, the start point of the delivery task can be the merchant's location, and the end point can be the user's location. Since order prediction data is predictive data, not actual order data, the actual delivery address cannot be obtained at this time; that is, the user's actual delivery address cannot be determined. Therefore, the end point in the first delivery origin and destination can be understood as the user's location within the merchant's delivery range, such as the user's corresponding Area of ​​Interest (AOI). The start point in the first delivery origin and destination can be understood as the merchant (SHOP). Although the order prediction data is not actual order data, the merchant can refer to an actual merchant. The start point in the first delivery origin and destination can also be the merchant's corresponding Point of Interest (POI). If the start point of the first delivery origin and destination is a SHOP, and the end point is an AOI, then the first delivery origin and destination can be understood as SHOP-AOI. The order forecast data and regulation benefit forecast data obtained by System 101 for the first delivery origin and destination can refer to order forecast data and regulation benefit forecast data for one, multiple, or all delivery origins and destinations within the capacity demand regulation area. Order forecast data can be forecast data for output orders, such as the expected order volume. Regulation benefit forecast data can represent the expected benefits obtained after regulating capacity demand, such as predicted revenue. Regulation benefit forecast data can also be the predicted order volume; in this case, the order forecast data and regulation benefit forecast data are the same data. Regulation benefit forecast data can also be other data used to represent order benefits, such as the predicted average delivery efficiency per order.

[0019] Capacity demand regulation feature data can be characteristic data used to regulate capacity demand. In the on-demand delivery sector, capacity demand regulation features may include at least one regulation feature used to regulate capacity demand, such as increasing order delivery costs or reducing the merchant's delivery range. These regulation features can also be called regulation measures. Increasing order delivery costs can include regulation measures such as increasing order delivery fees or increasing order delivery time. Capacity demand regulation feature data may include at least one regulation feature data, such as the increase in order delivery costs or the reduction in the merchant's delivery range. Capacity demand regulation feature data can include regulation factors and regulation factor values. Regulation factors can also be called regulation parameters, and regulation factor values ​​can also be called regulation parameter values. For example, if the regulation factor is to increase order delivery time and the regulation factor value is 15 minutes, the capacity demand regulation feature data could be used to increase the order delivery time by 15 minutes. Different capacity demand control feature data can refer to multiple capacity demand control feature data set for a single first delivery origin / destination, multiple first delivery origin / destination, or all first delivery origin / destination. In this case, the order prediction data and control benefit prediction data for a single first delivery origin / destination, multiple first delivery origin / destination, or all first delivery origin / destination under different capacity demand control feature data can be determined. For the case of multiple first delivery origin / destination or all first delivery origin / destination, multiple capacity demand control feature data can be set for any one of the first delivery origin / destination. Multiple capacity demand control feature data can also refer to different feature data corresponding to the same type of capacity demand control feature. For example, capacity demand control feature data such as increasing order delivery time by 5 minutes or increasing order delivery time by 10 minutes can be regarded as multiple capacity demand control feature data. These capacity demand control feature data belong to the same type of capacity demand control feature, only the control factor value or control parameter value is different. Multiple capacity demand regulation feature data can also refer to sets of different feature data corresponding to certain combinations of different types of capacity demand regulation features. For example, sets such as {order delivery time increases by 5 minutes, merchant's delivery range decreases by 1 kilometer}, {order delivery time increases by 10 minutes, merchant's delivery range decreases by 2 kilometers}, etc., can be considered as multiple capacity demand regulation feature data. These sets of capacity demand regulation feature data all belong to specific combinations of different types of capacity demand regulation features, but the data between sets are not entirely the same. Multiple capacity demand regulation feature data can also refer to sets of feature data corresponding to different combinations of different types of capacity demand regulation features. For example, sets such as {order delivery time increases by 5 minutes, merchant's delivery range decreases by 1 kilometer}, {order delivery time increases by 10 minutes}, {order delivery fee increases by 2 yuan, merchant's delivery range decreases by 3 kilometers}, etc., can be considered as multiple capacity demand regulation feature data. These sets of capacity demand regulation feature data belong to different combinations of different types of capacity demand regulation features, and the data between sets are naturally different.The multiple capacity demand control feature data set for different first delivery origin and destination can be the same or different.

[0020] Obtaining the order volume forecast data and regulation benefit forecast data for the first delivery origin and destination within a capacity demand regulation area under different capacity demand regulation feature data can be understood as obtaining the order volume and revenue data for the first delivery origin and destination granularity under different capacity demand regulation feature data. Since both order volume forecast data and regulation benefit forecast data are forecast data, obtaining the order volume and revenue data for the first delivery origin and destination granularity under different capacity demand regulation feature data can be specifically understood as obtaining the order volume and revenue data for the first delivery origin and destination granularity under different capacity demand regulation feature data within a future period (e.g., 30 minutes). Specifically, it can refer to obtaining the order volume and revenue data for the first delivery origin and destination granularity under each of multiple capacity demand regulation feature data within a future period. In other words, predicted order volume and revenue data can be obtained for each capacity demand regulation feature data. Therefore, the order volume under different capacity demand regulation feature data can refer to the order volume under each capacity demand regulation feature data, and the revenue data under different capacity demand regulation feature data can refer to the revenue data under each capacity demand regulation feature data.

[0021] For System 102, the starting point in the second delivery origin-destination can be the AOI or POI corresponding to the merchant, and the ending point can be the location of the user within the merchant's delivery range, such as the AOI or POI corresponding to the user. If the starting point in the second delivery origin-destination is the AOI corresponding to the merchant, and the ending point is the AOI corresponding to the user, then the second delivery origin-destination can be understood as AOI-AOI. The order fulfillment capacity forecast data for the second delivery origin-destination in the capacity demand control area can refer to the order fulfillment capacity forecast data for a certain second delivery origin-destination, multiple second delivery origin-destinations, or all second delivery origin-destinations in the capacity demand control area. In other words, the order fulfillment capacity forecast data can be the order fulfillment capacity forecast data at the granularity of the delivery origin-destination. If the first delivery origin-destination is SHOP-AOI, and the second delivery origin-destination is AOI-AOI, then the SHOP in the first delivery origin-destination can be the SHOP in the AOI corresponding to the merchant in the second delivery origin-destination, or in other words, the starting point location in the second delivery origin-destination includes the starting point location in the first delivery origin-destination. Waybill fulfillment requirements can be specific to the fulfillment of waybills. For example, they could require that the average delivery time of all waybills within a future period (e.g., 30 minutes) or the average delivery time of a specified number of waybills does not exceed a threshold for average delivery time per waybill (e.g., 38 minutes). This threshold can be represented as the "T delivery time standard." Another example is that waybill fulfillment requirements could also refer to a requirement that the average delivery cost of all waybills within a future period or the average delivery cost of a specified number of waybills does not exceed the average delivery cost per waybill. Waybill capacity forecasting data represents the predicted waybill handling capacity, such as the predicted number of waybills that can be handled, the predicted waybill pressure data that can be absorbed, or the predicted number of waybills with a specified delivery difficulty data that can be handled. Since waybill capacity forecasting data is forecast data, the obtained waybill capacity forecasting data can be waybill capacity forecasting data for a future period. Obtaining forecast data on the capacity to handle waybills at the second delivery origin and destination within the capacity demand control area, which meets the waybill fulfillment requirements, can refer to obtaining the predicted number of waybills that can be handled at the second delivery origin and destination granularity under the T delivery time standard within a future period.

[0022] System 101 can send production order forecast data and regulation benefit forecast data to system 103. System 102 can send waybill capacity forecast data to system 103. After obtaining the production order forecast data, regulation benefit forecast data, and waybill capacity forecast data, system 103 obtains reference data for capacity demand regulation in the capacity demand regulation area. Specifically, the waybill capacity forecast data can be used as a constraint condition for the production order forecast data. From the multiple obtained regulation benefit forecast data, the regulation benefit forecast data that meets the regulation benefit conditions is obtained as the determined regulation benefit forecast data. The capacity demand regulation characteristic data corresponding to the determined regulation benefit forecast data is determined as the capacity demand regulation reference data, which can also be simply referred to as regulation reference data. For example, if the waybill capacity forecast data is the predicted number of waybills that can be accepted, the production order forecast data is the expected order volume, and the regulation benefit forecast data is the predicted revenue, then the total revenue data under each capacity demand regulation characteristic data can be obtained based on the expected order volume and the predicted revenue. For example, the product of the projected order volume and the projected revenue under the same capacity demand control feature data can be obtained as the total revenue data under the same capacity demand control feature data. From the order volumes of multiple projected outputs, the order volume that does not exceed the logistics measurement's projected order volume under the same capacity demand control feature data is determined as the candidate projected output order volume. From the control benefit prediction data corresponding to the candidate projected output order volumes, the maximum control benefit prediction data is determined. The capacity demand control feature data corresponding to the maximum control benefit prediction data is determined as the capacity demand control reference data.

[0023] If the different capacity demand control feature data refers to multiple capacity demand control feature data set for a certain first delivery origin and destination within a capacity demand control area, then the capacity demand control reference data can refer to one capacity demand control feature data determined from the multiple capacity demand control feature data set for a certain first delivery origin and destination. In this case, the capacity demand control feature data determined from the multiple capacity demand control feature data can specifically refer to the capacity demand control feature data to be executed for a certain first delivery origin and destination. If the aforementioned different capacity demand control feature data refers to multiple capacity demand control feature data determined for multiple or all first delivery origins and destinations within a capacity demand control area, then the capacity demand control reference data here can include one capacity demand control feature data determined for each first delivery origin and destination from the multiple capacity demand control feature data set for each first delivery origin and destination. That is, the capacity demand control reference data in this case includes the capacity demand control reference data determined for each first delivery origin and destination.

[0024] Since system 103 obtains the capacity demand control reference data for the capacity demand control area after obtaining the production order prediction data, and the capacity demand control reference data can be determined from different capacity demand control characteristic data, system 103 can determine the production order prediction data corresponding to the capacity demand control reference data after obtaining it. System 103 can obtain the waybill prediction data corresponding to the capacity demand control reference data based on the production order prediction data corresponding to the capacity demand control reference data. For example, if the production order prediction data corresponding to the capacity demand control reference data refers to the order volume distribution data at the first delivery origin-end point granularity of the capacity demand control area after control according to the capacity demand control reference data, then the production order prediction data corresponding to the capacity demand control reference data can refer to the waybill volume distribution data at the second delivery origin-end point granularity of the capacity demand control area after control according to the capacity demand control reference data. The first delivery origin-end point granularity and the second delivery origin-end point granularity can be different. For example, the starting position of any first delivery origin / destination can fall within the range of the starting position of any second delivery origin / destination. The starting positions of any two first delivery origin / destinations may both fall within the same range of the starting position of the second delivery origin / destination, or they may fall within the range of the starting positions of different second delivery origin / destinations. For any second delivery origin / destination, the estimated order volume for that second delivery origin / destination can be determined by summing the estimated order volumes of all first delivery origin / destinations whose endpoints are the same as the endpoint of that second delivery origin / destination and whose starting positions fall within the range of the endpoint of that second delivery origin / destination. After obtaining the estimated order volumes of multiple or all second delivery origin / destinations within the capacity demand control area, the order volume distribution data at the second delivery origin / destination granularity within the capacity demand control area is obtained.

[0025] System 103 can send waybill forecast data to system 105. System 105 can determine whether the expected fulfillment status corresponding to the waybill forecast data meets the waybill fulfillment requirements. That is, it can determine whether, under the existing configuration of delivery resources in the area targeted by capacity demand regulation, the expected fulfillment status can meet the waybill fulfillment requirements if the waybill corresponding to the waybill forecast data is delivered, and feed back the determination result to system 103. If system 105 determines that the expected fulfillment status corresponding to the waybill forecast data can meet the waybill fulfillment requirements, it can send a confirmation result to system 103 indicating that the expected fulfillment status corresponding to the waybill forecast data meets the waybill fulfillment requirements; otherwise, it can feed back a feedback result indicating that the expected fulfillment status corresponding to the waybill forecast data cannot meet the waybill fulfillment requirements. Based on the above feedback results, it can also feed back information indicating the gap between the expected fulfillment status corresponding to the waybill forecast data and the waybill fulfillment requirements. If the expected fulfillment status corresponding to the waybill forecast data includes the average expected delivery time per waybill, and the waybill fulfillment requirements include an average delivery time threshold, and the average expected delivery time per waybill is greater than the average delivery time threshold, then the difference between the average expected delivery time per waybill and the average delivery time threshold can be fed back. Upon receiving the above confirmation result, system 103 can confirm adjusting capacity demand according to the capacity demand adjustment reference data. If the above feedback result is received, or if the above feedback result and the above difference information are received, the capacity demand adjustment reference data is re-obtained until the expected fulfillment status corresponding to the waybill forecast data meets the waybill fulfillment requirements. Because waybills and capacity in on-demand delivery scenarios change in real time, determining the reference data for capacity demand control and then assuming that capacity demand control is implemented according to this data verifies the effectiveness of the reference data by checking whether the expected fulfillment status of the anticipated waybills meets the fulfillment requirements. This ensures that the reference data balances capacity demand and available capacity while meeting actual delivery needs. Furthermore, implementing the reference data requires significant computing and storage resources. Executing effective reference data ensures efficient use of these resources, while ineffective data wastes them. Therefore, verifying the effectiveness of the reference data improves the utilization of computing and storage resources and prevents waste.

[0026] System 103 can proactively obtain waybill forecast data corresponding to the capacity demand control reference data. For example, System 103 can periodically obtain waybill forecast data corresponding to the capacity demand control reference data. When the cycle for obtaining waybill forecast data arrives, System 103 can proactively send a commercial flow order distribution data request message to System 101 to request the production order forecast data and control benefit forecast data for the first delivery origin and destination in the capacity demand control area under different capacity demand control characteristic data, and proactively send a logistics capacity distribution data request message to System 102 to request the logistics capacity distribution data of the waybill acceptance capacity forecast data that meets the waybill fulfillment requirements for the second delivery origin and destination in the capacity demand control area. System 101 can return the production order forecast data and control benefit forecast data for the first delivery origin and destination in the capacity demand control area under different capacity demand control characteristic data to System 103 based on the commercial flow order distribution data request message, and System 102 can return the waybill acceptance capacity forecast data that meets the waybill fulfillment requirements for the second delivery origin and destination in the capacity demand control area to System 103 based on the logistics capacity distribution data request message. After receiving the production order forecast data, regulation benefit forecast data, and waybill capacity forecast data, system 103 can obtain the capacity demand regulation reference data for the current cycle in the capacity demand regulation area. System 103 can also obtain waybill forecast data corresponding to the capacity demand regulation reference data after being triggered by other subsystems. For example, system 101 can proactively provide system 103 with production order forecast data and regulation benefit forecast data for the first delivery origin and destination in the capacity demand regulation area under different capacity demand regulation characteristic data; system 102 can proactively provide system 103 with waybill capacity forecast data for the second delivery origin and destination in the capacity demand regulation area that meets waybill fulfillment requirements. After obtaining this data, system 103 then obtains the waybill forecast data corresponding to the capacity demand regulation reference data.

[0027] Upon being triggered by System 106, System 103 can obtain waybill forecast data corresponding to the capacity demand control reference data. System 106 can trigger System 103 to obtain capacity demand control reference data for the capacity demand control area when it determines that the waybill forecast data for the second delivery origin-end point does not meet the waybill fulfillment requirements within a specified time period. System 106 can assess whether the waybill forecast data at the second delivery origin-end point granularity meets the waybill fulfillment requirements for a future time period, or it can assess this at the second delivery origin-end point granularity every cycle, for example, assessing every minute whether the waybill forecast data at the second delivery origin-end point granularity meets the waybill fulfillment requirements for the following 30 minutes. Assuming that system 106 assesses whether the waybill forecast data at the second delivery origin-destination level within the next 30 minutes meets the waybill fulfillment requirements in the current cycle, system 106 can obtain the currently accepted but not yet completed waybill data for at least one second delivery origin-destination in the capacity demand control area. This currently accepted but not yet completed waybill data is referred to as the currently incomplete waybill data. System 106 can also obtain the estimated waybill data to be accepted for at least one second delivery origin-destination within the next 30 minutes. The estimated waybill data refers to the data of waybills that are not currently generated but are estimated to still need to be accepted. System 106 can predict this estimated waybill data based on historical waybill data for the same period. After obtaining the currently incomplete waybill data and the estimated waybill data to be accepted, system 106 can obtain the total estimated data of incomplete waybills for at least one second delivery origin-destination within the next 30 minutes. This total estimated data of incomplete waybills can refer to the waybill forecast data for the second delivery origin-destination within a specified time period. Then, it can be determined whether the expected performance status of the waybill corresponding to the total expected data of the incomplete waybill meets the waybill performance requirements. If the requirements are met, it is not necessary to trigger system 103 to obtain the capacity demand adjustment reference data. Otherwise, system 103 can be triggered to obtain the capacity demand adjustment reference data.

[0028] Whether system 103 obtains reference data for capacity demand regulation in the region, system 101 obtains order forecast data and regulation benefit forecast data, or system 102 obtains order capacity forecast data, all require the support of computing and storage resources from the computer system. If system 106 determines that the expected fulfillment status of the orders corresponding to the total expected data of unfulfilled orders meets the order fulfillment requirements within a future time period, then it is not necessary to trigger system 103 to obtain the reference data for capacity demand regulation. If system 103 does not need to obtain the reference data for capacity demand regulation, then system 101 does not need to obtain order forecast data and regulation benefit forecast data, and system 102 does not need to obtain order capacity forecast data, thereby saving the computing and storage resources required to obtain the reference data for capacity demand regulation, order forecast data, regulation benefit forecast data, and order capacity forecast data. Therefore, the system 106 determines whether to trigger the system 103 to obtain reference data for capacity demand adjustment based on whether the expected performance status meets the waybill performance requirements. This can relatively save the use of computer system computing and storage resources and avoid wasting computer system computing and storage resources.

[0029] After obtaining the capacity demand control reference data, system 103 can send the reference data to control center 104, which will then execute the control data. If the capacity demand control reference data includes at least one adjustment value among the order delivery time adjustment value and the merchant's deliverable range adjustment value for each first delivery origin and destination in the capacity demand control area, control center 104 can adjust at least one delivery parameter among the order delivery time and the merchant's deliverable range according to the at least one adjustment value for each first delivery origin and destination. If control center 104 does not have the ability or authority to adjust the value of a certain parameter, it can send the control feature data corresponding to that parameter to a device or system with such ability or authority, which will then adjust the parameter value according to the control feature data. The control feature data corresponding to that parameter may include, for example, the adjustment value for that parameter.

[0030] To ensure the effectiveness of the capacity demand regulation reference data, the execution process and results can be monitored. Therefore, the capacity demand regulation system can also include an execution monitoring subsystem. Figure 1(Not shown in the diagram) This is used to detect whether the capacity demand regulation reference data has been executed and the results generated after execution. For example, the execution detection subsystem can obtain the capacity demand regulation reference data provided by system 103 or control center 104. If the capacity demand regulation reference data includes capacity demand regulation reference data for each first delivery origin and destination in the capacity demand regulation area, the execution detection subsystem can detect whether the capacity demand regulation reference data for each first delivery origin and destination has been executed. After detecting that the capacity demand regulation reference data has been executed, the execution detection subsystem can also detect the execution effect of the capacity demand regulation reference data. For example, it can detect whether the actual production order data matches the production order forecast data, whether the actual regulation benefit data matches the regulation benefit forecast data, and whether the actual waybill fulfillment data meets the waybill fulfillment requirements. If the regulation benefit forecast data is the predicted revenue, then the actual regulation benefit data can be the actual revenue. The production order forecast data and regulation benefit forecast data can be provided to the execution detection subsystem by system 101, system 103, or control center 104. Actual waybill fulfillment data may include average delivery time per waybill for each actual waybill received. If the execution detection subsystem determines that the execution effect of the capacity demand control reference data does not meet the preset expected execution effect, the execution detection subsystem can feed back the detection result indicating that the execution effect of the capacity demand control reference data does not meet the preset expected execution effect to the control center 104 or system 103. It can also feed back specific actual execution effect data such as actual order production data, actual control benefit data, or actual waybill fulfillment data, or specific execution effect difference data such as the difference between actual order production data and predicted order production data, the difference between actual control benefit data and predicted control benefit data, or the difference between actual waybill fulfillment data and waybill fulfillment requirements. After receiving the above feedback results, system 103 can determine whether it is necessary to re-obtain the capacity demand control reference data for the capacity demand control area.

[0031] Second Embodiment The logistics capacity distribution service subsystem may include, for example: Figure 2The logistics capacity model shown can also be called a waybill fulfillment data prediction model. The input data for the logistics capacity model can include at least one of the following: supply characteristic data, spatiotemporal environmental characteristic data, and AOI-AOI waybill data. Supply characteristic data can refer to the supply characteristic data of each shop in the starting point of the second delivery origin-destination region. Specifically, it can refer to the sum or average of the supply characteristic data of all shops, such as the sum or average of the order volume of all shops within a historical time period, or the sum or average of the user payment amount data received by the online-to-offline service system for all shops within a historical time period. The sum or average of the order volume within a historical time period, and the sum or average of the user payment amount data received for all shops within a historical time period, can be collectively referred to as the historical waybill acceptance data of the second delivery origin-destination region. The historical waybill acceptance data of the second delivery origin-destination region can also include the average delivery time data of all shops' orders within a historical time period. The historical time period is relative to the point in time when the logistics capacity model is used, that is, the historical time period before the point in time when the logistics capacity model is used. Spatiotemporal environmental characteristic data can refer to the spatiotemporal environmental characteristic data of the second delivery origin and destination at the time point when the logistics capacity model is used. Examples include characteristic data of the delivery period at which the logistics capacity model is used for each second delivery origin and destination (e.g., peak delivery periods), delivery difficulty data for each second delivery origin and destination (e.g., altitude difference between the origin AOI and destination AOI, delivery speed data from the origin AOI to the destination AOI), and environmental characteristic data for each second delivery origin and destination (e.g., meteorological characteristic data). AOI-AOI waybill data can refer to waybill data for at least one second delivery origin and destination over a future period, such as waybill volume. AOI-AOI waybill data can be the waybill data for each second delivery origin and destination within the capacity demand control area over a future period. AOI-AOI waybill data can also be referred to as multiple specified waybill data at the granularity of the second delivery origin and destination, or simply multiple specified waybill data for the second delivery origin and destination. AOI-AOI waybill data can be obtained from historical waybill data of the second delivery origin and destination, or from a trained waybill data prediction model, or from delivery parameters provided by other subsystems. Since system 101 can obtain production order forecast data, and since waybill data corresponds to order data, system 101 can provide production order forecast data to system 102. System 102 can then obtain AOI-AOI waybill data based on the production order forecast data. AOI-AOI waybill data can also be obtained based on configured logistics parameters. For example, logistics parameter configuration personnel can configure waybill data for a second delivery origin and destination, and system 102 can obtain AOI-AOI waybill data based on the configured waybill data for the second delivery origin and destination.

[0032] The input data for the logistics capacity model may also include at least one of the following: real-time distribution data of delivery resources, attribute data of AOIs (Area of ​​Interest), historical average fulfillment data per waybill, and data on incomplete real-time waybills. Real-time distribution data of delivery resources, also known as real-time capacity data, specifically refers to the real-time distribution data of delivery resources at the second delivery origin and destination points within the capacity demand control area, such as the number of delivery resources near the origin AOI. Attribute data of AOIs includes, for example, the location data of each origin AOI and each destination AOI, and the distance data between the origin and destination AOIs. Attribute data of AOIs can also be categorized under spatiotemporal environmental characteristic data. Historical average fulfillment data per waybill can refer to the historical average fulfillment data of waybills within the capacity demand control area, or the historical average fulfillment data of waybills at at least one second delivery origin and destination point, such as the historical average fulfillment data of waybills at each second delivery origin and destination point. Average fulfillment data per waybill may include, for example, historical average delivery time data. Incomplete real-time waybill data can refer to real-time data on waybills that have not yet been delivered, such as the real-time volume of incomplete waybills.

[0033] like Figure 2 As shown, after obtaining the above input data, the logistics capacity model can output average delivery time and / or wave duration. Average delivery time can refer to the average delivery time per order in the AOI-AOI waybill data. Wave duration can refer to the average duration of each delivery wave in the AOI-AOI waybill data. A delivery wave can refer to a complete delivery batch where the delivery resource picks up the delivery object from the originating AOI, delivers it to the destination AOI, and then returns to the originating AOI. The originating AOI reached by the delivery resource when picking up the delivery object may not be the same as the originating AOI it returns to. Average delivery time and / or wave duration can be called waybill fulfillment prediction data. If using... Figure 2The logistics capacity model shown obtains waybill capacity prediction data, and AOI-AOI waybill data can then be considered candidate waybill capacity prediction data. In practical applications, the first set of AOI-AOI waybill data can be input into the logistics capacity model. If the output average delivery time or wave duration meets the waybill fulfillment requirements, the first set of AOI-AOI waybill data can be identified as waybill capacity prediction data. If the output average delivery time or wave duration does not meet the waybill fulfillment requirements, the second set of AOI-AOI waybill data is then input into the logistics capacity model. This process continues until the AOI-AOI waybill data input into the logistics capacity model makes the average delivery time or wave duration output by the logistics capacity model meet the waybill fulfillment requirements, and the corresponding AOI-AOI waybill data at this point is identified as waybill capacity prediction data. That is, the system obtains the specified waybill data corresponding to the waybill fulfillment prediction data that matches the waybill fulfillment requirement data, uses this as the matching specified waybill data, and obtains the waybill capacity prediction data based on the matching specified waybill data. If system 105 utilizes... Figure 2 The logistics capacity model shown determines whether the expected fulfillment status corresponding to the waybill forecast data meets the waybill fulfillment requirements. Therefore, the AOI-AOI waybill data can refer to the waybill forecast data. Specifically, the waybill forecast data can be input into the logistics capacity model to obtain the average delivery time or wave duration output by the model. In other words, the logistics capacity model uses the average delivery time or wave duration required to complete the delivery of the waybill corresponding to the waybill forecast data. Afterwards, it can be determined whether the average delivery time or wave duration meets the waybill fulfillment requirements, thus realizing the determination of whether the expected fulfillment status corresponding to the waybill forecast data at the second delivery origin and destination corresponding to the capacity demand control reference data meets the waybill fulfillment requirements. If system 106 utilizes... Figure 2 The logistics capacity model shown determines whether the waybill forecast data meets the waybill fulfillment requirements. Therefore, the AOI-AOI waybill data can be the waybill forecast data. The relevant description of how System 106 uses the logistics capacity model to determine whether the waybill forecast data meets the waybill fulfillment requirements can be found in the above description and will not be repeated here.

[0034] Third Embodiment The commercial order distribution service subsystem may include, for example: Figure 3The commercial flow order model shown may include a first multi-head attention model 301 (hereinafter referred to as "Model 301"), a second multi-head attention model 302 (hereinafter referred to as "Model 302"), a vector concatenation model 303 (hereinafter referred to as "Model 303"), a vector high-order cross-model 304 (hereinafter referred to as "Model 304"), a parameter generation network model 305 (hereinafter referred to as "Model 305"), a feature cross-fusion model 306 (hereinafter referred to as "Model 306"), and a capacity demand control layer model 307 (hereinafter referred to as "Model 307"). Models 303, 304, and 305 can be combined or fused according to requirements. Figure 3 As shown, the long-term supply data of the starting position of the first delivery origin and destination, along with the long-term supply data sequence of the production order entities surrounding the starting position of the first delivery origin and destination, are input into model 301 (e.g., a multi-head attention model). This yields the output of model 301, which outputs a cross vector between the starting point and the supply around the starting point. In the multi-head attention model, each head has its own weight matrix and can learn different features, thus achieving parameter independence. It can learn multiple different "attention subspaces" in parallel, applying different linear transformations to the input vector, calculating the attention output for each head, and finally fusing the multi-head information to achieve information fusion and concatenation, ultimately outputting a cross vector, such as the cross vector between the starting point and the supply around the starting point. The starting position of the first delivery origin and destination (e.g., the starting position of the first delivery origin and destination) Figure 3 The "starting point location" shown can refer to the shop within the target SHOP-AOI (also known as the target OD). The target shop can also be referred to as the starting point location. Shops surrounding the target shop or the target AOI can be referred to as production order entities. The target SHOP-AOI can refer to the SHOP-AOI for which production order forecast data and regulatory benefit forecast data are needed; that is, the production order forecast data and regulatory benefit forecast data for the target SHOP-AOI need to be predicted using a commercial flow order model. The target SHOP-AOI can also be referred to as the SHOP-AOI to be predicted. Long-term supply data at the starting point location of the first delivery origin-destination point can refer to the long-term supply data of the shops within the target SHOP-AOI, such as at least one type of data including the target shop's order data and the target shop's user payment amount data. The target shop's order data can also be referred to as the target shop's historical production order data, and the target shop's user payment amount data can also be referred to as the target shop's historical operational data. Within a target shop's vicinity (within a preset distance range (e.g., 500 meters)), there can be at least one shop or order entity. Order entities existing around the target shop can be referred to as order entities surrounding the starting point of the first delivery origin-destination location. For example... Figure 3The inputs shown in the diagram are order entities 1 to n in model 301. The long-term supply data sequence of order entities surrounding the starting point may include at least one type of data, such as order data of order entities surrounding the starting point, distance data between order entities surrounding the starting point and the starting point of the first delivery origin-end point, and user payment data of order entities surrounding the starting point. Order data of order entities surrounding the starting point can also be referred to as the historical order count of order entities surrounding the starting point. Distance data between order entities surrounding the starting point and the starting point of the first delivery origin-end point may include the distance data between each order entity surrounding the starting point and the starting point of the first delivery origin-end point, or the average distance data between all order entities surrounding the starting point and the starting point of the first delivery origin-end point. User payment data of order entities surrounding the starting point can also be referred to as the historical operating data of order entities surrounding the starting point.

[0035] Please refer to Figure 3 The long-term supply data of the starting position of the first delivery origin-end point and the long-term supply data sequence of the order entities surrounding the ending position of the first delivery origin-end point are input into model 302 (e.g., a multi-head attention model) to obtain the supply cross vector of the starting point and the ending point perimeter output by model 302. The AOI in the target SHOP-AOI can be called the target AOI. There can be at least one SHOP or order entity in the vicinity of the target AOI (e.g., within a preset distance from the target AOI). The order entities existing in the vicinity of the target AOI can be called the order entities surrounding the ending position, for example... Figure 3 The input shown is to production order entities 1 to n in model 302. The description of the long-term supply data sequence of production order entities surrounding the endpoint can be found above in the description of the long-term supply data sequence of production order entities surrounding the starting point; it will not be repeated here.

[0036] Please refer to Figure 3The long-term auxiliary feature data of the first delivery origin and destination, the cross vector of supply between the origin and the surrounding area, and the cross vector of supply between the origin and the destination are input into model 303 (e.g., the Concat model) to obtain the long-term concatenated vector of the first delivery origin and destination. Model 303 can use feature concatenation as a neural network structure for fusing and regulating feature data. It can perform parallel combination of features along the feature dimension while retaining their original information, and concatenate different feature vectors end to end, allowing the subsequent network to automatically learn how to comprehensively utilize this information. The long-term auxiliary feature data of the first delivery origin and destination may include at least one type of data such as the origin location attribute feature data of the first delivery origin and destination, target AOI feature data, target OD feature data, and similar OD feature data. The origin location attribute feature data of the first delivery origin and destination can be simply referred to as the origin location feature data of the first delivery origin and destination, and may include at least one type of data such as the location feature data of the origin location of the first delivery origin and destination, the category data (e.g., fast food, hot pot, beverages, etc.) to which the origin location of the first delivery origin and destination belongs, the passenger flow data of the origin location of the first delivery origin and destination within a certain period of time (e.g., historical time period), and the profile feature data of the user group associated with the origin location of the first delivery origin and destination. Target AOI feature data may include at least one of the following: location feature data of the target AOI, category data of the order entities surrounding the target AOI, customer traffic data of the order entities surrounding the target AOI over a period of time, user profile feature data of the user groups associated with the order entities surrounding the target AOI, user profile feature data within the target AOI area, and user order data within the target AOI area. Target OD feature data may include at least one of the following: target OD order data, actual payment amount data of the target OD, distance data between the starting point of the first delivery origin and destination and the target AOI, and delivery difficulty data of the target OD. The target OD order data and the target OD actual payment amount data may not necessarily be included in the target OD feature data, depending on actual needs. In other words, the target OD order data and the target OD actual payment amount data may exist alongside the target OD feature data. Target OD order data may refer to order data generated from the starting point of the first delivery origin and destination with delivery addresses within the target AOI area. Target OD order data may also be referred to as the target OD's historical order data. The actual payment amount data for target ODs can refer to the actual payment amount data obtained for orders whose delivery addresses are within the target AOI range, based on the originating location of the first delivery origin-destination location. The actual payment amount data for target ODs can also be referred to as the historical operational data of the target ODs. Delivery difficulty data is used to represent the ease or difficulty of delivering orders whose delivery addresses are within the target AOI range, generated from the originating location of the first delivery origin-destination location.Similar ODs can refer to ODs surrounding the target OD, or ODs whose feature data is similar to the feature data of the target OD at a preset similarity threshold. At least one of the shops and areas of interest (AOIs) in the surrounding ODs can be within a preset distance range from at least one of the target shops and target AOIs in the target OD. Similar OD feature data can include at least one of the following: order data of similar ODs, actual payment amount data of users of similar ODs, distance data between shops and AOIs in similar ODs, and delivery difficulty data of similar ODs. Delivery difficulty data of the target OD does not necessarily need to be included in the target OD feature data; similarly, delivery difficulty data of similar ODs does not necessarily need to be included in the similar OD feature data. These two types of delivery difficulty data can be included as long-term auxiliary feature data. For related explanations of similar OD feature data, please refer to the above explanation of target OD feature data; it will not be repeated here.

[0037] Long-term auxiliary characteristic data for the first delivery origin and destination may also include at least one type of data, such as city characteristic data and trend characteristic data. City characteristic data can also be referred to as the characteristic data of the region where the capacity demand control area is located. City characteristic data may include at least one of the following: city order data and city user actual payment amount data. City order data may also be referred to as the city's historical order data, and city user actual payment amount data may also be referred to as the city's historical operational data. City characteristic data may also include at least one of the following: city level data, tourist season characteristic data, characteristic data of events held in the city, and city order time characteristic data. Trend characteristic data may include at least one type of data, such as the time data of the short-term order operation forecast data for the capacity demand control area to be predicted (also referred to as the specified forecast time data of the short-term order operation forecast data), the historical order data corresponding to the time data, and the historical operational data corresponding to the time data. The time data of the short-term order operation forecast data for the capacity demand control area to be predicted is used to indicate which time period's order operation forecast data is to be predicted. For example, the time data can be a specific date, indicating that the order operation forecast data for that date is to be predicted, that is, to predict the order operation forecast data for one day. Time data can also include the forecast time period for predicted production order data and regulatory benefit forecast data. For example, if you want to predict production order data and regulatory benefit forecast data within 30 minutes, then the time data needs to include this 30-minute time period. If the time data refers to a specific date, then that date should be the date of this 30-minute time period. The historical order data corresponding to the time data can refer to the order data corresponding to historical data at the same time. Historical data at the same time and the time data in the trend characteristic data can have the same attributes, such as both belonging to holidays or both belonging to weekdays. The order data corresponding to historical data at the same time can refer to the total number of orders or the average number of orders per unit time corresponding to historical data at the same time. The historical operational data corresponding to the time data can refer to the historical operational data corresponding to historical data at the same time.

[0038] Please refer to Figure 3The long-cycle concatenated vector of the first delivery origin and destination, output by model 303, is input into model 304 (e.g., the xDeepFM model) to obtain the high-order cross vector of the first delivery origin and destination, output by model 304. This high-order cross vector is then input into model 305 to obtain the short-cycle operational prediction parameters output by model 305 for predicting short-cycle order operational prediction data for the first delivery origin and destination. Model 304 can simultaneously model an end-to-end model of explicit and implicit high-order feature interactions, effectively integrating linear models, explicit high-order interactions (CIN), and implicit high-order interactions (DNN). This achieves explicit, controllable, and high-order modeling of feature interactions, overcoming the limitation of traditional DNNs in interpreting feature combinations. Model 305 can output parameter data corresponding to various data types, such as parameter data corresponding to at least one type of data, including order (Production) prediction data, revenue (Income) prediction data, user actual payment amount (GMV, Gross Merchandise Volume, also known as total merchandise transaction amount) prediction data, and cost (Cost) prediction data. Therefore, short-cycle operational forecasting parameters can include parameter data corresponding to at least one type of data, such as order forecasting data, revenue forecasting data, user actual payment forecasting data, and cost forecasting data.

[0039] Model 305 can employ a Dynamic Parameter Generative Network (PGN), which considers configuring different weights for different feature data. These different weights affect the model's output. For example, if the model considers configuring different weights for feature data from different cities, the prediction results for order operation forecast data corresponding to different cities will also differ, thus achieving a "hundred cities, thousand models" effect. Assuming that the user's actual payment amount corresponding to the starting position C of the first delivery origin-destination in the fourth OD is relatively large, then the weight of the user's actual payment amount at the starting position C of the first delivery origin-destination can be set higher. Since the higher-order cross vector of the first delivery origin-destination output by Model 304 already considers the differences between different cities, different delivery origin-destinations, and even different feature data—that is, considering configuring different weights for different feature data—the short-cycle order operation forecast data obtained by multiplying the higher-order cross vector with the short-cycle operation forecast parameters not only reflects the differences in scenarios such as cities and delivery origin-destinations but also better matches the actual scenario, thus resulting in more accurate prediction results.

[0040] Model 305 can multiply short-cycle operational forecasting parameter data with a higher-order cross vector to output short-cycle order operational forecasting data corresponding to the time data in the trend feature data. Short-cycle order operational forecasting data can include at least one of the following: order forecasting data, revenue forecasting data, user actual payment forecasting data, and cost forecasting data.

[0041] like Figure 3 As shown, the obtained short-cycle order operation forecast data can be used as input data for model 306. Model 306 can be a DNN (Deep Neural Network) model, which can automatically learn hierarchical feature representations from the original input through multi-layer nonlinear transformations, ultimately completing tasks such as classification and regression. In addition to short-cycle order operation forecast data, the input data of model 306 can also include real-time feature data, such as real-time order feature data. Real-time order feature data can include at least one of cumulative data and delivery cost data. Cumulative data can refer to the cumulative real-time data of orders at the first delivery origin and destination. Delivery cost data can refer to the real-time data of the average delivery cost per order. Real-time feature data can also include at least one of the following: capacity demand control period feature data, weather feature data, and logistics pressure data. Capacity demand control period feature data can include a time period data from the time data in the trend feature data. If, at the current point in time, models 306 and 307 are used to predict revenue and order volume for periods 1 to n (hereinafter referred to as T1 to Tn), and the current point in time is the start time of a certain period, then the capacity demand control period characteristic data includes the period data with a duration of 24 / n after the start of the current point in time. The data output by model 307 can include the expected revenue and expected order volume for T1 to Tn (e.g., each duration of T1 to Tn is 30 minutes) within the next day after the start of the current point in time. The capacity demand control period characteristic data can also include data reflecting the characteristics of the period, such as data indicating whether the period is a peak or off-peak delivery period, data indicating the time sequence of the period (morning, noon, evening, etc.), and data indicating the time sequence of the period among all periods of the day. Real-time weather characteristic data can include weather characteristic data when using models 306 and 307 to predict revenue and order volume. Real-time capacity pressure data can include logistics pressure coefficients used to represent logistical pressure. Specifically, these can refer to the logistics pressure coefficients used when predicting revenue and order volume using Models 306 and 307. The logistics pressure coefficients can be those for capacity demand control areas, AOI-AOI granularity, or SHOP-AOI granularity. If the capacity demand control period characteristic data, weather characteristic real-time data, and logistics pressure real-time data are all real-time data, then these can all be categorized as the aforementioned order characteristic real-time data, or collectively referred to as real-time characteristic data or real-time data.

[0042] After obtaining the aforementioned short-cycle order operation forecast data and real-time feature data, they are converted into corresponding short-cycle order operation forecast vectors and real-time feature vectors, respectively. The real-time feature vectors and short-cycle order operation forecast vectors are then input into model 306 for feature cross-fusion processing to obtain the real-time feature operation fusion vector output by model 306. Multiple capacity demand regulation feature data are then converted into multiple capacity demand regulation feature vectors. The real-time feature operation fusion vector and the multiple capacity demand regulation feature vectors are input into model 307. Model 307 uses the multiple capacity demand regulation feature vectors as regulation terms to perform vector regulation processing on the real-time feature operation fusion vector, obtaining the order production forecast data and regulation benefit forecast data for each time period from T1 to Tn within a future period. Model 307 can be an Uplift layer model used to quantify the impact of capacity demand regulation feature data on user ordering behavior. Based on capacity demand regulation feature data and whether or not intervention is accepted, this model extracts feature identifiers and predicts the predicted order volume of users in the "with intervention" and "without intervention" scenarios through two parallel branches. Through this model, the truly sensitive groups of users who can place orders can be identified, and more budget and resources can be invested in these users who can "influence" the ordering behavior.

[0043] Models 306 and 307 can be trained as follows: Multiple capacity demand regulation feature data samples for the first delivery origin-end sample, real-time feature data samples for the first delivery origin-end sample, short-cycle order operation prediction data samples for the first delivery origin-end sample, and order label data and regulation benefit label data for the first delivery origin-end sample in the future period, expected based on the multiple capacity demand regulation feature data samples, real-time feature data samples, and short-cycle order operation prediction data samples; input the real-time feature data samples and short-cycle order operation prediction data samples into the initial feature cross-fusion model to obtain the real-time feature operation fusion vector sample of the first delivery origin-end sample output by the initial feature cross-fusion model; input the real-time feature operation fusion vector sample and multiple capacity demand regulation feature data samples into the initial capacity demand regulation layer model to obtain the order prediction data and regulation benefit prediction data for the first delivery origin-end sample in the future period output by the initial capacity demand regulation layer model; input the order prediction data and regulation benefit prediction data of the first delivery origin-end sample in the future period, as well as the order label data and regulation benefit label data, into the model for... In the loss function that evaluates the degree of loss in order prediction and the degree of loss in regulation benefit prediction, the order prediction and regulation benefit prediction loss data between the order prediction data and regulation benefit prediction data of the first delivery origin-end sample in the future period and the order label data and regulation benefit label data are obtained. If the order prediction and regulation benefit prediction loss data are acceptable loss data, the initial feature cross-fusion model is determined as an applicable feature cross-fusion model, and the initial capacity demand regulation layer model is determined as an applicable capacity demand regulation layer model. Otherwise, the model parameters of the initial feature cross-fusion model and the initial capacity demand regulation layer model are adjusted until the order prediction and regulation benefit prediction loss data between the order prediction data and regulation benefit prediction data of the first delivery origin-end sample in the future period and the order label data and regulation benefit label data obtained according to the adjusted feature cross-fusion model and capacity demand regulation layer model are acceptable loss data. The adjusted feature cross-fusion model and the adjusted capacity demand regulation layer model are then determined as applicable feature cross-fusion models and capacity demand regulation layer models.

[0044] Models 306 and 307 can predict the revenue and order volume of any OD at any point in time and in any period of time, according to actual needs. This not only allows for real-time prediction of revenue and order volume, but also for prediction of revenue and order volume over a more precise geographical area, and for prediction of revenue and order volume over periods of time, thus meeting the timeliness requirements for revenue and order volume prediction and improving the accuracy of revenue and order volume over a geographical area.

[0045] Fourth embodiment The following is combined Figure 4 This section describes the operation of the real-time handshake service subsystem. In the near-real-time planning unit 401 (hereinafter referred to as "Unit 401"), the logistics capacity assessment subunit 4013 (hereinafter referred to as "Subunit 4013") inputs the waybill distribution (i.e., the merchant's waybill capacity prediction data) under the waybill fulfillment requirements over a certain period of time into the real-time handshake service subunit 4012 (hereinafter referred to as "Subunit 4012"). The commercial flow order assessment subunit 4011 (hereinafter referred to as "Subunit 4011") inputs the order volume and revenue at the OD granularity under different capacity demand control characteristic data into Subunit 4012. Subunit 4012 establishes a system based on the capacity demand control characteristic data and revenue... Figure 4 The "Control-Revenue" coordinate system shown depicts "Control" representing the characteristic data of capacity demand control, and "Revenue" representing revenue. Using waybill distribution and order volume as constraints, a first target value for revenue is determined, such as the maximum value. The capacity demand control characteristic data corresponding to the revenue of the first target value is then determined as the reference data for capacity demand control.

[0046] The following explains the method for determining reference data for capacity demand regulation using expressions. Expression (1) is used to calculate the maximum revenue (also called the maximum income). Expressions (2), (3), and (4) represent constraints. Expressions (5) and (6) explain some parameters in expression (1). Expressions (1), (2), (3), (4), (5), and (6) are as follows: (1) (2) (3) (4) (5) (6) In expression (1), i is the number of the first delivery origin and destination in the capacity demand control area, j is the number of the capacity demand control feature data for the capacity demand control area, I is the set of all the numbers of the first delivery origin and destination in the capacity demand control area, and J is the set of all the numbers of the capacity demand control feature data for the capacity demand control area. (Production) is the first delivery origin and destination with the number i (which can be denoted as...). The order volume under capacity demand regulation feature data numbered j (which can be denoted as capacity demand regulation feature data j) yes In the characteristic data of capacity demand regulation The revenue (also known as average revenue per order) is affected by logistics pricing.

[0047] In expression (1), For use to indicate targeting The indicator label for whether to use capacity demand regulation feature data j can be 0 or 1. When the value is 1, it indicates that it is for... Using the capacity demand regulation feature data j, when the value is 0, it indicates that the regulation is targeted at... The capacity demand control feature data j was not used. Since only one capacity demand control feature data can be used for any given first delivery origin and destination at any given time, it is possible to use a different feature data for each destination at the same time. Determine the corresponding capacity demand regulation feature data to be used Setting the value to 1 will target Other capacity demand regulation characteristic data corresponding to The value is set to 0. Assuming this is for... Three capacity demand regulation feature data were set, and the feature data numbers are 1, 2, and 3 respectively. Then, in expression (1), the following is involved: The expression is: This will involve The expression is called expression (7). If ,but Then expression (7) can be equivalent to Similarly, if ,but Then expression (7) can be equivalent to Similarly, if ,but Then expression (7) can be equivalent to Therefore, it can be seen that The value of determines the capacity demand control feature data set for all first delivery origins and destinations at the same time point in expression (1). In expression (1), the same capacity demand control feature data can also be set for at least two different first delivery origins and destinations at the same time point.

[0048] In expression (1), This represents the maximum total revenue (or total income) of the capacity demand control area obtained after applying determined capacity demand control reference data to each first delivery origin and destination within the capacity demand control area. The total revenue of the capacity demand control area is a predicted value. In expression (1), the total revenue of the capacity demand control area can be determined based on the product of the order volume, revenue, and indicator label value for each first delivery origin and destination under the capacity demand control feature data. After obtaining all the total revenues from the capacity demand control feature data, the maximum value is selected. Assume that the capacity demand control area has 3 first delivery origins and destinations, denoted as... , and For these three primary delivery origin and destination points, three capacity demand control feature data were set, numbered 1, 2, and 3 respectively. Expression (1) can be specifically expressed as: Let this expression be denoted as expression (8). If it is... Configure capacity demand regulation feature data 2, for Configure capacity demand regulation feature data 3, for Configure capacity demand regulation feature data 1, then , and The value of is 1. , , , , and The value of is 0. According to expression (8), the total revenue can be expressed as: It can be simultaneously for 2. Configuration of capacity demand regulation characteristic data Configured capacity demand regulation feature data 3 and for The configured capacity demand regulation feature data 1 is called a capacity demand regulation feature data combination. Similarly, if it is... , and At least one OD configuration in the data has different capacity demand regulation characteristic data from the capacity demand regulation characteristic data configured for that OD, or in other words, if for , and By configuring another combination of capacity demand regulation feature data, a different value for the total revenue can be obtained. And so on, when configuring... , and After configuring multiple combinations of capacity demand regulation feature data, multiple values ​​of total revenue can be obtained. According to expression (8), the maximum value of the total revenue value can be selected from the multiple values ​​of total revenue.

[0049] This indicates that expression (1) must simultaneously satisfy the constraints represented by expressions (2), (3), and (4), as well as the explanations for expressions (5) and (6). Since the above has already... The meaning and significance of the expressions have been explained, so only expressions (2), (3) and (4) will be explained below, and expressions (5) and (6) will not be repeated.

[0050] In expressions (2), (3), and (4), all have Such a function, It can represent the distribution of waybills provided by the logistics side. express The number of sampling orders, for The distance data between the starting and ending positions. In expression (2), This represents the minimum sum of all waybill volumes for which the distance between the starting and ending points of the first delivery origin and destination is less than or equal to 3 kilometers. This represents the sum of orders where the distance between the starting and ending points of the first delivery origin and destination is less than or equal to 3 kilometers. This represents the maximum sum of all waybill volumes within a distance of 3 kilometers between the origin and destination of the first delivery origin-destination point. Expression (2) generally indicates that the sum of order volumes within a 3-kilometer distance between the origin and destination of the first delivery origin-destination point is not less than the minimum sum of all waybill volumes within a 3-kilometer distance between the origin and destination of the first delivery origin-destination point, and does not exceed the maximum sum of all waybill volumes within a 3-kilometer distance between the origin and destination of the first delivery origin-destination point. In other words, the sum of order volumes within a 3-kilometer distance between the origin and destination of the first delivery origin-destination point needs to match the available transportation capacity provided by the logistics side for the first delivery origin-destination point within a 3-kilometer distance.

[0051] Expression (3) generally indicates that the sum of all orders with a distance greater than 3 kilometers between the starting and ending points of the first delivery origin and destination is greater than or equal to 0.9 times the minimum sum of all waybills with a distance greater than 3 kilometers between the starting and ending points of the first delivery origin and destination, and less than or equal to 0.9 times the maximum sum of all waybills with a distance greater than 3 kilometers between the starting and ending points of the first delivery origin and destination. In other words, the sum of orders with a distance greater than 3 kilometers between the starting and ending points of the first delivery origin and destination also needs to match the transportation capacity that the logistics side can provide for the first delivery origin and destination with a distance greater than 3 kilometers between the starting and ending points. 0.9 can be regarded as weight data. The reason why 0.9 is used as weight data in expression (3) and 1 is used as weight data in expression (2) is that the order volume with a distance of more than 3 kilometers between the starting and ending points may be relatively small and the delivery fee may be high. Therefore, the weight of the waybill volume with a distance of more than 3 kilometers between the starting and ending points is set lower than the weight of the waybill volume with a distance of less than 3 kilometers between the starting and ending points. The weight value can be set according to actual needs.

[0052] Expression (4) generally indicates that the sum of all order volumes at the first delivery origin and destination is greater than or equal to the minimum sum of all waybill volumes at the first delivery origin and destination, and less than or equal to the maximum sum of all waybill volumes at the first delivery origin and destination. In other words, without considering the distance between the origin and destination locations of the first delivery origin and destination, the sum of order volumes at the first delivery origin and destination needs to match the transportation capacity that the logistics side can provide for the first delivery origin and destination.

[0053] After configuring multiple combinations of capacity demand regulation feature data for multiple ODs, multiple values ​​of total revenue can be obtained. Due to the constraints of expressions (2), (3), and (4), the constraints of expressions (2), (3), and (4) should be satisfied simultaneously during the calculation of total revenue. In other words, the multiple values ​​of total revenue should be calculated under the constraints of expressions (2), (3), and (4). According to expression (1), the maximum value of the total revenue value can be selected, and then the capacity demand regulation feature data set to obtain the maximum total revenue value can be determined as the reference data for capacity demand regulation.

[0054] Please refer to this again. Figure 4After obtaining the capacity demand control reference data, subunit 4012 can send the capacity demand control reference data to real-time control unit 402. After receiving the capacity demand control reference data sent by subunit 4012, the waybill demand control execution reference data processing subunit 4021 (hereinafter referred to as "subunit 4021") can send the capacity demand control reference data to real-time control subunit 4022 (hereinafter referred to as "subunit 4022"). After executing the capacity demand control reference data at the control timing, subunit 4022 sends the capacity demand control reference data to the corresponding processing unit, which then executes the capacity demand control reference data. Subunit 4022 can also execute the capacity demand control reference data after being triggered by the logistics real-time risk assessment subunit 4023 (hereinafter referred to as "subunit 4023"). Subunit 4023 can assess logistics performance risks over a future period of time. If logistics performance risks exist, subunit 4023 can send information representing logistics performance risks or a capacity demand adjustment request message to subunit 4022 to trigger subunit 4022 to execute the capacity demand adjustment reference data obtained from subunit 4012.

[0055] During or after the execution of the capacity demand regulation reference data by subunit 4022, the monitoring and result review unit 403 (hereinafter referred to as "unit 403") detects the execution status and results of the capacity demand regulation reference data. The monitoring subunit 4031 within unit 403 detects whether the capacity demand regulation reference data has been effectively executed, for example, by detecting the difference between the parameter data corresponding to the capacity demand regulation reference data and the actual parameter data. The result review unit 4032 within unit 403 can detect the execution results of the capacity demand regulation reference data and can perform subsequent necessary work based on the execution results, such as adjusting logistics costs.

[0056] Fifth Embodiment In this embodiment of the application, before adjusting capacity demand, it is necessary to determine reference data for capacity demand adjustment. Therefore, this embodiment provides a method for obtaining reference data for capacity demand adjustment. Please refer to... Figure 5The method for obtaining reference data for capacity demand regulation provided in this application embodiment may include the following steps: Step S501: Obtaining order prediction data and regulation benefit prediction data for the first delivery origin and destination in the capacity demand regulation area under different capacity demand regulation characteristic data, including: obtaining long-cycle characteristic data of the first delivery origin and destination, which includes long-cycle supply data of the starting position of the first delivery origin and destination, long-cycle supply data sequence of order entities around the starting position of the first delivery origin and destination, and long-cycle supply data sequence of order entities around the destination of the first delivery origin and destination; obtaining short-cycle order operation prediction data of the first delivery origin and destination based on the long-cycle supply data of the starting position of the first delivery origin and destination, the long-cycle supply data sequence of order entities around the starting position of the first delivery origin and destination, and the long-cycle supply data sequence of order entities around the destination of the first delivery origin and destination, where the duration of the short cycle in the short-cycle order operation prediction data is shorter than that of the long cycle. The process involves: obtaining long-term characteristic data; acquiring real-time characteristic data for the first delivery origin and destination, and acquiring multiple capacity demand control characteristic data for the first delivery origin and destination; obtaining order production prediction data and control benefit prediction data for the first delivery origin and destination based on short-term order operation forecast data, real-time characteristic data, and multiple capacity demand control characteristic data for the first delivery origin and destination; Step S502: obtaining order acceptance capacity prediction data for the second delivery origin and destination in the capacity demand control area that meets the order fulfillment requirements, wherein the starting position of the second delivery origin and destination includes the starting position of the first delivery origin and destination, and the ending position of the second delivery origin and destination includes the ending position of the first delivery origin and destination; Step S503: obtaining capacity demand control reference data for the capacity demand control area based on the order acceptance capacity prediction data, order production prediction data, and control benefit prediction data, wherein the capacity demand control reference data includes capacity demand control reference data determined for at least one first delivery origin and destination.

[0057] Step S501 can be understood as determining order distribution data and order benefit data at the first delivery origin-destination level for a future period of time (e.g., 30 minutes) on the commercial flow side. For example, determining the order volume and revenue of each first delivery origin-destination in at least one first delivery origin-destination within the capacity demand control area for a future period of time, or the revenue of all first delivery origin-destinations. One implementation of step S501 is as follows: inputting short-cycle order operation forecast data and real-time feature data of the first delivery origin-destination into a feature cross-fusion model to obtain the real-time feature operation fusion vector of the first delivery origin-destination output by the feature cross-fusion model; inputting multiple capacity demand control feature data and the real-time feature operation fusion vector of the first delivery origin-destination into a capacity demand control layer model to obtain the order prediction data and control benefit prediction data of the first delivery origin-destination for a future period of time output by the capacity demand control layer model. The order prediction data and regulation benefit prediction data for the first delivery origin and destination can be obtained as follows: Obtain the real-time feature vector corresponding to the real-time feature data of the first delivery origin and destination; obtain the short-cycle order operation prediction vector corresponding to the short-cycle order operation prediction data of the first delivery origin and destination; perform feature cross-fusion processing on the real-time feature vector and the short-cycle order operation prediction vector to obtain the real-time feature operation fusion vector of the first delivery origin and destination; input multiple capacity demand regulation feature data and the real-time feature operation fusion vector of the first delivery origin and destination into the capacity demand regulation layer model to obtain the order prediction data and regulation benefit prediction data of the first delivery origin and destination for a future period output by the capacity demand regulation layer model, including: obtaining multiple capacity demand regulation feature vectors corresponding to the multiple capacity demand regulation feature data of the first delivery origin and destination; using the multiple capacity demand regulation feature vectors as regulation terms to perform vector regulation processing on the real-time feature operation fusion vector of the first delivery origin and destination to obtain the order prediction data and regulation benefit prediction data of the first delivery origin and destination for a future period.

[0058] The long-term characteristic data of the first delivery origin and destination may also include long-term auxiliary characteristic data of the first delivery origin and destination; the short-term order operation forecast data of the first delivery origin and destination can be obtained in the following manner: inputting the long-term supply data of the starting position of the first delivery origin and destination and the long-term supply data sequence of the order entities around the starting position of the first delivery origin and destination into the first multi-head attention model to obtain the starting point and surrounding supply cross vector output by the first multi-head attention model; inputting the long-term supply data of the starting position of the first delivery origin and destination and the long-term supply data sequence of the order entities around the ending position of the first delivery origin and destination into the second multi-head attention model to obtain the starting point and surrounding supply cross vector output by the second multi-head attention model; The long-term auxiliary feature data of the delivery origin and destination, the cross vectors of the origin and the supply around the origin, and the cross vectors of the origin and the destination around the supply are input into a vector concatenation model to obtain the long-term concatenation vector of the first delivery origin and destination. The long-term concatenation vector of the first delivery origin and destination is input into a higher-order vector cross model to obtain the higher-order cross vector of the first delivery origin and destination output by the higher-order vector cross model. The higher-order cross vector of the first delivery origin and destination is input into a parameter generation network model to obtain the short-term operation prediction parameter data output by the parameter generation network model for predicting the short-term order operation prediction data of the first delivery origin and destination. Based on the higher-order cross vector and the short-term operation prediction parameter data of the first delivery origin and destination, the short-term order operation prediction data of the first delivery origin and destination is obtained.

[0059] The first multi-head attention model, the second multi-head attention model, the vector concatenation model, the vector high-order cross model, and the parameter generation network model can be trained as follows: Obtain long-term auxiliary feature data samples of the first delivery origin-end point sample, long-term supply data samples of the starting position of the first delivery origin-end point sample, long-term supply data samples of the order entities surrounding the starting position of the first delivery origin-end point sample, long-term supply data samples of the order entities surrounding the ending position of the first delivery origin-end point sample, and based on the long-term auxiliary feature data samples, the long-term supply data samples of the starting position, the long-term supply data samples of the order entities surrounding the starting position, and the ending position... The first delivery origin-end point sample is expected to obtain short-cycle order operation tag data by taking long-cycle supply data samples of surrounding order entities; the long-cycle supply data samples of the origin position and the surrounding order entities are input into the initial first multi-head attention model to obtain the origin and surrounding supply cross vector output by the initial first multi-head attention model; the long-cycle supply data samples of the origin position and the surrounding order entities are input into the initial second multi-head attention model to obtain the origin and surrounding supply cross vector output by the initial second multi-head attention model; the long-cycle supply data samples of the origin position and the surrounding order entities are input into the initial second multi-head attention model to obtain the origin and surrounding supply cross vector output by the initial second multi-head attention model; the long-cycle supply data samples of the first delivery origin-end point sample are input into the initial second multi-head attention model to obtain the short-cycle order operation tag data of the surrounding order entities; the long-cycle supply data samples of the origin and ... short-cycle order operation tag data of the surrounding order entities; the long-cycle supply data samples of the origin and surrounding order entities are input into the initial second multi-head attention model to obtain the short-cycle order operation tag data of the surrounding order entities; the long-cycle supply data samples of the origin and surrounding order entities are input into the initial second multi-head attention model to obtain the short-cycle order operation tag data of the surrounding order entities; the long-cycle supply data samples of the origin and surrounding order entities are input into the initial second multi-head attention model to obtain the short-cycle order operation tag data of the surrounding order entities; the long-cycle supply data samples of the origin and surrounding order entities are input into the initial second multi-head attention model to obtain the short-cycle order operation tag data of the surrounding order entities; the long Auxiliary feature data samples, the start-point and start-point periphery supply cross vectors output by the initial first multi-head attention model, and the start-point and end-point periphery supply cross vectors output by the initial second multi-head attention model are input into the initial vector concatenation model to obtain the long-period concatenation vector of the first delivery start-point and end-point samples. The long-period concatenation vector of the first delivery start-point and end-point samples is then input into the initial vector high-order cross model to obtain the high-order cross vector of the first delivery start-point and end-point samples output by the initial vector high-order cross model. Finally, the high-order cross vector of the first delivery start-point and end-point samples is input into the initial parameter generation network model to obtain the short-period order vector output by the initial parameter generation network model for predicting the first delivery start-point and end-point samples. The short-cycle operational forecasting parameter data of the first delivery origin-end point sample of the single operational forecasting data; based on the higher-order cross vector of the first delivery origin-end point sample and the short-cycle operational forecasting parameter data of the first delivery origin-end point sample, the short-cycle order operational forecasting data of the first delivery origin-end point sample is obtained; the short-cycle order operational forecasting data of the first delivery origin-end point sample and the short-cycle order operational label data of the first delivery origin-end point sample are input into a loss function used to evaluate the degree of short-cycle order operational forecasting loss, and the short-cycle order operational forecasting loss data between the short-cycle order operational forecasting data of the first delivery origin-end point sample and the short-cycle order operational label data of the first delivery origin-end point sample is obtained;If the predicted loss data for short-cycle order operations is acceptable, then the initial first multi-head attention model, initial second multi-head attention model, initial vector concatenation model, initial higher-order vector cross model, and initial parameter generation network model are respectively determined as the applicable first multi-head attention model, applicable second multi-head attention model, applicable vector concatenation model, applicable higher-order vector cross model, and applicable parameter generation network model. Otherwise, the model parameters of the initial first multi-head attention model, initial second multi-head attention model, initial vector concatenation model, initial higher-order vector cross model, and initial parameter generation network model are adjusted until the adjusted model parameters are used for the first multi-head attention model. The short-term order operation prediction loss data between the short-term operation prediction parameter data of the first delivery origin-end point sample obtained by the force model, the second multi-head attention model, the vector concatenation model, the vector higher-order cross model, and the parameter generation network model, and the short-term order operation label data of the first delivery origin-end point sample, is determined to be acceptable until the adjusted first multi-head attention model, second multi-head attention model, vector concatenation model, vector higher-order cross model, and parameter generation network model are respectively identified as applicable first multi-head attention model, applicable second multi-head attention model, applicable vector concatenation model, applicable vector higher-order cross model, and applicable parameter generation network model.

[0060] The long-term auxiliary characteristic data of the first delivery origin and destination includes at least one of the following: short-term order operation forecast data of the first delivery origin and destination at a specified forecast time; characteristic data of the first delivery origin and destination; historical order data and historical operation data of the first delivery origin and destination at a historical time matching the specified forecast time data; characteristic data of the starting position of the first delivery origin and destination; characteristic data of the ending position of the first delivery origin and destination; historical order data and historical operation data of similar delivery origins and destinations of the first delivery origin and destination; and historical order data and historical operation data of the region where the capacity demand control area is located. The long-term supply data includes historical order data and historical operational data of the starting point of the first delivery origin and destination; the long-term supply data sequence of order entities surrounding the starting point of the first delivery origin and destination includes historical order data and historical operational data of order entities surrounding the starting point of the first delivery origin and destination; the long-term supply data sequence of order entities surrounding the ending point of the first delivery origin and destination includes historical order data and historical operational data of order entities surrounding the ending point of the first delivery origin and destination; the short-term order operation forecast data of the first delivery origin and destination includes short-term order operation forecast data of the first delivery origin and destination at a specified forecast time.

[0061] The starting point and surrounding supply cross vector output by the first multi-head attention model can also be obtained as follows: Obtain the long-term supply vector corresponding to the long-term supply data of the starting point of the first delivery origin-end point; obtain the long-term supply vector of the production order entities surrounding the starting point of the first delivery origin-end point corresponding to the long-term supply data sequence of the production order entities surrounding the starting point of the first delivery origin-end point; perform multi-head attention cross processing on the long-term supply vector of the starting point and the long-term supply vector of the production order entities surrounding the starting point of the starting point to obtain the starting point and surrounding supply cross vector. The starting point and surrounding supply cross vector output by the second multi-head attention model can also be obtained as follows: Obtain the long-term supply vector corresponding to the long-term supply data of the starting point of the first delivery origin-end point; obtain the long-term supply vector of the production order entities surrounding the ending point corresponding to the long-term supply data sequence of the production order entities surrounding the ending point of the first delivery origin-end point; perform multi-head attention cross processing on the long-term supply vector of the starting point and the long-term supply vector of the production order entities surrounding the ending point of the first delivery origin-end point to obtain the starting point and surrounding supply cross vector. Furthermore, the long-cycle auxiliary feature vector corresponding to the long-cycle auxiliary feature data of the first delivery origin and destination is obtained. The cross vectors of the origin and the supply around the origin, the cross vectors of the origin and the supply around the destination, and the long-cycle auxiliary feature vector are concatenated to obtain the long-cycle concatenated vector of the first delivery origin and destination.

[0062] Short-cycle order operation forecast data for the first delivery origin and destination can also be obtained in the following ways: obtain the specified forecast time data for the order operation forecast data of the first delivery origin and destination; obtain the characteristic data of the first delivery origin and destination; obtain the historical production order data and historical operation data of the starting position of the first delivery origin and destination; obtain the historical production order data and historical operation data of the production order entities surrounding the starting position of the first delivery origin and destination; obtain the historical production order data and historical operation data of the production order entities surrounding the destination position of the first delivery origin and destination; based on the specified forecast time data, the characteristic data of the first delivery origin and destination, the historical production order data and historical operation data of the starting position, the historical production order data and historical operation data of the production order entities surrounding the starting position, and the historical production order data and historical operation data of the production order entities surrounding the destination position, obtain the order operation forecast data for the first delivery origin and destination at the specified forecast time.

[0063] Furthermore, it is also possible to obtain historical order data and historical operation data of the first delivery origin and destination at historical times that match the specified forecast time data, characteristic data of the starting position of the first delivery origin and destination, characteristic data of the ending position of the first delivery origin and destination, historical order data and historical operation data of similar delivery origins and destinations of the first delivery origin and destination, and historical order data and historical operation data of the region where the capacity demand control area is located. At this time, order operation forecast data can also be obtained in the following ways: based on the specified forecast time data of the order operation forecast data of the first delivery origin and destination, the historical order data and historical operation data of the first delivery origin and destination at the historical time matching the specified forecast time data, the characteristic data of the starting position of the first delivery origin and destination, the characteristic data of the ending position of the first delivery origin and destination, the characteristic data of the first delivery origin and destination, the historical order data and historical operation data of the starting position, the historical order data and historical operation data of the order entities around the starting position, the historical order data and historical operation data of the order entities around the ending position, the historical order data and historical operation data of similar delivery origins and destinations of the first delivery origin and destination, and the historical order data and historical operation data of the region where the capacity demand control area is located, the order operation forecast data of the first delivery origin and destination at the specified forecast time can be obtained.

[0064] In practical implementation, a feature cross-fusion model and a capacity demand regulation layer model can be used to obtain order prediction data and regulation benefit prediction data. For example, short-cycle order operation prediction data and real-time feature data are input into the feature cross-fusion model to obtain a real-time feature operation fusion vector. Then, the real-time feature operation fusion vector and multiple capacity demand regulation feature data are input into the capacity demand regulation layer model to obtain the order prediction data and regulation benefit prediction data for the first delivery origin and destination under multiple capacity demand regulation feature data output by the capacity demand regulation layer model.

[0065] To obtain order forecast data and regulation benefit forecast data for the first delivery origin and destination under multiple capacity demand regulation characteristic data, at least one of the following data can also be obtained: capacity demand regulation period characteristic data for the first delivery origin and destination, real-time weather characteristic data for the first delivery origin and destination, and real-time capacity pressure data for the first delivery origin and destination. Furthermore, based on at least one of the following data: capacity demand regulation period characteristic data, real-time weather characteristic data, and real-time capacity pressure data, as well as capacity demand regulation factors, real-time characteristic data, and short-cycle order operation forecast data, order forecast data and regulation benefit forecast data for the first delivery origin and destination under multiple capacity demand regulation characteristic data can be obtained.

[0066] For a description of step S501, please refer to the descriptions of the first and third embodiments, which will not be repeated here.

[0067] Step S502 can be understood as obtaining the forecast data of the waybill acceptance capacity at the second delivery origin and destination granularity in the capacity demand control area within a future period, which meets the waybill fulfillment requirements. For example, obtaining the expected waybill acceptance volume at at least one second delivery origin and destination in the capacity demand control area within a future period, which meets the waybill fulfillment requirements. One implementation of step S502 is as follows: obtaining multiple designated waybill data for the second delivery origin and destination; obtaining historical waybill acceptance data for the second delivery origin and destination; obtaining spatiotemporal environmental characteristic data for the second delivery origin and destination; obtaining waybill fulfillment forecast data for the capacity demand control area under multiple designated waybill data based on the multiple designated waybill data, the historical waybill acceptance data, and the spatiotemporal environmental characteristic data of the second delivery origin and destination; obtaining waybill fulfillment requirement data; obtaining designated waybill data corresponding to the waybill fulfillment forecast data that matches the waybill fulfillment requirement data, as the matching designated waybill data; obtaining waybill acceptance capacity forecast data based on the matching designated waybill data. To obtain waybill fulfillment forecast data for the capacity demand control area under multiple specified waybill data, it is also possible to obtain real-time capacity data of the second delivery origin and destination, historical average waybill fulfillment data of the second delivery origin and destination, and real-time incomplete waybill data of the second delivery origin and destination. Furthermore, based on multiple specified waybill data of the second delivery origin and destination, real-time capacity data of the second delivery origin and destination, historical waybill acceptance data of the second delivery origin and destination, historical average waybill fulfillment data of the second delivery origin and destination, real-time incomplete waybill data of the second delivery origin and destination, and spatiotemporal environmental characteristic data of the second delivery origin and destination, waybill fulfillment forecast data for the capacity demand control area under multiple specified waybill data can be obtained.

[0068] In practical applications, neural network models can be used to obtain waybill fulfillment prediction data for capacity demand control areas under multiple specified waybill data. For example, multiple specified waybill data for the second delivery origin and destination, historical waybill acceptance data for the second delivery origin and destination, and spatiotemporal environmental characteristic data of the second delivery origin and destination can be input into the waybill fulfillment data prediction model to obtain the waybill fulfillment prediction data for the capacity demand control area under multiple specified waybill data output by the waybill fulfillment data prediction model.

[0069] The waybill fulfillment data prediction model can be trained as follows: Obtain multiple specified waybill data samples of the second delivery origin-destination sample, historical waybill acceptance data samples of the second delivery origin-destination sample, spatiotemporal environmental characteristic data samples of the second delivery origin-destination sample, and waybill fulfillment label data of the capacity demand control area under the multiple specified waybill data samples, obtained based on these data samples; input the multiple specified waybill data samples, historical waybill acceptance data samples, and spatiotemporal environmental characteristic data samples into the initial logistics capacity model to obtain the waybill fulfillment prediction data of the capacity demand control area under the multiple specified waybill data samples output by the initial logistics capacity model; combine the waybill fulfillment prediction data and the waybill fulfillment data... The label data is input into a loss function used to evaluate the degree of predicted loss based on waybill fulfillment data. This yields the loss data between the predicted waybill fulfillment data and the waybill fulfillment label data, which is then used as the waybill fulfillment loss data. If the waybill fulfillment loss data is acceptable, the initial logistics capacity model is determined as an applicable logistics capacity model. Otherwise, the model parameters of the initial logistics capacity model are adjusted until the loss data between the predicted waybill fulfillment data and the waybill fulfillment label data output by the logistics capacity model obtained after adjusting the model parameters is acceptable. The logistics capacity model obtained after adjusting the model parameters is then determined as an applicable logistics capacity model.

[0070] The explanation of step S502 here is relatively simple. For specific implementation details, please refer to the relevant descriptions of the first and second embodiments, which will not be repeated here. It should be noted that the above steps S501 and S502 do not necessarily have a specific order.

[0071] Step S503 can be understood as follows: given the order distribution data and order benefit data at the first delivery origin-end point granularity over a future period, and the predicted order acceptance capacity at the second delivery origin-end point granularity to meet the order fulfillment requirements, reference data for capacity demand regulation at the first delivery origin-end point granularity can be determined over a future period. For example, given the order distribution data for each or a portion of the first delivery origin-end points within the capacity demand regulation area over a future period, the order benefit data for all or a portion of the first delivery origin-end points, and the predicted order acceptance capacity at all or a portion of the second delivery origin-end points to meet the order fulfillment requirements, characteristic data for capacity demand regulation to be implemented for each of all or a portion of the first delivery origin-end points over a future period can be determined. All starting locations in a portion of the second delivery origin-destination area (all starting AOIs in this AOI-AOI section) can include all starting locations in this portion of the first delivery origin-destination area (all starting SHOPs in this SHOP-AOI section), and all ending locations in this portion of the second delivery origin-destination area (all ending AOIs in this AOI-AOI section) include all ending locations in this portion of the first delivery origin-destination area (all ending AOIs in this SHOP-AOI section). It should be noted that the order distribution data mentioned here can refer to the production order forecast data in step S501, and the order benefit data can refer to the control benefit forecast data in step S501.

[0072] The regulation benefit prediction data may include average order benefit prediction data. One implementation of step S503 is as follows: Based on the production order prediction data and average order benefit prediction data of any one of the at least one first delivery origin and destination under the specified capacity demand regulation characteristic data in different capacity demand regulation characteristic data, obtain the order benefit prediction data of any one of the first delivery origin and destination under the specified capacity demand regulation characteristic data; after obtaining the order benefit prediction data of each of the at least one first delivery origin and destination under different capacity demand regulation characteristic data, obtain multiple total order benefit prediction data of all first delivery origin and destination; obtain the total order benefit prediction data that meets the regulation benefit target and matches the production order prediction data and the freight volume capacity prediction data of the first delivery origin and destination from the multiple total order benefit prediction data, and use it as the total order benefit target prediction data; obtain the capacity demand regulation characteristic data of each first delivery origin and destination corresponding to the total order benefit target prediction data, and use it as the capacity demand regulation reference data determined for at least one first delivery origin and destination; generate the capacity demand regulation reference data of the capacity demand regulation area based on the capacity demand regulation reference data determined for at least one first delivery origin and destination. For example, the order forecast data is the expected order volume, and the average order benefit forecast data is the revenue. By multiplying the expected order volume and the revenue, we can obtain the total revenue of the first delivery origin and destination under the capacity demand control characteristic data, which is to say, obtain the order benefit forecast data of the first delivery origin and destination under the specified capacity demand control characteristic data.

[0073] After obtaining order benefit prediction data for each first delivery origin and destination under different capacity demand control characteristic data, a capacity demand control characteristic data can be selected for each first delivery origin and destination from the different capacity demand control characteristic data, serving as the selected capacity demand control characteristic data for each first delivery origin and destination in one round. The sum of the order benefit prediction data for each first delivery origin and destination under the selected demand control characteristic data in this round is obtained as a total order benefit prediction data for all first delivery origins and destinations. Then, another capacity demand control characteristic data can be selected for each first delivery origin and destination from the different capacity demand control characteristic data, serving as the selected capacity demand control characteristic data for each first delivery origin and destination in another round. The sum of the order benefit prediction data for each first delivery origin and destination under the selected demand control characteristic data in this second round is obtained as another total order benefit prediction data for all first delivery origins and destinations. This process is repeated to obtain multiple total order benefit prediction data for all first delivery origins and destinations. Specifically, a capacity demand control characteristic data can be selected for each first delivery origin and destination as the first round of selected capacity demand control characteristic data for each first delivery origin and destination. Obtain the order benefit prediction data for each first delivery origin and destination under the selected capacity demand control feature data in the first round. Based on this data, sum the order benefit prediction data for each first delivery origin and destination under the selected capacity demand control feature data in the first round. This sum is used as the total order benefit prediction data for all first delivery origins and destinations, and can be called the first round total order benefit prediction data. For example, the order benefit prediction data for each first delivery origin and destination under the selected capacity demand control feature data in the first round can be added together to obtain the sum of the first round total order benefit prediction data. After selecting the first round of selected capacity demand control feature data for each first delivery origin and destination, select another selected capacity demand control feature data for each first delivery origin and destination as the second round of selected capacity demand control feature data for each first delivery origin and destination. Obtain the order benefit prediction data for each first delivery origin and destination under the second round of selected capacity demand control feature data. Based on the order benefit prediction data of each first delivery origin and destination under the selected capacity demand control feature data in the second round, the sum of the order benefit prediction data of each first delivery origin and destination under the selected demand control feature data in the second round is obtained, which serves as another total order benefit prediction data for all first delivery origins and destinations. This total order benefit prediction data can be called the second round total order benefit prediction data.For example, the order benefit forecast data for each first delivery origin and destination under the selected demand control feature data in the second round can be summed to obtain the sum of the order benefit forecast data for each first delivery origin and destination under the selected demand control feature data in the second round, which is the total order benefit forecast data for the second round. Similarly, the sum of the order benefit forecast data for each first delivery origin and destination under the selected demand control feature data in other rounds can be obtained, which is the total order benefit forecast data for other rounds. Obtaining the total order benefit forecast data for multiple rounds is equivalent to obtaining multiple total order benefit forecast data for all first delivery origins and destinations.

[0074] After obtaining multiple total order benefit forecast data, the total order benefit forecast data that meets the control benefit target and matches the order volume forecast data and waybill capacity forecast data at the first delivery origin and destination is selected as the target total order benefit forecast data. For example, the total order benefit forecast data with the largest total order benefit forecast data that matches the order volume forecast data and waybill capacity forecast data at the first delivery origin and destination is selected as the target total order benefit forecast data. The matching of the order volume forecast data and waybill capacity forecast data at the first delivery origin and destination indicates that the expected order volume at the first delivery origin and destination can be transported by logistics based on the logistics' expected capacity. In addition, the largest total order benefit forecast data is based on the premise that the order volume forecast data and waybill capacity forecast data at the first delivery origin and destination match. In other words, the total order benefit forecast data with the largest total order benefit forecast data can be selected from the total order benefit forecast data that meets the condition of matching the order volume forecast data and waybill capacity forecast data at the first delivery origin and destination.

[0075] Matching the order forecast data and the waybill capacity forecast data of the first delivery origin and destination can include at least one of the following matching methods. The first matching method involves matching the order forecast data of all first delivery origins and destinations whose distance between the origin and destination is no greater than a preset origin-endpoint distance threshold with the waybill capacity forecast data of all second delivery origins and destinations whose distance between the origin and destination is no greater than the preset origin-endpoint distance threshold. For example, if the preset origin-endpoint distance threshold is set to 3 kilometers, the first matching method can be understood as the difference between the sum of the order forecast data of all first delivery origins and destinations whose distance between the origin and destination is no more than 3 kilometers and the sum of the waybill capacity forecast data of all second delivery origins and destinations whose distance between the origin and destination is no more than a preset first threshold for forecast data difference. The second matching method involves matching the order forecast data of all first delivery origins and destinations whose distance between the origin and destination is greater than the preset origin-endpoint distance threshold with the waybill capacity forecast data of all second delivery origins and destinations whose distance between the origin and destination is greater than the preset origin-endpoint distance threshold. Similarly, if the preset origin-destination distance threshold is set to 3 kilometers, the second matching method can be understood as follows: the difference between the sum of the order prediction data for all first delivery origin-destination points with a distance exceeding 3 kilometers and the sum of the order capacity prediction data for all second delivery origin-destination points with a distance exceeding 3 kilometers does not exceed the preset second threshold for prediction data difference. The third matching method matches the order prediction data for all first delivery origin-destination points with the order capacity prediction data for all second delivery origin-destination points.

[0076] After obtaining the total order benefit target prediction data, the capacity demand control characteristic data for each first delivery origin and destination corresponding to the total order benefit target prediction data can be obtained, serving as reference data for capacity demand control determined for at least one first delivery origin and destination. As mentioned earlier, after selecting capacity demand control characteristic data for each first delivery origin and destination in each round, a total order benefit prediction data is obtained, which is the total order benefit prediction data for that round. Since the total order benefit target prediction data is determined from multiple total order benefit prediction data, after determining the total order benefit target prediction data, it is possible to reverse-engineer the capacity demand control characteristic data required to be configured for each first delivery origin and destination, that is, to determine the capacity demand control reference data for each first delivery origin and destination corresponding to the total order benefit target prediction data. After obtaining the capacity demand control reference data determined for at least one first delivery origin and destination, capacity demand control reference data for the capacity demand control area can be generated based on the capacity demand control reference data determined for at least one first delivery origin and destination. The capacity demand control reference data may include capacity demand control reference data determined for at least one first delivery origin and destination. In other words, the capacity demand control reference data may be expressed as capacity demand control reference data determined for each of the at least one first delivery origin and destination.

[0077] For further explanation of step S503, please refer to the descriptions of the first and fourth embodiments, which will not be repeated here.

[0078] As an optional implementation, to minimize the possibility that the actual number of waybills at the second delivery origin and destination exceeds the logistics side's carrying capacity after the implementation of the capacity demand control reference data, after obtaining the capacity demand control reference data, it can be determined whether the logistics side has the capacity to handle the expected number of waybills generated after the implementation of the capacity demand control reference data. If so, it indicates that the capacity demand control reference data is appropriate or acceptable, and thus it can be implemented. Otherwise, it indicates that the expected number of waybills generated after the implementation of the capacity demand control reference data exceeds the logistics side's carrying capacity, so the capacity demand control reference data needs to be adjusted until the expected number of waybills generated after the implementation of the adjusted capacity demand control reference data can be handled by the logistics side. Afterward, the adjusted capacity demand control reference data can be implemented. In practical applications, it can be determined whether the forecast data for the second delivery origin and destination corresponding to the capacity demand control reference data meets the waybill fulfillment requirements. If so, it is determined that the capacity demand control reference data matches the waybill fulfillment requirements; otherwise, the capacity demand control reference data for the capacity demand control area is obtained again until the obtained capacity demand control reference data matches the waybill fulfillment requirements. Afterward, the capacity demand control reference data that matches the waybill fulfillment requirements can be executed.

[0079] The waybill prediction data of the second delivery origin and destination corresponding to the capacity demand control reference data can be obtained in the following way: obtain the order prediction data of any one of the first delivery origins and destinations under the capacity demand control characteristic data determined for any one of the first delivery origins and destinations included in the capacity demand control reference data, and use it as the order prediction data of any one of the first delivery origins and destinations under the capacity demand control reference data. The waybill forecast data for the second delivery origin and destination under the capacity demand control reference data can be obtained as follows: Based on the pre-set attribution relationship between the first and second delivery origin and destinations, determine the second delivery origin and destination to which each of the at least one first delivery origin and destination belongs. For any given second delivery origin and destination, obtain the waybill forecast data for that given second delivery origin and destination based on the waybill forecast data of at least one first delivery origin and destination belonging to that given second delivery origin and destination. For example, the waybill forecast data of at least one first delivery origin and destination belonging to that given second delivery origin and destination can be added together, and the sum of the obtained waybill forecast data of at least one first delivery origin and destination belonging to that given second delivery origin and destination can be determined as the waybill forecast data for that given second delivery origin and destination. The attribution relationship between the first and second delivery origin and destinations can indicate the first delivery origin and destinations included in the second delivery origin and destination.

[0080] After obtaining the waybill forecast data for the second delivery origin and destination corresponding to the capacity demand control reference data, it can be determined whether the waybill forecast data for the second delivery origin and destination corresponding to the capacity demand control reference data meets the waybill fulfillment requirements. This includes: obtaining historical waybill acceptance data for the second delivery origin and destination; obtaining spatiotemporal environmental characteristic data of the second delivery origin and destination; obtaining waybill fulfillment forecast data for the second delivery origin and destination under the capacity demand control reference data based on the waybill forecast data for the second delivery origin and destination corresponding to the capacity demand control reference data, the historical waybill acceptance data for the second delivery origin and destination, and the spatiotemporal environmental characteristic data of the second delivery origin and destination; obtaining waybill fulfillment requirement data; and determining whether the waybill forecast data for the second delivery origin and destination corresponding to the capacity demand control reference data meets the waybill fulfillment requirements based on the waybill fulfillment forecast data and the waybill fulfillment requirement data of the second delivery origin and destination under the capacity demand control reference data. For instructions on how to determine whether the waybill forecast data for the second delivery origin and destination corresponding to the capacity demand control reference data meets the waybill fulfillment requirements, please refer to the relevant description in the second embodiment. After obtaining the reference data for capacity demand regulation, how to determine whether the waybill prediction data of the second delivery origin and destination corresponding to the reference data for capacity demand regulation meets the waybill fulfillment requirements, and how to determine the subsequent processing method based on the judgment result, can be referred to the relevant description of system 105 in the first embodiment and the relevant description of the second embodiment, which will not be repeated here.

[0081] Obtaining reference data for capacity demand regulation can be subject to certain triggering conditions, or triggering opportunities. When these conditions are met, the reference data can be obtained. For example, it can be determined whether the forecast data for waybills at the second delivery origin and destination within a specified time period meets the waybill fulfillment requirements. Step S503 may include: if it is determined that the forecast data for waybills at the second delivery origin and destination within the specified time period does not meet the waybill fulfillment requirements, then based on the waybill capacity forecast data, production forecast data, and regulation benefit forecast data, obtaining reference data for capacity demand regulation in the capacity demand regulation area. Determining whether the waybill forecast data for the second delivery origin and destination within a specified time period meets the waybill fulfillment requirements includes: obtaining the waybill forecast data for the second delivery origin and destination within the specified time period; obtaining historical waybill acceptance data for the second delivery origin and destination; obtaining spatiotemporal environmental characteristic data of the second delivery origin and destination within the specified time period; obtaining waybill fulfillment forecast data for the second delivery origin and destination within the specified time period based on the waybill forecast data, historical waybill acceptance data, and spatiotemporal environmental characteristic data; obtaining waybill fulfillment requirement data; and determining whether the waybill forecast data for the second delivery origin and destination within the specified time period meets the waybill fulfillment requirements based on the waybill fulfillment forecast data and the waybill fulfillment requirement data. For explanations on how to determine the triggering conditions for obtaining capacity demand control reference data and how to obtain capacity demand control reference data based on the triggering conditions, please refer to the relevant descriptions of the logistics delivery time assessment service subsystem 106 in the first embodiment and the relevant descriptions in the second embodiment; these will not be repeated here.

[0082] Sixth Embodiment This application also provides an embodiment of a device for obtaining reference data for capacity demand regulation. For example... Figure 6As shown, the capacity demand regulation reference data acquisition device provided in the sixth embodiment of this application includes: a first delivery origin-end prediction data acquisition unit 601, used to acquire order prediction data and regulation benefit prediction data of the first delivery origin-end in the capacity demand regulation area under different capacity demand regulation characteristic data, including: acquiring long-cycle characteristic data of the first delivery origin-end, the long-cycle characteristic data of the first delivery origin-end including long-cycle supply data of the starting position of the first delivery origin-end, long-cycle supply data sequence of order entities around the starting position of the first delivery origin-end, and long-cycle supply data sequence of order entities around the ending position of the first delivery origin-end; based on the long-cycle supply data of the starting position of the first delivery origin, the long-cycle supply data sequence of order entities around the starting position of the first delivery origin, and the long-cycle supply data sequence of order entities around the ending position of the first delivery origin, acquiring short-cycle order operation prediction data of the first delivery origin-end, the duration of the short cycle in the short-cycle order operation prediction data is less than the long cycle of the long-cycle characteristic data. The system obtains the following data: The duration of the first delivery origin and destination; real-time characteristic data of the first delivery origin and destination, and multiple capacity demand control characteristic data for the first delivery origin and destination; based on the short-cycle order operation forecast data of the first delivery origin and destination, the real-time characteristic data of the first delivery origin and destination, and the multiple capacity demand control characteristic data for the first delivery origin and destination, the system obtains the order production forecast data and control benefit forecast data of the first delivery origin and destination; the second delivery origin and destination forecast data acquisition unit 602 is used to obtain the order acceptance capacity forecast data of the second delivery origin and destination in the capacity demand control area that meets the order fulfillment requirements, wherein the starting position of the second delivery origin and destination includes the starting position of the first delivery origin and destination, and the ending position of the second delivery origin and destination includes the ending position of the first delivery origin and destination; the capacity demand control reference data acquisition unit 603 is used to obtain the capacity demand control reference data of the capacity demand control area based on the order acceptance capacity forecast data, order production forecast data, and control benefit forecast data, wherein the capacity demand control reference data includes capacity demand control reference data determined for at least one first delivery origin and destination. It should be noted that the relevant descriptions of the sixth embodiment can be referred to the relevant descriptions of the fifth embodiment above, and will not be repeated here.

[0083] Seventh Embodiment This application also provides an embodiment of an electronic device. For example... Figure 7 As shown, the electronic device includes: at least one processor 701, at least one communication interface 702, at least one memory 703, and at least one communication bus 704; the memory 703 stores a computer program, and the processor 701 calls the program stored in the memory 703 to execute the method provided in the fifth embodiment of this application and the steps that need to be executed by a computer program mentioned in the first to fourth embodiments of this application.

[0084] Eighth embodiment The eighth embodiment of this application also provides a computer storage device. This computer storage device stores a computer program that is executed by a processor to perform the method provided in the fifth embodiment of this application and the steps mentioned in the first to fourth embodiments of this application that require execution by a computer program.

[0085] It should be noted that the embodiments of this application may involve the use of user data. In practical applications, user-specific personal data may be used within the scope permitted by applicable laws and regulations of the country in which the application is located (e.g., with the user's explicit consent and effective notification to the user). In practical applications, corresponding access points may be provided for users to choose to authorize or refuse the use of their personal data.

Claims

1. A method for obtaining reference data for capacity demand regulation, characterized in that, include: Obtaining order prediction data and regulation benefit prediction data for the first delivery origin and destination within a capacity demand regulation area under different capacity demand regulation characteristic data includes: obtaining long-term characteristic data of the first delivery origin and destination, which includes long-term supply data at the starting position of the first delivery origin and destination, long-term supply data sequences of order entities surrounding the starting position of the first delivery origin and destination, and long-term supply data sequences of order entities surrounding the destination position of the first delivery origin and destination; based on the long-term supply data at the starting position of the first delivery origin and destination, the long-term supply data sequences of order entities surrounding the starting position of the first delivery origin and destination, and... The system obtains short-cycle order operation forecast data for the first delivery origin and destination from long-cycle supply data sequences of order entities surrounding the destination location. The duration of the short-cycle in the short-cycle order operation forecast data is shorter than the duration of the long-cycle in the long-cycle feature data. It also obtains real-time feature data for the first delivery origin and destination and multiple capacity demand control feature data for the first delivery origin and destination. Based on the short-cycle order operation forecast data, the real-time feature data, and the multiple capacity demand control feature data for the first delivery origin and destination, the system obtains order forecast data and control benefit forecast data for the first delivery origin and destination. Obtain the predicted data of the waybill acceptance capacity of the second delivery origin and destination in the capacity demand control area to meet the waybill fulfillment requirements. The starting position of the second delivery origin and destination includes the starting position of the first delivery origin and destination, and the ending position of the second delivery origin and destination includes the ending position of the first delivery origin and destination. Based on the order acceptance capacity prediction data, the production order prediction data, and the regulation benefit prediction data, the transportation capacity demand regulation reference data for the transportation capacity demand regulation area is obtained, wherein the transportation capacity demand regulation reference data includes transportation capacity demand regulation reference data determined for at least one first delivery origin and destination.

2. The method for obtaining reference data for capacity demand regulation according to claim 1, characterized in that, The step of obtaining order prediction data and regulation benefit prediction data for the first delivery origin and destination based on short-cycle order operation prediction data, real-time characteristic data of the first delivery origin and destination, and multiple capacity demand regulation characteristic data for the first delivery origin and destination includes: The short-cycle order operation prediction data of the first delivery origin and destination and the real-time feature data of the first delivery origin and destination are input into the feature cross-fusion model to obtain the real-time feature operation fusion vector of the first delivery origin and destination output by the feature cross-fusion model. Multiple capacity demand regulation feature data for the first delivery origin and destination and the real-time feature operation fusion vector of the first delivery origin and destination are input into the capacity demand regulation layer model to obtain the order prediction data and regulation benefit prediction data of the first delivery origin and destination in the future period output by the capacity demand regulation layer model.

3. The method for obtaining reference data for capacity demand regulation according to claim 2, characterized in that, The step of inputting the short-cycle order operation prediction data and the real-time feature data of the first delivery origin and destination into the feature cross-fusion model to obtain the real-time feature operation fusion vector of the first delivery origin and destination output by the feature cross-fusion model includes: obtaining the real-time feature vector corresponding to the real-time feature data of the first delivery origin and destination, obtaining the short-cycle order operation prediction vector corresponding to the short-cycle order operation prediction data of the first delivery origin and destination, and performing feature cross-fusion processing on the real-time feature vector and the short-cycle order operation prediction vector to obtain the real-time feature operation fusion vector of the first delivery origin and destination; The step of inputting multiple capacity demand regulation feature data and the real-time characteristic operation fusion vector of the first delivery origin and destination into the capacity demand regulation layer model to obtain the order prediction data and regulation benefit prediction data of the first delivery origin and destination for a future period output by the capacity demand regulation layer model includes: obtaining multiple capacity demand regulation feature vectors corresponding to multiple capacity demand regulation feature data of the first delivery origin and destination; using the multiple capacity demand regulation feature vectors as regulation terms to perform vector regulation processing on the real-time characteristic operation fusion vector of the first delivery origin and destination to obtain the order prediction data and regulation benefit prediction data of the first delivery origin and destination for a future period.

4. The method for obtaining reference data for capacity demand regulation according to claim 2, characterized in that, The feature cross-fusion model and the capacity demand regulation layer model are trained in the following manner: The system obtains multiple capacity demand regulation feature data samples for the first delivery origin-end sample, real-time feature data samples for the first delivery origin-end sample, short-cycle order operation prediction data samples for the first delivery origin-end sample, and production order label data and regulation benefit label data of the first delivery origin-end sample expected to be obtained in the future based on the multiple capacity demand regulation feature data samples, the real-time feature data samples and the short-cycle order operation prediction data samples. The real-time feature data sample and the short-cycle order operation prediction data sample are input into the initial feature cross-fusion model to obtain the real-time feature operation fusion vector sample of the first delivery origin and destination sample output by the initial feature cross-fusion model. The real-time feature operation fusion vector sample and the multiple capacity demand regulation feature data samples are input into the initial capacity demand regulation layer model to obtain the order prediction data and regulation benefit prediction data of the first delivery origin and destination sample in the future period of the output of the initial capacity demand regulation layer model. The production order prediction data and regulation benefit prediction data of the first delivery origin and destination sample in the future period, as well as the production order label data and the regulation benefit label data, are input into a loss function used to evaluate the degree of production order prediction loss and regulation benefit prediction loss, so as to obtain the production order prediction and regulation benefit prediction loss data between the production order prediction data and regulation benefit prediction data of the first delivery origin and destination sample in the future period and the production order label data and the regulation benefit label data. If the loss data of the predicted production order and the predicted regulation benefit are acceptable loss data, then the initial feature cross-fusion model is determined as an applicable feature cross-fusion model, and the initial capacity demand regulation layer model is determined as an applicable capacity demand regulation layer model. Otherwise, the model parameters of the initial feature cross-fusion model and the initial capacity demand regulation layer model are adjusted until the loss data of the predicted production order and the predicted regulation benefit of the first delivery origin-end sample obtained from the feature cross-fusion model and the capacity demand regulation layer model after adjusting the model parameters are acceptable loss data. Then, the feature cross-fusion model after adjusting the model parameters is determined as an applicable feature cross-fusion model, and the capacity demand regulation layer model after adjusting the model parameters is determined as an applicable capacity demand regulation layer model.

5. The method for obtaining reference data for capacity demand regulation according to claim 1, characterized in that, The long-term feature data of the first delivery origin and destination also includes long-term auxiliary feature data of the first delivery origin and destination; The step of obtaining short-cycle order operation forecast data for the first delivery origin and destination based on long-cycle supply data of the starting position of the first delivery origin and destination, long-cycle supply data sequences of production order entities surrounding the starting position of the first delivery origin and destination, and long-cycle supply data sequences of production order entities surrounding the destination of the first delivery origin and destination includes: The long-term supply data of the starting position of the first delivery origin and destination and the long-term supply data sequence of the production order entities around the starting position of the first delivery origin and destination are input into the first multi-head attention model to obtain the starting point and the supply cross vector around the starting point output by the first multi-head attention model. The long-term supply data of the starting position of the first delivery origin and destination and the long-term supply data sequence of the production order entities around the ending position of the first delivery origin and destination are input into the second multi-head attention model to obtain the supply cross vector of the starting point and the ending point output by the second multi-head attention model. The long-term auxiliary feature data of the first delivery origin and destination, the supply cross vector between the origin and the surrounding area of ​​the origin, and the supply cross vector between the origin and the destination are input into the vector splicing model to obtain the long-term splicing vector of the first delivery origin and destination. The long-period spliced ​​vector of the first delivery origin and destination is input into the vector high-order cross model to obtain the high-order cross vector of the first delivery origin and destination output by the vector high-order cross model. The high-order cross vector of the first delivery origin and destination is input into the parameter generation network model to obtain the short-cycle operation prediction parameter data output by the parameter generation network model for predicting the short-cycle order operation prediction data of the first delivery origin and destination. Based on the high-order cross vector of the first delivery origin and destination and the short-cycle operation prediction parameter data, the short-cycle order operation prediction data of the first delivery origin and destination is obtained.

6. The method for obtaining reference data for capacity demand regulation according to claim 5, characterized in that, The step of inputting the long-cycle supply data of the starting position of the first delivery origin and destination and the long-cycle supply data sequence of the production order entities around the starting position of the first delivery origin and destination into the first multi-head attention model to obtain the starting point and surrounding supply cross vector output by the first multi-head attention model includes: obtaining the long-cycle supply vector of the starting position corresponding to the long-cycle supply data of the starting position of the first delivery origin and destination; obtaining the long-cycle supply vector of the production order entities around the starting position corresponding to the long-cycle supply data sequence of the production order entities around the starting position of the first delivery origin and destination; and performing multi-head attention cross processing on the long-cycle supply vector of the starting position and the long-cycle supply vector of the production order entities around the starting position to obtain the starting point and surrounding supply cross vector. The step of inputting the long-cycle supply data of the starting position of the first delivery origin and destination and the long-cycle supply data sequence of the production order entities around the ending position of the first delivery origin and destination into the second multi-head attention model to obtain the starting point and ending point periphery supply cross vector output by the second multi-head attention model includes: obtaining the long-cycle supply vector of the starting position corresponding to the long-cycle supply data of the starting position of the first delivery origin and destination, obtaining the long-cycle supply vector of the production order entities around the ending position corresponding to the long-cycle supply data sequence of the production order entities around the ending position of the first delivery origin and destination, and performing multi-head attention cross processing on the long-cycle supply vector of the starting position and the long-cycle supply vector of the production order entities around the ending position to obtain the starting point and ending point periphery supply cross vector; The step of inputting the long-cycle auxiliary feature data of the first delivery origin and destination, the cross vector of the supply around the origin and the supply around the origin and the destination into the vector splicing model to obtain the long-cycle spliced ​​vector of the first delivery origin and destination includes: obtaining the long-cycle auxiliary feature vector corresponding to the long-cycle auxiliary feature data of the first delivery origin and destination, and splicing the cross vector of the supply around the origin and the destination, the cross vector of the supply around the origin and the destination, and the long-cycle auxiliary feature vector to obtain the long-cycle spliced ​​vector of the first delivery origin and destination.

7. The method for obtaining reference data for capacity demand regulation according to claim 5, characterized in that, The first multi-head attention model, the second multi-head attention model, the vector concatenation model, the higher-order vector cross model, and the parameter generation network model are trained in the following manner: The system obtains long-cycle auxiliary feature data samples of the first delivery origin-end point sample, long-cycle supply data samples of the starting position of the first delivery origin-end point sample, long-cycle supply data samples of order entities around the starting position of the first delivery origin-end point sample, long-cycle supply data samples of order entities around the ending position of the first delivery origin-end point sample, and short-cycle order operation tag data of the first delivery origin-end point sample expected to be obtained based on the long-cycle auxiliary feature data samples, long-cycle supply data samples of the starting position, long-cycle supply data samples of order entities around the starting position, and long-cycle supply data samples of order entities around the ending position. The long-term supply data sample at the starting position and the long-term supply data sample of the production order entity around the starting position are input into the initial first multi-head attention model to obtain the starting point and the supply cross vector around the starting point output by the initial first multi-head attention model. The long-term supply data sample at the starting position and the long-term supply data sample of the production order entity around the ending position are input into the initial second multi-head attention model to obtain the supply cross vector around the starting and ending positions output by the initial second multi-head attention model. The long-period auxiliary feature data sample of the first delivery origin and destination sample, the starting point and the supply cross vector around the starting point output by the initial first multi-head attention model, and the starting point and the supply cross vector around the starting point output by the initial second multi-head attention model are input into the initial vector splicing model to obtain the long-period splicing vector of the first delivery origin and destination sample. The long-period spliced ​​vector of the first delivery origin and destination sample is input into the high-order cross-vector model of the initial vector to obtain the high-order cross-vector of the first delivery origin and destination sample output by the high-order cross-vector model of the initial vector model. The high-order cross vector of the first delivery origin-end point sample is input into the initial parameter generation network model to obtain the short-cycle order operation prediction parameter data of the first delivery origin-end point sample output by the initial parameter generation network model for predicting the short-cycle order operation prediction data of the first delivery origin-end point sample. Based on the high-order cross vector of the first delivery origin-end point sample and the short-cycle operation prediction parameter data of the first delivery origin-end point sample, the short-cycle order operation prediction data of the first delivery origin-end point sample is obtained. The short-cycle order operation forecast data of the first delivery origin and destination sample and the short-cycle order operation tag data of the first delivery origin and destination sample are input into a loss function used to evaluate the degree of short-cycle order operation forecast loss, so as to obtain the short-cycle order operation forecast loss data between the short-cycle order operation forecast data of the first delivery origin and destination sample and the short-cycle order operation tag data of the first delivery origin and destination sample. If the predicted loss data for short-cycle order operations is acceptable, then the initial first multi-head attention model, the initial second multi-head attention model, the initial vector concatenation model, the initial higher-order vector crossover model, and the initial parameter generation network model are respectively determined as the applicable first multi-head attention model, the applicable second multi-head attention model, the applicable vector concatenation model, the applicable higher-order vector crossover model, and the applicable parameter generation network model. Otherwise, the model parameters of the initial first multi-head attention model, the initial second multi-head attention model, the initial vector concatenation model, the initial higher-order vector crossover model, and the initial parameter generation network model are adjusted until the model parameters are adjusted. The short-cycle order operation prediction loss data between the short-cycle operation prediction parameter data of the first delivery origin-end point sample obtained by the first multi-head attention model, the second multi-head attention model, the vector concatenation model, the vector higher-order cross model, and the parameter generation network model after parameter adjustment, and the short-cycle order operation label data of the first delivery origin-end point sample, is determined to be acceptable loss data until the first multi-head attention model, the second multi-head attention model, the vector concatenation model, the vector higher-order cross model, and the parameter generation network model after parameter adjustment are respectively determined as the applicable first multi-head attention model, the applicable second multi-head attention model, the applicable vector concatenation model, the applicable vector higher-order cross model, and the applicable parameter generation network model.

8. The method for obtaining reference data for capacity demand regulation according to claim 1, characterized in that, The method of obtaining the predicted waybill capacity for the second delivery origin and destination within the capacity demand control area, which meets the waybill fulfillment requirements, includes: Obtain multiple designated waybill data for the second delivery origin and destination; Obtain historical waybill acceptance data for the second delivery origin and destination; Obtain the spatiotemporal environmental characteristics data of the second delivery origin and destination; Based on the multiple designated waybill data of the second delivery origin and destination, the historical waybill acceptance data of the second delivery origin and destination, and the spatiotemporal environmental characteristic data of the second delivery origin and destination, the waybill fulfillment prediction data of the capacity demand control area under the multiple designated waybill data is obtained. Obtain data on waybill fulfillment requirements; Obtain the specified waybill data corresponding to the waybill fulfillment prediction data that matches the waybill fulfillment requirement data, and use it as the specified matching waybill data; The predicted load factor of the waybill is obtained based on the specified matching waybill data.

9. The method for obtaining reference data for capacity demand regulation according to claim 1, characterized in that, The predicted data for the regulation benefits include predicted data for the average benefit per unit. The step of obtaining reference data for transport capacity demand regulation in the transport capacity demand regulation area based on the transport order acceptance capacity forecast data, the production order forecast data, and the regulation benefit forecast data includes: Based on the order prediction data and the average order benefit prediction data of any one of the at least one first delivery origin and destination under the specified capacity demand control characteristic data in the different capacity demand control characteristic data, the order benefit prediction data of any one of the first delivery origin and destination under the specified capacity demand control characteristic data is obtained. After obtaining the order benefit prediction data for each of the at least one first delivery origin and destination under the different capacity demand control characteristic data, multiple total order benefit prediction data for all first delivery origin and destination are obtained. From the total order benefit prediction data, obtain the total order benefit prediction data that meets the control benefit target and matches the order prediction data with the order acceptance capacity prediction data at the first delivery origin and destination, and use it as the total order benefit target prediction data. Obtain the capacity demand control characteristic data for each first delivery origin and destination corresponding to the total order benefit target prediction data, and use it as reference data for capacity demand control determined for at least one first delivery origin and destination; Based on the capacity demand control reference data determined for at least one first delivery origin and destination, capacity demand control reference data for the capacity demand control area is generated.

10. The method for obtaining reference data for capacity demand regulation according to claim 9, characterized in that, The acquisition of total order benefit prediction data for all first delivery origin and destination points includes: From the different capacity demand control feature data, select one capacity demand control feature data for each first delivery origin and destination, as a round of selected capacity demand control feature data for each first delivery origin and destination; The sum of the order benefit prediction data for each first delivery origin and destination under the selected capacity demand control feature data in the first round is obtained as a total order benefit prediction data for all first delivery origins and destinations; From the different capacity demand control feature data, select one capacity demand control feature data for each first delivery origin and destination, as another round of selected capacity demand control feature data for each first delivery origin and destination; The sum of the order benefit prediction data for each first delivery origin and destination under the selected capacity demand control feature data of the other round is obtained as another total order benefit prediction data for all first delivery origins and destinations; By doing so, we can obtain total order benefit forecasts for all first delivery origins and destinations.

11. The method for obtaining reference data for capacity demand regulation according to claim 9, characterized in that, Matching the order forecast data at the first delivery origin and destination with the order capacity forecast data includes at least one of the following matching methods: The order prediction data of all first delivery origin and destination locations whose distance between the origin and destination is not greater than the preset origin-end point distance threshold are matched with the order acceptance capacity prediction data of all second delivery origin and destination locations whose distance between the origin and destination is not greater than the preset origin-end point distance threshold. The order prediction data of all first delivery origins and destinations whose distance between the origin and destination is greater than the preset origin-end point distance threshold are matched with the order acceptance capacity prediction data of all second delivery origins and destinations whose distance between the origin and destination is greater than the preset origin-end point distance threshold. The order forecast data for all first delivery origin and destination points are matched with the order capacity forecast data for all second delivery origin and destination points.

12. The method for obtaining reference data for capacity demand regulation according to claim 1, characterized in that, Also includes: Determine whether the waybill prediction data at the second delivery origin and destination corresponding to the capacity demand control reference data meets the waybill fulfillment requirements; If so, then it is determined that the reference data for adjusting the transport capacity demand matches the waybill fulfillment requirements; Otherwise, the capacity demand control reference data for the aforementioned capacity demand control area is obtained again until the obtained capacity demand control reference data matches the waybill fulfillment requirements.

13. The method for obtaining reference data for capacity demand regulation according to claim 1, characterized in that, Also includes: Determine whether the waybill prediction data for the second delivery origin and destination within a specified time period meets the waybill fulfillment requirements; The step of obtaining reference data for capacity demand regulation in the capacity demand regulation area based on the waybill acceptance capacity forecast data, the production order forecast data, and the regulation benefit forecast data includes: if it is determined that the waybill forecast data at the second delivery origin and destination in a specified time period does not meet the waybill fulfillment requirements, then obtaining reference data for capacity demand regulation in the capacity demand regulation area based on the waybill acceptance capacity forecast data, the production order forecast data, and the regulation benefit forecast data.

14. An electronic device, comprising: processor; The electronic device also includes a memory for storing a computer program, which, when powered on and running through the processor, executes the method according to any one of claims 1-13.

15. A storage device storing a computer program, the computer program being executed by a processor to perform the method of any one of claims 1-13.

Images