WO2020108465A1 - 订单分配方法、装置、电子设备和存储介质 - Google Patents

订单分配方法、装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2020108465A1
WO2020108465A1 PCT/CN2019/120838 CN2019120838W WO2020108465A1 WO 2020108465 A1 WO2020108465 A1 WO 2020108465A1 CN 2019120838 W CN2019120838 W CN 2019120838W WO 2020108465 A1 WO2020108465 A1 WO 2020108465A1
Authority
WO
WIPO (PCT)
Prior art keywords
order
wave
timeout
distribution
probability
Prior art date
Application number
PCT/CN2019/120838
Other languages
English (en)
French (fr)
Inventor
徐小丽
叶畅
Original Assignee
拉扎斯网络科技（上海）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 拉扎斯网络科技（上海）有限公司 filed Critical 拉扎斯网络科技（上海）有限公司
Publication of WO2020108465A1 publication Critical patent/WO2020108465A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • G06Q10/063112—Skill-based matching of a person or a group to a task
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/083—Shipping
    • G06Q10/0838—Historical data

Definitions

  • the communication technology field of the embodiments of the present application particularly relates to an order distribution method, device, electronic equipment, and storage medium.
  • the judgment on whether the rider delivery order may time out is based on the following two schemes.
  • the second is to perform a path planning simulation on the existing back orders of the rider, that is, to simulate the rider's order of sending orders to determine whether the rider has a risk of overtime, so as to realize the rider's order dispatching.
  • a path planning simulation on the existing back orders of the rider, that is, to simulate the rider's order of sending orders to determine whether the rider has a risk of overtime, so as to realize the rider's order dispatching.
  • the order and path planning are not completely accurate, and there will be errors.
  • the larger the order quantity the overtime estimated deviation for each order delivered will always be passed back and accumulated.
  • the simulation of path planning and delivery order may not necessarily coincide with what actually happened. Based on these two points, it is impossible to accurately estimate the rider's timeout judgment.
  • the purpose of the embodiments of the present application is to provide an order distribution method, device, electronic equipment, and storage medium.
  • a wave concept based on the rider from no load (not a single order) to no load (all orders are delivered)
  • a wave of orders is used to evaluate the timeout probability of the assigned order, which improves the accuracy of the judgment of whether the order delivery is overtime, thereby improving the effectiveness of order allocation and improving the user experience.
  • the embodiment of the present application provides an order allocation method, which includes: obtaining the order parameter set of the distribution resource at the target wave, and the wave is the distribution resource from one no-load state to the next no-load state During the period; according to the order parameter set of the distribution resource at the target wave, evaluate the timeout probability of the distribution resource at the target wave; allocate the order to the distribution resource according to the timeout probability of the distribution resource at the target wave.
  • An embodiment of the present application also provides an order allocation device, including: a parameter acquisition module, which acquires an order parameter set of a distribution resource at a target wave, and the wave is a period from a no-load state of a distribution resource to a next no-load state ; Probability evaluation module, based on the order parameter set of the distribution resource at the target wave, evaluates the timeout probability of the distribution resource at the target wave; Order allocation module, orders the distribution resource according to the probability of the timeout of the distribution resource at the target wave.
  • An embodiment of the present application also provides an electronic device including at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are at least one
  • the processor executes to achieve: obtain the order parameter set of the delivery resource in the target wave, and the wave is the period from one empty state to the next empty state of the delivery resource; evaluate the delivery based on the order parameter set of the delivery resource in the target wave Probability of resource overtime at target wave; order distribution of distribution resources based on the probability of overtime of distribution resource at target wave.
  • An embodiment of the present application further provides a non-volatile storage medium for storing a computer-readable program, the computer-readable program is for a computer to execute the above-mentioned order distribution method.
  • the distribution time of the latest order in a wave is earlier than the meal-taking time of the earliest order in a wave.
  • the above method also includes: acquiring the historical order parameter set of the historical wave of the distribution resource; constructing the wave timeout model of the distribution resource, the wave timeout model includes the wave timeout function, and the wave timeout function uses the order parameter set of the distribution resource
  • the parameters in are variables, and each variable is assigned a corresponding weight value; using a logistic regression algorithm, the wave timeout model is trained according to the historical order parameter set of the distribution resource.
  • the timeout probability of the order to be allocated at the target wave is evaluated according to the order parameter set of the delivery resource at the target wave, including: inputting the order parameter set of the target wave of the delivery resource in the trained wave timeout model, and calculating the delivery The timeout probability of the resource at the target wave.
  • the wave timeout model is trained based on the set of historical order parameters of the distribution resources, including: using the sigmoid function as a conversion function to convert the wave timeout function; enter the historical wave of the distribution resource in the converted wave timeout function The set of historical order parameters for a timeout timeout value; compare the timeout probability value with the timeout result of the historical wave, and continuously adjust the weight value of the wave timeout function according to the comparison result until the weight value changes less than the amplitude threshold or reaches The preset number of adjustments.
  • the above method further includes: receiving orders to be distributed; and allocating the orders to be distributed to multiple distribution resources at the distribution site.
  • the distribution of orders for distribution resources according to the timeout probability includes comparing the timeout probabilities of the orders to be distributed among multiple distribution resources; assigning the orders to be distributed to the distribution resources with the lowest timeout probability.
  • FIG. 1 is a flowchart of an order allocation method according to the first embodiment of the present application
  • FIG. 2-1 is a flowchart of an order allocation method according to the second embodiment of the present application.
  • 2-3 is a graph of the igmoid function according to the second embodiment of the present application.
  • FIG. 3 is a flowchart of an order allocation method according to the third embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of an order distribution device according to a fourth embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present application.
  • the first embodiment of the present application relates to an order allocation method, as shown in FIG. 1, including the following steps.
  • Step 103 Obtain the order parameter set of the distribution resource in the target wave, and the wave is the period from one no-load state to the next no-load state of the distribution resource.
  • step 104 the timeout probability of the order to be allocated at the target wave is evaluated according to the order parameter set of the distribution resource at the target wave.
  • Step 105 Perform order allocation on the distribution resources according to the timeout probability of the order to be allocated at the target wave.
  • one wave is the period from one no-load state to the next no-load state of the rider, that is, one wave means that the rider has no order from one time
  • the specific status of the rider during a wave is from the time the order has not been taken until it is completely delivered.
  • the rider’s pick-up point should be It is very dense, and the delivery points should be very smooth, so that the rider can get the best business performance.
  • rider A there is no order (no load) at 10:00 am. At this time, rider A receives 30 orders and all of them are not taken. These 30 orders belong to two adjacent orders (for example, Delivery site), and the distance between the delivery points of these 30 orders does not exceed 5 kilometers. During the subsequent 10:00-12:00, rider A performs the delivery service, which is expected to end at 12:00, Rider A delivered all these 30 orders, that is, at 12:00, Rider A was in the state of no next order (empty load), then this stage during 10:00-12:00 is considered a wave.
  • the distribution time of the latest order in a wave in this embodiment is earlier than the meal-taking time of the earliest order in a wave.
  • the allocation time of the 30th order is 9:50
  • the earliest order taking time of Rider A is 10:01, so that Rider A can The system allocates all orders and then picks up the order, so that the rider can better plan the route and time of order picking and delivery, thus improving the efficiency of order delivery.
  • the target wave in this embodiment refers to the current wave of distribution resources when an order to be allocated is newly added, that is, the period from which the unretrieved orders of the distribution resources and the to-be-allocated orders are all taken from the time when they are not taken to all delivered. For example, for rider A, 30 orders have been received at 10:00 am and all have not been taken. At this time, if a new order is added, rider A will take all 31 orders at 10:30, at 12: 00 will deliver all 31 orders, then take 10:00-12:00 as the target wave, and the newly added order to be allocated belongs to the order in the target wave.
  • the method before obtaining the order parameter set of the distribution resource at the target wave, as shown in FIG. 1, the method further includes the following steps.
  • Step 101 Receive an order to be distributed.
  • Step 102 Allocate orders to be distributed to multiple distribution resources at the distribution site.
  • the order to be allocated is first distributed to multiple candidate riders at the delivery site, and then the target wave of each candidate rider is obtained, and the target wave is obtained according to the target wave To get the order parameter set of the candidate rider at the target wave. For example, at 10:00 am, a new order to be allocated is added. At this time, the candidate rider has received 30 orders and all of them have not been taken. Then the 31 orders are divided into the orders of the candidate rider's target wave, and these 31 orders are obtained. The order parameter set of the order.
  • the order parameter set includes at least the maximum back ability of the candidate rider, the current back order quantity of the candidate rider, the number of overtime orders planned for the candidate rider's simulated path, the distance from the candidate rider to the order taking point and the current order taking point
  • the pressure coefficient of (for example, the pressure coefficient of the distribution station).
  • the distribution site is a place where logistics downstream distributors, retailers, and customers of a certain business district (such as Jing'an Temple business district and Wujiaochang business district) make distribution processes in the logistics supply chain. It uses circulation facilities and information systems.
  • the platform is used for flipping, sorting, circulation processing, matching, designing transportation routes and transportation methods for the goods handled by logistics, so as to provide tailor-made distribution services for service objects.
  • the maximum back order ability of the rider refers to the maximum order quantity that the rider can bear at one time
  • the current back order quantity of the rider refers to the order quantity currently undertaken by the rider
  • the overtime odd number of the rider's simulated path planning refers to the simulated path
  • the number of overtime orders in the orders delivered by the rider under planning, the distance from the rider to the order taking point usually refers to the distance between the rider and the restaurant taking the order
  • the pressure coefficient of the current order taking point refers to the number of push orders and the current order taking point
  • the above parameters have a great influence on whether the order delivery is overtime. Through these parameters to evaluate the rider's overtime probability in a wave, the evaluation result is more accurate and the reference value is greater.
  • the timeout probability of the order to be allocated at the target wave is evaluated based on the acquired order parameter set of the rider at the target wave, and the order is assigned to the rider according to the timeout probability. For example, compared with other candidate riders at the delivery site, if the estimated probability of the order to be assigned is higher in the target wave of candidate Rider A, the candidate rider A is not assigned the order to be assigned, if the evaluated order to be assigned is Candidate Rider A’s target wave has a low probability of overtime, and the order to be allocated is allocated to candidate Rider A for delivery.
  • the rider is a distribution resource.
  • the distribution resource also includes distribution vehicles, distribution robots, etc., and is not limited to the content exemplified in this embodiment.
  • This embodiment obtains the order parameter set of the rider at the target wave, and evaluates the time-out probability of the rider to allocate the order at the target wave according to the order parameter set, avoiding the variability caused by a single order parameter for time-out evaluation It improves the accuracy of the evaluation of whether the order delivery is overtime. In this way, allocating orders to the rider according to the overtime probability obtained by the evaluation can reduce the probability of overtime of the order and improve the delivery efficiency.
  • the second embodiment of the present application provides an order allocation method, as shown in FIG. 2-1, which includes the following steps.
  • Step 201 Obtain the order parameter set of the distribution resource in the target wave, and the wave is the period from one empty state to the next empty state of the distribution resource.
  • Step 205 Input the order parameter set of the target wave of the distribution resource in the trained wave timeout model, and calculate the timeout probability of the order to be allocated at the target wave.
  • Step 206 Perform order allocation on the distribution resources according to the timeout probability of the order to be allocated at the target wave.
  • step 205 as shown in FIG. 2-1, the order allocation method based on wave timeout estimation in this embodiment further includes the following steps.
  • Step 202 Acquire a historical order parameter set of historical waves of distribution resources.
  • Step 203 Construct a wave timeout model of the distribution resource.
  • the wave timeout model includes a wave timeout function.
  • the wave timeout function takes the parameters in the order parameter set of the distribution resource as variables, and each variable is assigned a corresponding weight value.
  • Step 204 Use a logistic regression algorithm to train the wave timeout model according to the historical order parameter set of the distribution resource.
  • Step 201 and step 206 of this embodiment are the same as those of step 101 and step 103 of the first embodiment, and are not repeated here.
  • this embodiment essentially uses a machine learning algorithm model to build a rider's wave timeout model based on multiple wave timeout factors that may affect the rider, and trains the wave timeout model To get the optimized wave timeout model, and use the optimized wave timeout model to calculate the timeout probability of the order to be allocated at the rider's target wave.
  • the parameters of the function are at least related to the existing backorder amount of the distribution resource at the target wave, and may also involve other influences of the distribution resource at the target wave. Variables of efficiency. For example, for the wave of rider A between 10:00 and 12:00, when evaluating the timeout probability of a new order to be assigned to this wave of rider A, first obtain multiple order parameters of these 31 orders ( The order parameters include at least the existing back order quantity of rider A. 30), multiple order parameters are calculated in combination with the wave timeout model. Specifically, the wave timeout model in this embodiment may be trained by a logistic regression algorithm. Calculate to get the estimated timeout probability.
  • each variable of the wave timeout model is assigned a corresponding weight value to reflect the influence of the parameter in predicting the timeout probability of distribution resources in a wave. For example, it can be estimated from the historical parameter set of historical waves that the rider’s ability to back orders has a greater impact on the rider’s time-out probability in a wave, and a larger weight value can be assigned to the parameter of back orders; if the rider The order delivery range of a wave is not far from the order taking point (such as the delivery station) within 3 kilometers, that is, the distance from the rider to the order taking point. This parameter has little effect on the rider's timeout probability in a wave , Then the distance from the rider to the single point can be given a smaller weight value. According to the different influences of different parameters on the calculation results, the corresponding weight values are assigned to the different parameters, which can obtain a better wave time-out model, which is beneficial to obtain more accurate calculation results after training.
  • the process of training the wave timeout model using a logistic regression algorithm includes the following steps.
  • step 2041 the wave timeout function is converted using the sigmoid function as the conversion function.
  • Step 2042 Input the historical order parameter set of the historical wave of the distribution resource in the converted wave timeout function to obtain a timeout probability value.
  • Step 2043 Compare the time-out probability value with the time-out result of the historical wave, and continuously adjust the weight value of the wave time-out function according to the comparison result until the weight value changes less than the amplitude threshold or reaches a preset number of adjustments.
  • Logistic regression is a simple and effective classification algorithm.
  • the sigmoid function is usually used as the conversion function.
  • the calculation result is to obtain the classification label of the sample data.
  • the expression of the sigmoid function is:
  • the graph of the sigmoid function is shown in Figure 2-3.
  • the wave timeout function is converted by the sigmoid function so that the calculated value of the wave timeout function and the actual value of the time-out result of the historical wave The difference between them is compressed to between 0 and 1 cells, which can compress the huge vibration of the data, and it is convenient to obtain the classification label of the actual value (the classification is based on whether the calculation result of the sigmoid function is greater than 0.5), and the wave timeout The optimal weight value of the function.
  • X1, X2, X3... are the order parameters in the order parameter set
  • W1, W2, W3... are the weight values of the construction.
  • the weight values W1, W2, W3... should be adjusted so that the calculated value of the fitted wave timeout function is less than 0.3, which is closer to the other history
  • the actual value of the time-out result of the wave; according to this method, the historical order parameter set of the rider's historical wave is used to continuously adjust the weight value of the wave time-out function until the weight value changes less than the preset amplitude threshold, For example, the weight value of each adjustment changes by less than the amplitude threshold of 0.02, or until the preset number of adjustments is reached. For example, 100 times.
  • the above adjustment process can be carried out using the gradient ascent algorithm.
  • the gradient ascent algorithm refers to finding the growth direction of the wave timeout function, and iterating through the small step forward-adjusting the direction-continuing the small step forward-and finally approaching the optimal solution. Until the weight value hardly changes, the gradient ascent algorithm belongs to the prior art and will not be repeated here.
  • the wave timeout model is trained according to the historical order parameter set of the distribution resource, and the order parameter set of the target wave of the distribution resource is entered in the trained wave timeout model to calculate the distribution resource at the target wave Probability of timeout, and then allocate orders to the distribution resources according to the probability of timeout of the distribution resources at the target wave.
  • the order parameter set in this embodiment is not a specific order parameter set, but a wave parameter set for the rider. That is to say, when calculating the approximate probability of a rider's overtime to treat assigned orders, it is not specific to a certain order, but a timeout probability of the total amount of orders (including orders to be assigned) that the rider is carrying on the current wave.
  • the time-out probability is calculated through the wave time-out model, which improves the accuracy of the judgment of whether the order is delivered over time, thus providing a more accurate basis for order allocation.
  • the third embodiment of the present application provides an order allocation method, as shown in FIG. 3, including the following steps.
  • Step 301 Obtain the order parameter set of the delivery resource in the target wave, and the wave is the period from one empty state to the next empty state of the delivery resource.
  • step 302 the timeout probability of the order to be allocated at the target wave is evaluated according to the order parameter set of the distribution resource at the target wave.
  • Step 303 Compare the timeout probabilities of the target waves of the orders to be distributed among multiple distribution resources.
  • Step 304 Assign the order to be allocated to the distribution resource with the lowest timeout probability.
  • step 304 when the time-out probability of the order to be allocated at the target wave of the delivery resource is the lowest, the order to be allocated is allocated to the delivery resource, otherwise the order to be allocated is not allocated to the delivery resource.
  • Step 301 and step 302 in this embodiment are the same as those in step 101 and step 102 in the first embodiment, and will not be described in detail.
  • step 303 and step 304 of this embodiment taking the distribution resource as a rider as an example, when judging whether to assign an order to a candidate rider according to the timeout probability, it is necessary to compare multiple candidate riders at the delivery site, for the delivery site For each candidate rider, calculate the time-out probability of the order to be assigned at the target wave of the candidate rider, that is, evaluate the time-out probability of each candidate rider to deliver the order to be assigned, select the candidate rider with the lowest timeout probability, and assign the The order is assigned to this candidate rider for delivery.
  • the time-out probability of the target wave of A is 10%; the five orders of rider B and one order to be allocated are divided into orders in the target wave, and the order to be allocated is calculated based on the order parameter set of these six orders in rider B
  • the target time-out probability of the target wave is 8%; the eight orders of rider C and one order to be allocated are divided into orders in the target wave, and the order to be allocated to rider A is calculated according to the order parameter set of these nine orders
  • the time-out probability of the target wave is 15%; then, in comparison, the time-out probability of the target wave to be allocated in Rider B is the lowest, then the order to be distributed is allocated to Rider B for delivery, without giving Rider A and Rider C allocates the pending order.
  • the judgment is made by setting a timeout probability threshold, which can reasonably allocate order scheduling to the rider according to the precise timeout probability.
  • a person skilled in the art may think that when determining whether to allocate an order for the rider according to the estimated timeout probability, the present application may also use other methods of determination, which are not limited to this embodiment.
  • the fourth embodiment of the present application provides an order distribution device. As shown in FIG. 4, it includes: a parameter acquisition module 401, which obtains the order parameter set of the distribution resource in the target wave, and the wave is the distribution resource from a no-load state to The period of the next no-load state; the probability evaluation module 405, based on the order parameter set of the delivery resource at the target wave, evaluates the timeout probability of the order to be allocated at the target wave; the order allocation module 406, according to the order to be allocated at the target wave Overtime probability allocates orders to distribution resources.
  • a parameter acquisition module 401 which obtains the order parameter set of the distribution resource in the target wave, and the wave is the distribution resource from a no-load state to The period of the next no-load state
  • the probability evaluation module 405 based on the order parameter set of the delivery resource at the target wave, evaluates the timeout probability of the order to be allocated at the target wave
  • the order allocation module 406 according to the order to be allocated at the target wave Overtime probability allocates orders to distribution resources
  • the distribution time of the latest order in a wave is earlier than the meal-taking time of the earliest order in a wave. This can limit the delivery status of all orders in a wave, and further improve the accuracy of overtime probability assessment based on the wave.
  • the order parameter set acquired by the parameter acquiring module 401 includes at least one or more of the following factors: the maximum backorder capacity of the distribution resource, the existing backorder quantity of the distribution resource, the number of overtime orders for the simulated path planning, and the distribution resource The distance to the single point and the pressure coefficient of the distribution resources. Therefore, relevant parameters are provided for calculating the wave timeout probability, and the accuracy of the calculation is improved.
  • the order allocation device of this embodiment further includes: a historical parameter acquisition module 402 to acquire a historical order parameter set of historical waves of distribution resources; a model construction module 403 to construct a wave timeout model of the distribution resources, the wave timeout model includes The wave timeout function takes the parameters in the order parameter set of the distribution resource as variables, and each variable is assigned a corresponding weight value; the model training module 404 uses a logistic regression algorithm to match the historical order parameter set pair of the distribution resource Wave time-out model for training.
  • the probability evaluation module 405 evaluates the timeout probability of the delivery resource at the target wave based on the order parameter set of the delivery resource at the target wave, including: inputting the order of the target wave of the delivery resource in the trained wave timeout model Set of parameters to calculate the timeout probability of the order to be allocated at the target wave.
  • the model training module 404 trains the wave timeout model according to the historical order parameter set of the distribution resources, including: using the sigmoid function as a conversion function to convert the wave timeout function; input in the converted wave timeout function The historical order parameter set of the historical wave of the distribution resource to obtain the timeout probability value; compare the timeout probability value with the historical wave timeout result, and continuously adjust the weight value of the wave timeout function according to the comparison result until the weight value changes The amplitude is less than the amplitude threshold or the preset number of adjustments is reached.
  • the order allocation device of this embodiment further includes: an order receiving module 407, which receives the order to be allocated; an order processing module 408, which allocates the order to be allocated to a plurality of delivery resources of the delivery site.
  • the order allocation module 406 allocates orders to distribution resources according to the timeout probability, which may specifically include: comparing the timeout probability of the order to be allocated to the target wave of multiple distribution resources; allocating the order to be allocated to the delivery with the lowest probability of timeout Resources.
  • the order distribution device of this embodiment evaluates the time-out probability of the rider's wave from no-load (no order) to no-load (all orders are delivered), thereby improving the accuracy of the judgment of whether the order delivery times out. Improve the effectiveness of order allocation and enhance the user experience.
  • This embodiment provides an electronic device, as shown in FIG. 5, including at least one processor 501; and a memory 502 communicatively connected to the at least one processor 501; wherein, the memory 502 stores executable by at least one processor 501
  • the instruction is executed by at least one processor 501 to achieve: obtain the order parameter set of the distribution resource at the target wave, the wave is the period from one empty state of the distribution resource to the next empty state; according to the distribution resource at the target
  • the order parameter set of the wave evaluates the time-out probability of the order to be allocated at the target wave; the order allocation of the distribution resources is based on the time-out probability of the order to be allocated at the target wave.
  • the memory 502 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs, and modules.
  • the processor 501 runs non-volatile software programs, instructions, and modules stored in the memory 502 to execute various functional applications and data processing of the device, that is, to implement the above-mentioned order distribution method.
  • the memory 502 may include a storage program area and a storage data area, wherein the storage program area may store an operating system and application programs required by at least one function; the storage data area may store historical data of shipping network transportation, and the like.
  • the memory 502 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.
  • the memory 502 may optionally include memories remotely set with respect to the processor 501, and these remote memories may be connected to an external device through a network. Examples of the above network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.
  • One or more modules are stored in the memory 502, and when executed by one or more processors 501, execute the order allocation method in any of the above method embodiments.
  • the sixth embodiment of the present application relates to a non-volatile storage medium for storing a computer readable program, the computer readable program being used for a computer to execute some or all of the above method embodiments.
  • a program which is stored in a storage medium and includes several instructions to make a device ( It may be a single chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in the embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code .

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Educational Administration (AREA)
  • Game Theory and Decision Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Debugging And Monitoring (AREA)

Abstract

一种订单分配方法、装置、电子设备及存储介质，属于通信技术领域。所述方法包括：获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间(103)；根据配送资源在目标波次的订单参数集合评估待分配订单在目标波次的超时概率(104)；根据待分配订单在目标波次的超时概率对配送资源进行订单分配(105)。

Description

订单分配方法、装置、电子设备和存储介质
交叉引用
本申请引用于2018年11月30日递交的名称为“订单分配方法、装置、电子设备和存储介质”的第201811458962.6号中国专利申请，其通过引用被全部并入本申请。
技术领域
本申请实施例通信技术领域，特别涉及一种订单分配方法、装置、电子设备和存储介质。
背景技术
在订单分配系统分配订单时，通常调度分配的不合理会导致配送超时的情况出现，因此，在分配订单时会全局考虑最适合的骑手，不会把订单分配给存在超时风险的骑手。
目前，对骑手配送订单是否可能超时进行判断基于以下两种方案。
一是根据骑手的最大背单能力(骑手负载订单的最大量)来预估骑手是否存在超时风险，从而实现骑手的分单调度。然而，最大背单量无法体现出骑手目前背单的具体送单路径和送单顺序，每个订单的难易程度(例如是否顺路)不同，因此无法对骑手是否超时进行精准评估。
二是对骑手的现有背单进行一个路径规划模拟，即模拟骑手的送单顺序判断骑手是否存在超时风险，从而实现骑手的分单调度。然而,对骑手送单路径模拟时，送单顺序和路径规划不是完全准确的、会存在误差。订单量越多，对每配送一个订单的超时预估偏差就会一直向后传递并积累。此外，路径规划和送单顺序的模拟跟真实发生的也不一定相符。基于这两点，在对骑手超时判断上也无法做到准确预估。
发明内容
本申请实施例的目的在于提供一种订单分配方法、装置、电子设备和存储介质，通过提出一个波次的概念，基于骑手从空载(没单)到空载(全部订单送完)的这一个波次的订单来对待分配订单的超时概率进行评估，提高了订单配送是否超时判断的精准性，从而提高了订单分配的有效性，提升了用户体验。
为解决上述技术问题，本申请的实施例提供了一种订单分配方法，包括：获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间；根据配送资源在目标波次的订单参数集合评估配送资源在目标波次的超时概率；根据配送资源在目标波次的超时概率对配送资源进行订单分配。
本申请的实施例还提供了一种订单分配装置，包括：参数获取模块，获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间；概率评估模块，根据配送资源在目标波次的订单参数集合评估配送资源在目标波次的超时概率；订单分配模块，根据配送资源在目标波次的超时概率对配送资源进行订单分配。
本申请的实施例还提供了一种电子设备，包括至少一个处理器；以及，与至少一个处理器通信连接的存储器；其中，存储器存储有可被至少一个处理器执行的指令，指令被至少一个处理器执行以实现：获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间；根据配送资源在目标波次的订单参数集合评估配送资源在目标波次的超时概率；根据配送资源在目标波次的超时概率对配送资源进行订单分配。
本申请的实施例还提供了一种非易失性存储介质，用于存储计算机可读程序，所述计算机可读程序用于供计算机执行上述的订单分配方法。
另外，一个波次中最晚订单的分配时间早于一个波次中最早订单的取餐时间。
另外，上述方法还包括：获取配送资源的历史波次的历史订单参数集合；构建配送资源的波次超时模型，波次超时模型包含波次超时函数，波次超时函数以配送资源的订单参数集合中的参数为变量，每个变量赋予相应的权重值；利用逻辑回归算法，根据配送资源的历史订单参数集合对波次超时模型进行训 练。
另外，根据配送资源在目标波次的订单参数集合评估待分配订单在目标波次的超时概率，包括：在训练后的波次超时模型中输入配送资源的目标波次的订单参数集合，计算配送资源在目标波次的超时概率。
另外，根据配送资源的历史订单参数集合对波次超时模型进行训练，包括：使用sigmoid函数作为转换函数对波次超时函数进行转换；在转换后的波次超时函数中输入配送资源的历史波次的历史订单参数集合，得到超时概率值；将超时概率值与历史波次的超时结果进行比较，并根据比较结果来不断调整波次超时函数的权重值，直到权重值变化幅度小于幅度阈值或者达到预设的调整次数。
另外，上述方法还包括：接收待分配订单；将待分配订单分配给配送站点的多个配送资源。
另外，根据超时概率对配送资源进行订单分配，包括比较待分配订单在多个配送资源的目标波次的超时概率；将待分配订单分配给超时概率最低的配送资源。
附图说明
一个或多个实施例通过与之对应的附图中的图片进行示例性说明，这些示例性说明并不构成对实施例的限定，附图中具有相同参考数字标号的元件表示为类似的元件，除非有特别申明，附图中的图不构成比例限制。
图1是根据本申请第一实施例中的一种订单分配方法的流程图；
图2-1是根据本申请第二实施例中的一种订单分配方法的流程图；
图2-2是根据本申请第二实施例中利用逻辑回归算法对波次超时模型进行训练的方法流程图；
图2-3是是根据本申请第二实施例中s igmoid函数的曲线图；
图3是根据本申请第三实施例中的一种订单分配方法的流程图；
图4是根据本申请第四实施例中的一种订单分配装置的结构示意图；
图5是根据本申请第五实施例中的一种电子设备的结构示意图。
具体实施例
为使本申请实施例的目的、技术方案和优点更加清楚，下面将结合附图对本申请的各实施例进行详细的阐述。然而，本领域的普通技术人员可以理解，在本申请各实施例中，为了使读者更好地理解本申请而提出了许多技术细节。但是，即使没有这些技术细节和基于以下各实施例的种种变化和修改，也可以实现本申请所要求保护的技术方案。
第一实施例
本申请的第一实施例涉及一种订单分配方法，如图1所示，包括以下步骤。
步骤103，获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间。
步骤104，根据配送资源在目标波次的订单参数集合评估待分配订单在目标波次的超时概率。
步骤105，根据待分配订单在目标波次的超时概率对配送资源进行订单分配。
具体地，以配送资源为一个骑手为例进行说明，步骤103中，一个波次为骑手从一个空载状态到下一个空载状态的期间，即一个波次即该骑手从一次没有订单的状态到下一次订单全部送完的期间，具体执行时，该骑手的在一个波次期间的状态是将订单从尚未取单到全部送完，优选地，在实际应用中，该骑手的取点应很密集，送点应很顺路，以便于骑手能够获得最佳的业务表现。
例如对于骑手A，在上午10：00时处于一次没有订单(空载)的状态，此时骑手A接收到30个订单并且全部未取，这30个订单属于两个邻近的取单点(例如配送站点)所分配的订单，并且这30个订单的送点之间的距离不超过5公里，在之后的10：00-12：00期间，骑手A执行送单业务，预计截至12：00，骑手A将这30个订单全部配送完毕，即12：00时骑手A处于下一次没有订单(空载)的状态，那么10：00-12：00期间的这个阶段被认为是一个波次。
作为进一步优化，本实施例的一个波次中最晚订单的分配时间早于一个波次中最早订单的取餐时间。例如骑手A在10：00-12：00期间的这个波次的订单中，第30个订单的分配时间为9：50，骑手A最早订单的取餐时间为10：01，这样骑手A能够在系统将订单全部分配完成之后再去取单，方便骑手更好 地去规划取单和送单的路径和时间，从而提高了订单配送的效率。
本实施例的目标波次，是指新增待分配订单时配送资源的当前波次，即将配送资源未取的订单和待分配订单共同从全部未取到全部送完的期间。例如同样对于骑手A，在上午10：00时已接收30个订单且全部未取，此时若新增一个待分配订单，骑手A在10：30将这31个订单全部取完，在12：00将这31个订单全部送完，那么将10：00-12：00作为目标波次，该新增的一个待分配订单就属于目标波次中的订单。
进一步地，本实施例的订单分配方法，在获取配送资源在目标波次的订单参数集合之前，如图1所示，还包括以下步骤。
步骤101，接收待分配订单。
步骤102，将待分配订单分配给配送站点的多个配送资源。
具体来讲，本实施例的上述实例中，新增待分配订单时，首先将该待分配订单分配给配送站点的多个候选骑手，然后获取每个候选骑手的目标波次，根据目标波次来获取候选骑手在目标波次的订单参数集合。例如上午10：00新增一个待分配订单，此时候选骑手已接收30个订单且全部未取，那么将这31个订单共同划分为候选骑手的目标波次中的订单，并获取这31个订单的订单参数集合。更为具体地，订单参数集合至少包括候选骑手的最大背单能力、候选骑手现有背单量、针对候选骑手的模拟路径规划的超时单数、候选骑手到取单点的距离和当前取单点的压力系数(例如配送站点的压力系数)。配送站点是在物流供应链环节中,为某个商圈(例如静安寺商圈、五角场商圈)的物流下游经销商、零售商、客户作配送工序的场所，其利用流通设施、信息系统平台，对物流经手的货物,作倒装、分类、流通加工、配套、设计运输路线、运输方式,从而为服务对象提供度身配送服务。
其中，骑手的最大背单能力是指该骑手一次能够承担的最大的订单量，骑手的现有背单量是指该骑手当前承担的订单量，骑手的模拟路径规划的超时单数是指模拟路径规划下骑手配送的单中超时单的数量，骑手到取单点的距离是通常指骑手到取单的餐厅之间的距离，当前取单点的压力系数是指当前取单点的推单数与当前取单点在线骑手最大背单能力之和的比值。上述几个参数对于订单配送是否超时具有很大的影响作用，通过这几个参数来评估骑手在一个波次的超时概率，评估结果更为准确，参考价值更大。
进一步地，本实施例根据获取的骑手在目标波次的订单参数集合评估待 分配订单在目标波次的超时概率，并根据超时概率对骑手分配订单。例如相比较于配送站点的其他候选骑手，如果评估得到的待分配订单在候选骑手A的目标波次的超时概率较高则不为候选骑手A分配待分配订单，如果评估得到的待分配订单在候选骑手A的目标波次的超时概率较低，则将待分配订单分配给候选骑手A进行配送。
值得一提的是，本实施例中，骑手为一种配送资源，本领域技术人员可以理解，配送资源还包括配送车辆、配送机器人等，不以本实施例所例举的内容为限。
本实施例获取了骑手在目标波次的订单参数集合，并根据订单参数集合来评估骑手对待分配订单在目标波次的超时概率，避免了通过单一的订单参数进行超时评估所带来的多变性，提高了订单配送是否超时评估的精准性，如此根据评估得到的超时概率对骑手分配订单，能够降低订单超时的概率，提高配送效率。
第二实施例
本申请第二实施例提供了一种订单分配方法，如图2-1所示，包括以下步骤。
步骤201，获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间。
步骤205，在训练后的波次超时模型中输入配送资源的目标波次的订单参数集合，计算待分配订单在目标波次的超时概率。
步骤206，根据待分配订单在目标波次的超时概率对配送资源进行订单分配。
进一步地，在步骤205之前，如图2-1所示，本实施例的基于波次超时预估的订单分配方法还包括以下步骤。
步骤202，获取配送资源的历史波次的历史订单参数集合。
步骤203，构建配送资源的波次超时模型，波次超时模型包含波次超时函数，波次超时函数以配送资源的订单参数集合中的参数为变量，每个变量赋予相应的权重值。
步骤204，利用逻辑回归算法，根据配送资源的历史订单参数集合对波次超时模型进行训练。
本实施例的步骤201、步骤206与第一实施例的步骤101、步骤103内容 相同，不再赘述。
具体地，在步骤202-205中，本实施例实质上是通过机器学习算法模型，根据多个可能影响骑手的波次超时因素来构建骑手的波次超时模型，通过对波次超时模型进行训练来得到优化的波次超时模型，并且利用优化的波次超时模型来计算待分配订单在骑手的目标波次的超时概率。
可选地，本实施例的波次超时模型的波次超时函数中，函数的参数至少涉及配送资源在目标波次的现有背单量，还可以涉及配送资源在目标波次的其他影响配送效率的变量。例如对于骑手A在10：00-12：00期间的波次，在评估新增的一个待分配订单在骑手A的这个波次的超时概率时，首先获取这31个订单的多个订单参数(订单参数中至少包括骑手A的现有背单量30)，将多个订单参数结合波次超时模型进行计算，具体地，可通过逻辑回归算法来训练本实施例中的波次超时模型，通过计算即可得到预估的超时概率。
需要说明的是，波次超时模型的每个变量赋予相应的权重值，以体现参数在预测一个波次中配送资源的超时概率时所具备的影响力大小。例如从历史波次的历史参数集合中可预估，骑手的背单能力对骑手在一个波次的超时概率具有较大的影响，可为背单能力这个参数赋予较大的权重值；若骑手的在一个波次的订单配送范围均未远离取单点(例如配送站点)3公里的范围内，即骑手到取单点的距离这个参数对于骑手在一个波次的超时概率的影响作用较小，那么可为骑手到取单点的距离这个参数赋予较小的权重值。根据不同参数对计算结果的影响力的不同来对不同参数赋予相应的权重值，能够获得较优的波次超时模型，有利于通过训练之后得到更为准确的计算结果。
本实施例中，如图2-2所示，利用逻辑回归算法对波次超时模型进行训练的过程包括以下步骤。
步骤2041，使用sigmoid函数作为转换函数对波次超时函数进行转换。
步骤2042，在转换后的波次超时函数中输入配送资源的历史波次的历史订单参数集合，得到超时概率值。
步骤2043，将超时概率值与历史波次的超时结果进行比较，并根据比较结果来不断调整波次超时函数的权重值，直到权重值变化幅度小于幅度阈值或者达到预设的调整次数。
逻辑回归实质是一种简单而效果好的分类算法，在逻辑回归算法的思想中，通常使用sigmoid函数作为转换函数，在该场景中，计算结果是要得到对 样本数据的分类标签。sigmoid函数的表达式为：
f(x)＝1/(1+e^-x)(-x是幂数)
进一步地，sigmoid函数的曲线图如图2-3所示，本实施例中，通过sigmoid函数对波次超时函数进行转换，使得波次超时函数的计算值与历史波次的超时结果的实际值之间的差值，压缩到0～1的小区间，从而能够压缩数据的巨幅震荡，方便得到实际值的分类标签(分类以sigmoid函数的计算结果是否大于0.5为依据)，和波次超时函数的最优权重值。
本实施例中，假设构建的拟合波次超时函数为：
Y(拟合)＝W1*X1+W2*X2+W3*X3···
其中，X1、X2、X3···为订单参数集合中的各订单参数，W1、W2、W3···为构建的权重值，若将骑手的一个历史波次的历史订单参数集合代入上述拟合波次超时函数，得到该一个历史波次的超时概率值Y1(拟合)，例如Y1(拟合)为0.6，将0.6代入sigmoid函数中，得到sigmoid函数的函数值f(X)为0.51，即拟合波次超时函数的计算值与该一个历史波次的超时结果的实际值之间的差值为0.51，由于0.51大于0.5，可以判断该一个历史波次的超时结果的实际值应当是位于拟合波次超时函数的函数曲线的上方，此时应当调整构建的权重值W1、W2、W3···，使得拟合波次超时函数的计算值大于0.6，从而更为接近该一个历史波次的超时结果的实际值；一次调整之后，将骑手的另一个历史波次的历史订单参数集合代入上述拟合波次超时函数得到该一个历史波次的超时概率值Y1(拟合)，例如Y1(拟合)为0.3，将0.3代入sigmoid函数中，得到sigmoid函数的函数值f(X)为0.48，由于0.48小于0.5，可以判断该一个历史波次的超时结果的实际值应当位于拟合波次超时函数的函数曲线的下方，此时应当调整构建的权重值W1、W2、W3···，使得拟合波次超时函数的计算值小于0.3，从而更为接近该另一个历史波次的超时结果的实际值；根据这样的方法，利用骑手的历史波次的历史订单参数集合不断对波次超时函数的权重值进行调整，直到权重值的变化幅度小于预设的幅度阈值，例如每次调整的权重值的变化幅度小于幅度阈值0.02，或者直到达到了预设的调整次数.例如100次。
上述调整的过程可以采用梯度上升算法来进行，梯度上升算法是指找到波次超时函数的增长方向，通过小步前进-调整方向-继续小步前进-最终逼近最优解的方式不断迭代运算直到权重值几乎不再变化为止，梯度上升算法属于现有技术，在此不做赘述。
如此，根据配送资源的历史订单参数集合对波次超时模型进行训练，在训练后的波次超时模型中输入配送资源的目标波次的订单参数集合，即可计算获得配送资源在目标波次的超时概率，继而根据配送资源在目标波次的超时概率对配送资源进行订单分配。
本实施例的订单参数集合不是具体某个订单的参数集合，而是骑手一个波次的参数集合。即计算一个骑手对待分配订单的超时概率大概是多少时，不是具体到某一单上，而是计算该骑手在当前波次所背的订单总量(包括待分配订单)的一个超时概率。通过波次超时模型计算超时概率，提高了订单配送是否超时判断的精准性，从而为订单分配提供了更加准确的依据。
第三实施例
本申请第三实施例提供了一种订单分配方法，如图3所示，包括以下步骤。
步骤301，获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间。
步骤302，根据配送资源在目标波次的订单参数集合评估待分配订单在目标波次的超时概率。
步骤303，比较待分配订单在多个配送资源的目标波次的超时概率。
步骤304，将待分配订单分配给超时概率最低的配送资源。
具体地，步骤304中，当待分配订单在配送资源的目标波次的超时概率为最低时，对配送资源分配该待分配订单，否则不对配送资源分配该待分配订单。
本实施例的步骤301、步骤302与第一实施例中的步骤101、步骤102内容相同，不再赘述。
具体地，本实施例的步骤303和步骤304中，以配送资源为骑手为例，当根据超时概率判断是否对候选骑手分配订单时，需要对配送站点的多个候选骑手进行比较，针对配送站点的每一个候选骑手，分别计算待分配订单在该候选骑手的目标波次的超时概率，即评估每一个候选骑手配送该待分配订单的超时概率，选取其中超时概率最低的候选骑手，将待分配订单分配给这个候选骑手进行配送。
例如当新增一个待分配订单时，配送站点当前在线的有3个骑手，其中骑手A已接收且未取单的订单为三个，骑手B已接收且未取单的订单有五个， 骑手C已接收且未取单的订单有八个，将骑手A的三个订单和一个待分配订单共同划分为目标波次中的订单，根据这四个订单的订单参数集合计算待分配订单在骑手A的目标波次的超时概率为10％；将骑手B的五个订单和一个待分配订单共同划分为目标波次中的订单，根据这六个订单的订单参数集合计算待分配订单在骑手B的目标波次的超时概率为8％；将骑手C的八个订单和一个待分配订单共同划分为目标波次中的订单，根据这九个订单的订单参数集合计算待分配订单在骑手A的目标波次的超时概率为15％；那么比较而言，待分配订单在骑手B的目标波次的超时概率是最低的，则将待分配订单分配给骑手B进行配送，而不向骑手A和骑手C分配该待分配订单。
本实施例通过设置超时概率阈值来进行判断，能根据精准的超时概率合理的对骑手分配订单调度。此外，本领域技术人员可以想到，本申请在根据预估超时概率判断是否为骑手分配订单时，还可以采用其他的判断方法，并不为本实施例所限。
第四实施例
本申请第四实施例提供了一种订单分配装置，如图4所示，包括：参数获取模块401，获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间；概率评估模块405，根据配送资源在目标波次的订单参数集合评估待分配订单在目标波次的超时概率；订单分配模块406，根据待分配订单在目标波次的超时概率对配送资源进行订单分配。
本实施例中，参数获取模块401中，一个波次中最晚订单的分配时间早于一个波次中最早订单的取餐时间。这样能够限制了一个波次中的所有订单的配送状态，进一步提高基于波次来进行超时概率评估的准确性。
更为具体地，参数获取模块401获取的订单参数集合至少包括以下因素的一个或多个：配送资源的最大背单能力、配送资源的现有背单量、模拟路径规划的超时单数、配送资源到取单点的距离和配送资源的压力系数。从而为波次超时概率的计算提供了相关参数，提高了计算的准确率。
另外，本实施例的订单分配装置还包括：历史参数获取模块402，获取配送资源的历史波次的历史订单参数集合；模型构建模块403，构建配送资源的波次超时模型，波次超时模型包含波次超时函数，波次超时函数以配送资源的订单参数集合中的参数为变量，每个变量赋予相应的权重值；模型训练模块404，利用逻辑回归算法，根据配送资源的历史订单参数集合对波次超时模型进行训 练。
作为优选地，概率评估模块405根据配送资源在目标波次的订单参数集合评估配送资源在目标波次的超时概率，包括：在训练后的波次超时模型中输入配送资源的目标波次的订单参数集合，计算待分配订单在目标波次的超时概率。
进一步地，模型训练模块404根据配送资源的历史订单参数集合对波次超时模型进行训练，包括：使用s igmoid函数作为转换函数对波次超时函数进行转换；在转换后的波次超时函数中输入配送资源的历史波次的历史订单参数集合，得到超时概率值；将超时概率值与历史波次的超时结果进行比较，并根据比较结果来不断调整波次超时函数的权重值，直到权重值变化幅度小于幅度阈值或者达到预设的调整次数。
另外，本实施例的订单分配装置还包括：订单接收模块407，接收待分配订单；订单处理模块408，将待分配订单分配给配送站点的多个配送资源。
在实际应用中，订单分配模块406根据超时概率对配送资源分配订单，具体可以包括：比较待分配订单在多个配送资源的目标波次的超时概率；将待分配订单分配给超时概率最低的配送资源。
本实施例的订单分配装置，通过对骑手从空载(没单)到空载(全部订单送完)的这一个波次的超时概率进行评估，提高了订单配送是否超时判断的精准性，从而提高了订单分配的有效性，提升了用户体验。
第五实施例
本实施例提供了一种电子设备，如图5所示，包括至少一个处理器501；以及与至少一个处理器501通信连接的存储器502；其中，存储器502存储有可被至少一个处理器501执行的指令，指令被至少一个处理器501执行以实现：获取配送资源在目标波次的订单参数集合，波次为配送资源从一个空载状态到下一个空载状态的期间；根据配送资源在目标波次的订单参数集合评估待分配订单在目标波次的超时概率；根据待分配订单在目标波次的超时概率对配送资源进行订单分配。
具体地，图5中以通过总线连接为例。存储器502作为一种非易失性计算机可读存储介质，可用于存储非易失性软件程序、非易失性计算机可执行程序以及模块。处理器501通过运行存储在存储器502中的非易失性软件程序、指令以及模块，从而执行设备的各种功能应用以及数据处理，即实现上述订单 分配方法。
存储器502可以包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需要的应用程序；存储数据区可存储航运网络运输的历史数据等。此外，存储器502可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。在一些实施例中，存储器502可选包括相对于处理器501远程设置的存储器，这些远程存储器可以通过网络连接至外接设备。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
一个或者多个模块存储在存储器502中，当被一个或者多个处理器501执行时，执行上述任意方法实施例中的订单分配方法。
上述产品可执行本申请实施例所提供的方法，具备执行方法相应的功能模块和有益效果，未在本实施例中详尽描述的技术细节，可参见本申请实施例所提供的方法。
第六实施例
本申请的第六实施例涉及一种非易失性存储介质，用于存储计算机可读程序，所述计算机可读程序用于供计算机执行上述部分或全部的方法实施例。
即，本领域技术人员可以理解，实现上述实施例方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序存储在一个存储介质中，包括若干指令用以使得一个设备(可以是单片机，芯片等)或处理器(processor)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。

Claims (16)

  1. 一种订单分配方法，包括：
    获取配送资源在目标波次的订单参数集合，所述波次为所述配送资源从一个空载状态到下一个空载状态的期间；
    根据所述配送资源在所述目标波次的订单参数集合评估待配送订单在所述目标波次的超时概率；
    根据所述待配送订单在所述目标波次的超时概率对所述配送资源进行订单分配。
  2. 根据权利要求1所述的订单分配方法，其中，一个所述波次中最晚订单的分配时间早于一个所述波次中最早订单的取单时间。
  3. 根据权利要求1所述的订单分配方法，其中，所述方法还包括：
    获取所述配送资源的历史波次的历史订单参数集合；
    构建所述配送资源的波次超时模型，所述波次超时模型包含波次超时函数，所述波次超时函数以所述配送资源的订单参数集合中的参数为变量，每个所述变量赋予相应的权重值；
    利用逻辑回归算法，根据所述历史订单参数集合对所述波次超时模型进行训练。
  4. 根据权利要求3所述的方法，其中，所述根据所述配送资源在目标波次的订单参数集合评估待分配订单在所述目标波次的超时概率，包括：
    在训练后的所述波次超时模型中输入所述配送资源的目标波次的订单参数集合，计算所述待分配订单在所述目标波次的超时概率。
  5. 根据权利要求3所述的订单分配方法，其中，所述根据所述配送资源的历史订单参数集合对所述波次超时模型进行训练，包括：
    使用sigmoid函数作为转换函数对所述波次超时函数进行转换；
    在转换后的波次超时函数中输入所述历史波次的历史订单参数集合，得到超时概率值；
    将所述超时概率值与所述历史波次的超时结果进行比较，并根据比较结果来不断调整所述波次超时函数的权重值，直到所述权重值变化幅度小于幅度阈值或者达到预设的调整次数。
  6. 根据权利要求1至5中任一项所述的订单分配方法，其中，所述方法还 包括：
    接收待分配订单；
    将所述待分配订单分配给配送站点的多个配送资源。
  7. 根据权利要求6所述的订单分配方法，其中，所述根据所述待分配订单在所述目标波次的超时概率对所述配送资源进行订单分配，包括：
    比较所述待分配订单在所述多个配送资源的所述目标波次的超时概率；
    将所述待分配订单分配给所述超时概率最低的所述配送资源。
  8. 一种订单分配装置，
    Figure PCTCN2019120838-appb-100001
    包括：
    参数获取模块，获取配送资源在目标波次的订单参数集合，所述波次为所述配送资源从一个空载状态到下一个空载状态的期间；
    概率评估模块，根据所述配送资源在所述目标波次的订单参数集合评估所述配送资源在所述目标波次的超时概率；
    订单分配模块，根据所述配送资源在所述目标波次的超时概率对所述配送资源进行订单分配。
  9. 一种电子设备，包括至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；
    其中，所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行以实现：
    获取配送资源在目标波次的订单参数集合，所述波次为所述配送资源从一个空载状态到下一个空载状态的期间；
    根据所述配送资源在所述目标波次的订单参数集合评估待分配订单在所述目标波次的超时概率；
    根据所述待分配订单在所述目标波次的超时概率对所述配送资源进行订单分配。
  10. 根据权利要求9所述的电子设备，其中，一个所述波次中最晚订单的分配时间早于一个所述波次中最早订单的取单时间。
  11. 根据权利要求9所述的电子设备，其中，所述处理器还执行：
    获取所述配送资源的所有历史波次的历史订单参数集合；
    构建所述配送资源的波次超时模型，所述波次超时模型包含波次超时函数，所述波次超时函数以所述配送资源的订单参数集合中的参数为变量，每个所述变量赋予相应的权重值；
    利用逻辑回归算法，根据所述配送资源的历史订单参数集合对所述波次超时模型进行训练。
  12. 根据权利要求11所述的电子设备，其中，所述处理器执行根据所述配送资源在目标波次的订单参数集合评估待分配订单在所述目标波次的超时概率，包括：
    在训练后的所述波次超时模型中输入所述配送资源的目标波次的订单参数集合，计算所述配送资源在所述目标波次的超时概率。
  13. 根据权利要求11所述的电子设备，其中，所述处理器执行根据所述配送资源的历史订单参数集合对所述波次超时模型进行训练，包括：
    使用sigmoid函数作为转换函数对所述波次超时函数进行转换；
    在转换后的波次超时函数中输入所述配送资源的历史波次的历史订单参数集合，得到超时概率值；
    将所述超时概率值与所述历史波次的超时结果进行比较，并根据比较结果来不断调整所述波次超时函数的权重值，直到所述权重值变化幅度小于幅度阈值或者达到预设的调整次数。
  14. 根据权利要求9至13中任一项所述的订单分配方法，其中，所述处理器还用于执行：
    接收待分配订单；
    将所述待分配订单分配给配送站点的多个配送资源。
  15. 根据权利要求14所述的电子设备，其中，所述处理器执行所述根据所述待分配订单在所述目标波次的超时概率对所述配送资源进行订单分配，包括：
    比较所述待分配订单在所述多个配送资源的所述目标波次的超时概率；
    将所述待分配订单分配给所述超时概率最低的所述配送资源。
  16. 一种非易失性存储介质，用于存储计算机可读程序，所述计算机可读程序用于供计算机执行如权利要求1至7中任一项所述的订单分配方法。
PCT/CN2019/120838 2018-11-30 2019-11-26 订单分配方法、装置、电子设备和存储介质 WO2020108465A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811458962.6A CN109615201B (zh) 2018-11-30 2018-11-30 订单分配方法、装置、电子设备和存储介质
CN201811458962.6 2018-11-30

Publications (1)

Publication Number Publication Date
WO2020108465A1 true WO2020108465A1 (zh) 2020-06-04

Family

ID=66005217

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/120838 WO2020108465A1 (zh) 2018-11-30 2019-11-26 订单分配方法、装置、电子设备和存储介质

Country Status (2)

Country Link
CN (1) CN109615201B (zh)
WO (1) WO2020108465A1 (zh)

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109615201B (zh) * 2018-11-30 2021-02-09 拉扎斯网络科技（上海）有限公司 订单分配方法、装置、电子设备和存储介质
CN110070289B (zh) * 2019-04-19 2023-03-24 苏州达家迎信息技术有限公司 任务分配方法、装置、设备及存储介质
CN111062553B (zh) * 2019-05-16 2020-12-29 拉扎斯网络科技（上海）有限公司 订单分配方法、装置、服务器和非易失性存储介质
CN111861614A (zh) * 2019-05-28 2020-10-30 北京嘀嘀无限科技发展有限公司 一种订单处理方法、装置、电子设备及存储介质
CN110516872B (zh) * 2019-08-27 2022-10-14 拉扎斯网络科技（上海）有限公司 一种信息处理方法、装置、存储介质和电子设备
CN112561112B (zh) * 2019-09-10 2024-08-27 北京三快在线科技有限公司 订单分配的方法、装置、计算机可读存储介质及电子设备
CN113077299A (zh) * 2020-01-03 2021-07-06 北京三快在线科技有限公司 订单处理方法、装置、设备及存储介质
CN111325503B (zh) * 2020-02-10 2023-09-26 拉扎斯网络科技（上海）有限公司 订单处理方法、装置、服务器和非易失性存储介质
CN111461832A (zh) * 2020-03-31 2020-07-28 拉扎斯网络科技（上海）有限公司 数据处理的方法、装置、可读存储介质和电子设备
CN111756602B (zh) * 2020-06-29 2022-09-27 上海商汤智能科技有限公司 神经网络模型训练中的通信超时检测方法和相关产品
CN112036697B (zh) * 2020-07-28 2024-06-11 拉扎斯网络科技（上海）有限公司 一种任务分配的方法、装置、可读存储介质和电子设备
CN112258129A (zh) * 2020-11-12 2021-01-22 拉扎斯网络科技（上海）有限公司 配送路径预测网络训练、配送资源调度方法及装置
CN113139755A (zh) * 2021-05-17 2021-07-20 拉扎斯网络科技（上海）有限公司 配送任务的调度方法及系统
CN114358694A (zh) * 2022-01-11 2022-04-15 拉扎斯网络科技（上海）有限公司 运单配送处理方法、装置及计算设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107748923A (zh) * 2016-08-29 2018-03-02 北京三快在线科技有限公司 订单处理方法、装置及服务器
CN108364085A (zh) * 2018-01-02 2018-08-03 拉扎斯网络科技（上海）有限公司 一种外卖配送时间预测方法和装置
CN108564269A (zh) * 2018-04-09 2018-09-21 北京小度信息科技有限公司 配送任务分配方法、装置、电子设备及计算机存储介质
CN108734390A (zh) * 2018-05-07 2018-11-02 北京顺丰同城科技有限公司 一种订单配送的调度处理方法及装置
CN109615201A (zh) * 2018-11-30 2019-04-12 拉扎斯网络科技（上海）有限公司 订单分配方法、装置、电子设备和存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107748923A (zh) * 2016-08-29 2018-03-02 北京三快在线科技有限公司 订单处理方法、装置及服务器
CN108364085A (zh) * 2018-01-02 2018-08-03 拉扎斯网络科技（上海）有限公司 一种外卖配送时间预测方法和装置
CN108564269A (zh) * 2018-04-09 2018-09-21 北京小度信息科技有限公司 配送任务分配方法、装置、电子设备及计算机存储介质
CN108734390A (zh) * 2018-05-07 2018-11-02 北京顺丰同城科技有限公司 一种订单配送的调度处理方法及装置
CN109615201A (zh) * 2018-11-30 2019-04-12 拉扎斯网络科技（上海）有限公司 订单分配方法、装置、电子设备和存储介质

Also Published As

Publication number Publication date
CN109615201A (zh) 2019-04-12
CN109615201B (zh) 2021-02-09

Similar Documents

Publication Publication Date Title
WO2020108465A1 (zh) 订单分配方法、装置、电子设备和存储介质
US11053076B1 (en) Carton induction optimization in order fulfillment picking system
Sakhaii et al. A robust optimization approach for an integrated dynamic cellular manufacturing system and production planning with unreliable machines
He et al. Modeling berth allocation and quay crane assignment considering QC driver cost and operating efficiency
CN103745107B (zh) 基于故障模式的装备基层级维修保障仿真系统构建方法
CN109598368B (zh) 路径决策评估、订单分配方法及装置、电子设备及介质
CN103942609B (zh) 产品服务供应链优化设计方法
CN109409739B (zh) 一种基于pomdp模型的众包平台任务分配方法
CN108369675A (zh) 用于案例分配的技术
WO2011105997A1 (en) Simulating supply and demand realization in workforce plan evaluation
Fazli et al. Addressing a coordinated quay crane scheduling and assignment problem by red deer algorithm
Zhao et al. Integrated facility layout design and flow assignment problem under uncertainty
RamachandranPillai et al. Spiking neural firefly optimization scheme for the capacitated dynamic vehicle routing problem with time windows
EP4371008A1 (en) Reducing the environmental impact of distributed computing
Zhao et al. A hierarchical parallel evolutionary algorithm of distributed and multi-threaded two-level structure for multi-satellite task planning
CN117217665A (zh) 智能化库存管理及补货策略生成系统
CN116738239B (zh) 模型训练方法、资源调度方法及装置、系统、设备及介质
CN111144673A (zh) 组织人员结构评估方法、装置、设备及计算机可读介质
Gocgun et al. A Lagrangian approach to dynamic resource allocation
Zaman et al. An efficient methodology for robust assignment problem
CN114819577A (zh) 面向市场需求不确定性的生产计划与调度集成优化方法
Wang et al. A Portrait-Based Method for Constructing Multi-Time Scale Demand Response Resource Pools
CN113807714A (zh) 资源分配的方法、装置、设备、存储介质及程序产品
CN111582408A (zh) 数据处理方法、数据处理装置、存储介质和电子设备
CN113537853A (zh) 订单分配方法、订单分配装置、可读存储介质及电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19891170

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19891170

Country of ref document: EP

Kind code of ref document: A1