CN114066105B - Training method of waybill distribution timeout estimation model, storage medium and electronic equipment - Google Patents

Training method of waybill distribution timeout estimation model, storage medium and electronic equipment Download PDF

Info

Publication number
CN114066105B
CN114066105B CN202210029170.7A CN202210029170A CN114066105B CN 114066105 B CN114066105 B CN 114066105B CN 202210029170 A CN202210029170 A CN 202210029170A CN 114066105 B CN114066105 B CN 114066105B
Authority
CN
China
Prior art keywords
waybill
data
timeout
estimated
value
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN202210029170.7A
Other languages
Chinese (zh)
Other versions
CN114066105A (en
Inventor
周凯荣
朱麟
王鹏宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang Koubei Network Technology Co Ltd
Original Assignee
Zhejiang Koubei Network Technology Co Ltd
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 Zhejiang Koubei Network Technology Co Ltd filed Critical Zhejiang Koubei Network Technology Co Ltd
Priority to CN202210029170.7A priority Critical patent/CN114066105B/en
Publication of CN114066105A publication Critical patent/CN114066105A/en
Application granted granted Critical
Publication of CN114066105B publication Critical patent/CN114066105B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • 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
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • 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

Abstract

The application discloses training method of a waybill distribution timeout estimation model, a storage medium and electronic equipment, wherein the training method comprises the following steps: extracting characteristic data in the collected waybill sample data; determining a real label value corresponding to the sample data according to a defined label value of the estimated waybill distribution overtime level; inputting the real label value and the characteristic data serving as a sample data set into a constructed waybill distribution overtime estimation model for training to obtain an estimated value of the waybill sample data overtime; calculating the loss between the estimated value and the real label value according to a first loss function and a second loss function to obtain a loss parameter value; adjusting the waybill distribution timeout pre-estimation model according to the loss parameter value, and determining the adjusted waybill distribution timeout pre-estimation model as a target waybill distribution timeout pre-estimation model; therefore, the estimated overtime risk is more accurate when the trained model is used subsequently.

Description

Training method of waybill distribution timeout estimation model, storage medium and electronic equipment
Technical Field
The application relates to the technical field of computer application, in particular to a training method and a device of a waybill delivery timeout estimation model. The application also relates to a method and a device for estimating the delivery timeout of the waybill. The application also relates to a waybill scheduling method and device. The application also relates to a computer storage medium and an electronic device.
Background
With the continuous development of internet technology and computer technology, the realization of goods purchase through a network application platform has become a daily life style. Generally, a rough process of online shopping includes logging in an application service platform provided on an electronic device, selecting a required article from the application service platform, and then performing online payment or selective delivery payment of the corresponding article. At this time, the order is formed and transferred to a subsequent logistics distribution stage, which may generally determine a distribution duration according to one or more combinations of the attributes of the item, the distribution distance of the item, the location of the distributor, the order receiving time of the distributor, and the like, and provide the distribution time to the user for the user to refer.
Disclosure of Invention
The application provides a training method of a waybill delivery overtime estimation model, which aims to solve the problems that delivery overtime estimation is inaccurate in the prior art, and the subsequent waybill delivery efficiency is reduced.
The application provides a training method of a waybill distribution timeout estimation model, which comprises the following steps:
extracting characteristic data in the collected waybill sample data;
determining a real label value corresponding to the sample data according to a defined label value of the estimated waybill distribution overtime level;
inputting the real label value and the characteristic data as a sample data set into a constructed waybill distribution timeout estimation model for training to obtain a predicted value of the waybill sample data timeout;
calculating the loss between the estimated value and the real label value according to a first loss function and a second loss function to obtain a loss parameter value;
and adjusting the waybill distribution timeout pre-estimation model according to the loss parameter value, and determining the adjusted waybill distribution timeout pre-estimation model as a target waybill distribution timeout pre-estimation model.
In some embodiments, the extracting feature data in the collected waybill sample data includes:
collecting the waybill sample data;
and extracting the characteristic data according to the waybill sample data.
In some embodiments, said collecting said waybill sample data comprises:
determining the collection range of the waybill sample data according to the time from the creation of the waybill to the time from the receiving of the waybill by a distributor;
and collecting the waybill sample data in the collection range according to the set collection time.
In some embodiments, said extracting said feature data according to said waybill sample data comprises:
and extracting the feature data from the waybill sample data according to the set extraction dimension.
In some embodiments, the extracting the feature data from the waybill sample data according to the set extraction dimension includes at least one of:
extracting at least one characteristic data of the unit price of passengers, the distribution distance, the number of resource objects and the duration time of the waybill corresponding to the waybill dimension from the waybill sample data according to the waybill dimension;
according to the dimension of a resource object provider, extracting at least one characteristic data of the real-time unfinished quantity of the resource objects, the finished quantity of the resource objects in historical time, the quantity of orders in historical time and the delivery overtime risk rate in historical time, which correspond to the resource object provider, from the waybill sample data;
extracting at least one characteristic data of the type of a freight line, the number of distribution parties, the number of unfinished distribution freight notes, a distribution pressure value and a distribution overtime risk rate in historical time corresponding to the freight waybill sample data according to the freight waybill dimension;
and extracting at least one characteristic data of a weather grade, a distribution path road condition and a distribution time point corresponding to the environmental dimension from the waybill sample data according to the distribution environmental dimension.
In some embodiments, said determining a true tag value corresponding to said sample data according to a tag defining a pre-estimated waybill distribution timeout level comprises:
defining a label value of the estimated waybill distribution overtime level;
and determining the real label value corresponding to the sample data according to the label value.
In some embodiments, said calculating the loss between said estimated value and said real tag according to a first loss function and a second loss function to obtain a loss parameter value comprises:
calculating a mean square error loss function according to the real label value and the estimated value;
calculating a sorting loss function according to the real label value and the estimated value;
calculating the loss parameter value using the mean square error loss function as the first loss function and the ordering loss function as the second loss function.
In some embodiments, said calculating said loss parameter value using said mean square error loss function as said first loss function and said ordering loss function as said second loss function comprises:
and determining the sum of the quotient of the mean square error loss function and the set reference value and the sorting loss function as the loss parameter value.
The application also provides a training device of the waybill delivery overtime estimation model, which comprises:
the extraction unit is used for extracting characteristic data in the collected waybill sample data;
the label determining unit is used for determining a real label value corresponding to the sample data according to the defined label value of the estimated waybill delivery overtime level;
the training unit is used for inputting the real label value and the characteristic data as a sample data set into a constructed waybill distribution timeout estimation model for training to obtain a predicted value of the waybill sample data timeout;
an obtaining unit, configured to calculate a loss between the predicted value and the true tag value according to a first loss function and a second loss function, and obtain a loss parameter value;
and the target determining unit is used for adjusting the waybill distribution timeout estimation model according to the loss parameter value and determining the adjusted waybill distribution timeout estimation model as a target waybill distribution timeout estimation model.
The application also provides a method for pre-estimating delivery timeout of the waybill, which comprises the following steps:
inputting waybill data to be estimated into a waybill distribution timeout estimation model to obtain a timeout estimated value of the waybill data to be estimated;
determining a called mapping relation table according to attribute information in the waybill data to be estimated; the mapping relation table is used for describing a corresponding relation between an actual sample estimated value of actual waybill sample data of the waybill distribution timeout estimation model and an actual timeout risk rate;
determining data sub-buckets in the mapping relation table corresponding to the timeout estimated value;
and determining the actual overtime risk data corresponding to the data sub-buckets as target overtime risk data of the waybill to be estimated.
In some embodiments, further comprising:
selecting actual waybill transport order sample data according to a data generation area range and/or a time range of the actual waybill transport order sample data determined by waybill transport order attributes of waybill transport order data to be estimated;
inputting the actual waybill sample data into the waybill distribution timeout estimation model to obtain the actual sample estimated value;
dividing the actual waybill sample data according to the data sub-buckets according to the sequencing of the actual sample pre-estimated values;
acquiring the actual sample estimated value interval range of the data sub-bucket;
determining the actual overtime risk rate of the data sub-bucket according to the actual sample estimated value interval;
and establishing the mapping relation table according to the corresponding relation between the actual sample pre-evaluation value interval and the actual overtime risk rate.
In some embodiments, the determining the actual timeout risk rate waybill for the data sub-bucket according to the actual sample pre-evaluation interval includes:
and determining the actual overtime risk rate according to the sum of the selected actual sample estimated values and the total singular number of the freight note in the selected range when the actual sample estimated values are selected.
In some embodiments, the determining a called mapping relationship table according to the attribute information in the waybill data to be estimated includes:
calling the mapping relation table matched with the waybill generation area according to the waybill generation area in the waybill data to be estimated; or calling the mapping relation table matched with the waybill creation time information according to the waybill creation time in the waybill data to be estimated.
In some embodiments, the determining the data sub-bucket in the mapping relation table corresponding to the timeout estimation value includes:
determining a target actual sample estimated value interval according to the actual sample estimated value interval in which the overtime estimated value falls;
and determining the target actual sample estimated value interval as a data sub-bucket in the mapping relation table corresponding to the overtime estimated value.
In some embodiments, further comprising:
and outputting the target overtime risk data.
The application also provides a device is estimated to too time-out of waybill delivery, includes:
the acquisition unit is used for inputting the waybill data to be estimated into the waybill distribution timeout estimation model and acquiring the timeout estimated value of the waybill data to be estimated;
the first determining unit is used for determining the called mapping relation table according to the attribute information in the waybill data to be estimated; the mapping relation table is used for describing a corresponding relation between an actual sample estimated value of actual waybill sample data of the waybill distribution timeout estimation model and an actual timeout risk rate;
a second determining unit, configured to determine data buckets in the mapping relation table corresponding to the timeout estimated value;
and the third determining unit is used for determining the actual overtime risk data corresponding to the data sub-bucket as the target overtime risk data of the waybill to be estimated.
The application also provides a waybill scheduling method, which comprises the following steps:
acquiring waybill information to be estimated, inputting the waybill information to be estimated into a waybill distribution overtime estimation model, and determining target overtime risk data of the waybill to be estimated;
arranging the freight notes to be estimated in a descending order according to the risk data values in the target overtime risk data;
and taking the arranged freight notes to be estimated as freight notes to be scheduled, and scheduling according to the arrangement sequence.
The present application further provides a waybill scheduling device, including:
the acquisition unit is used for acquiring the waybill information to be estimated, inputting the waybill information to be estimated into the waybill distribution overtime estimation model, and determining the target overtime risk data of the waybill to be estimated;
the sorting unit is used for sorting the waybills to be pre-estimated in a descending order according to the risk data values in the target overtime risk data;
and the scheduling unit is used for taking the arranged freight notes to be estimated as freight notes to be scheduled and scheduling according to the arrangement sequence.
The application also provides a computer storage medium for storing the data generated by the network platform and a program for processing the data generated by the network platform;
when the program is read and executed by the processor, the method for training the waybill distribution timeout estimation model is executed; or, executing the estimation method of waybill distribution timeout; alternatively, the steps of the waybill scheduling method as described above are performed.
The present application further provides an electronic device, comprising:
a processor;
a memory for storing a program for processing data generated by the network platform, wherein the program, when being read and executed by the processor, executes the steps of the training method of the waybill distribution timeout estimation model; or, executing the estimation method of waybill distribution timeout; alternatively, the steps of the waybill scheduling method as described above are performed.
Compared with the prior art, the method has the following advantages:
according to the training method embodiment of the waybill delivery overtime estimation model, optimization updating of the model is achieved by fusing the first loss function and the second loss function, estimated overtime risks of the trained model are more accurate when the trained model is used subsequently, stable and reliable overtime risk reference data are provided for a scheduling system, efficiency and quality of scheduling work are improved beneficially, and in addition, stability of the waybill delivery overtime estimation model under a scheduling application scene is better due to the two loss functions.
The estimation method for the waybill distribution timeout provided by the application obtains the timeout estimated value of the waybill data to be estimated by inputting the waybill data to be estimated into a waybill distribution timeout estimation model, determining a called mapping relation table according to attribute information in the waybill data to be estimated, determining data sub-buckets in the mapping relation table corresponding to the timeout estimated value, determining the actual timeout risk data corresponding to the data sub-buckets as target timeout risk data of the waybill to be estimated, thereby not only predicting the overtime risk of the waybill but also calibrating the data of the estimated value under the service scene, so that the overtime risk rate of the waybill can be determined according to the estimated value, reference is provided for subsequent scheduling, and the waybill flow is adjusted to pass the reference without being influenced by the replacement of the estimation model.
Drawings
FIG. 1 is a flowchart of an embodiment of a training method for a waybill distribution timeout estimation model provided in the present application;
FIG. 2 is a structural diagram of an embodiment of a training apparatus for a waybill distribution timeout estimation model provided in the present application;
FIG. 3 is a flowchart of an embodiment of a method for estimating waybill dispatch timeout provided herein;
FIG. 4 is a schematic diagram illustrating estimated value display divisions of different estimated models in an embodiment of a method for estimating waybill distribution timeout according to the present application;
FIG. 5 is a schematic structural diagram of an embodiment of an estimation apparatus for waybill distribution timeout provided in the present application;
FIG. 6 is a flowchart of an embodiment of a waybill scheduling method provided herein;
fig. 7 is a schematic structural diagram of an embodiment of an waybill scheduling device provided in the present application;
fig. 8 is a schematic structural diagram of an embodiment of an electronic device provided in the present application.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. This application is capable of implementation in many different ways than those herein set forth and of similar import by those skilled in the art without departing from the spirit of this application and is therefore not limited to the specific implementations disclosed below.
The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The description used in this application and in the appended claims is for example: the terms "a," "an," "first," and "second," etc., are not intended to be limiting in number or order, but rather are used to distinguish one type of information from another.
With reference to the above description of the background art, it is known that the inventive concept of the present application is derived from the problem in the aspect of online shopping application, and specifically, the following description is made with reference to the application scenario:
in the takeaway scenario, the waybill scheduling system will make a judgment on the waybill (i.e. the back order) that the rider is responsible for delivery and the waybill to be scheduled and assigned to the rider during the assignment of the waybill scheduling, for example: the current overtime risk of riding a back order is high; when the current waybill in the distribution stage has higher overtime risk, the algorithm for scheduling and assigning the waybill can adjust the scheduling and assigning of the waybill in a targeted manner, so that the proportion of the whole overtime waybill is reduced. However, when the scheduling system determines the overtime risk of the waybill, if the estimation of the overtime risk is inaccurate, the scheduling assignment of the whole waybill scheduling chain is unstable, so that the scheduling system has the problems of low efficiency and inaccurate waybill assignment.
The same problem still exists in the logistics distribution scenario in fact, particularly in the specific subject scenario, the rapid increase of the logistics distribution amount requires the advance and accurate estimation of the distribution timeout condition, so that the scheduling system can adaptively adjust the distribution of logistics distribution according to the estimated timeout.
In both take-away and logistics distribution scenarios, accurate estimation of the timeout risk of the waybill needs to be assumed, so as to improve the efficiency of the dispatch system in dispatching the waybill. The process of delivering an article from one location to another location can be the source of the inventive concept of the present application, for example: besides the takeaway scene and the logistics distribution scene, the method can also be used for predicting overtime of vehicle arrival, so that the travel route can be timely adjusted according to the overtime prediction, or the travel route and/or the travel mode can be timely adjusted according to the overtime prediction (such as navigation) from the departure place to the destination, and the like.
Therefore, the invention concept of the training method of the waybill distribution timeout estimation model provided by the application is derived from a takeout application scenario, but the application scenario with the same technical problem is not limited to the takeout application scenario. Any problem that the prediction of the timeout risk is inaccurate is applicable to the inventive concept of the present application. The following describes in detail a training method of a waybill distribution timeout estimation model provided in the present application.
As shown in fig. 1, fig. 1 is a flowchart of an embodiment of a training method for a waybill distribution timeout estimation model provided in the present application, where the embodiment of the training method includes:
step S101: extracting characteristic data in the collected waybill sample data;
the waybill in step S101 may be data information of delivery service for an order generated by the takeout application service platform; or data information generated by other application service platforms with delivery service requirements. In this embodiment, a local life service application or a takeaway application service is mainly used as an example of an application scenario for explanation.
The step S101 is to extract feature data of waybill sample data, and the specific implementation process may include:
step S101-1: collecting the waybill sample data;
step S101-2: and extracting the characteristic data according to the waybill sample data.
In this embodiment, the collection of the waybill sample data in the step S101-1 may be to collect waybill sample data in a whole area of the country, certainly, for predicting the waybill timeout risk in different regions, the waybill sample data in a corresponding country may be collected, or an area range for collecting the waybill sample data may be set according to a requirement, and a specific range for collecting the waybill sample data is not limited. The specific implementation process of the acquisition in the step can comprise the following steps:
step S101-11: determining the collection range of the waybill sample data according to the time from the creation of the waybill to the time from the receiving of the waybill by a distributor;
step S101-12: and collecting the waybill sample data in the collection range according to the set collection time.
In this embodiment, the combination of the steps S101-11 and the takeaway application scenario may be understood as that the waybill data in the time period from creation of the waybill to the time period before the rider (distributor) takes the waybill is the collection range of the waybill sample data. The step S101-12 may be configured to collect one waybill sample data in the collection range every minute. That is, the sample data of the constructed waybill can be the sample data of the waybill collected every minute from the creation of the waybill to the time point of the receipt of the dispatched party. The set acquisition time can be set according to the scheduling requirements of the waybill, for example: if the waybill schedule is a waybill timeout per minute query, it may be set to sample once a minute, and may sample at other time lengths if the query is other time lengths. Of course, the query according to the schedule is only one implementation manner, and other manners may also be included, for example: the collection time is set according to the number of freight notes, the number of distribution parties and the like.
The specific implementation process of the step S101-2 may include:
step S101-21: and extracting the feature data from the waybill sample data according to the set extraction dimension. Wherein the extraction dimension comprises at least one of: waybill dimensions, resource object provider dimensions, capacity dimensions, distribution environment dimensions, and the like. The specific implementation process of step S101821 may include at least one of the following:
extracting at least one characteristic data of the unit price of the client, the distribution distance, the number of resource objects, the duration time of the waybill and the like corresponding to the waybill dimension from the waybill sample data according to the waybill dimension; wherein, the unit price of the guest can be understood as the data of the amount of the commodity which is averagely purchased by the user; the distribution distance can be understood as the distribution distance of a distributor for distributing the freight notes; the quantity of the resource objects can be understood as the quantity of commodities in the waybill, that is, the resource objects can be commodities, dishes, or the like, for example: in the take-out scene, the number of dishes in the order may be determined, it should be noted that the waybill and the order are located in different process stages, the displayed commodity information is the same, and the waybill may further include distribution side information and the like with respect to the order. The relevant information of the waybill and the order is not described herein. The waybill duration time can be understood as the time length after the waybill is created, namely the time length determined by subtracting the creation time point of the waybill from the sampling time point of the waybill. That is, at least one kind of characteristic data of the unit price of the guest, the delivery distance, the number of resource objects, the elapsed time of the unit, and the like is extracted in the unit dimension.
According to the dimension of a resource object provider, extracting at least one of the characteristic data of the real-time unfinished quantity of the resource objects, the completion duration of the resource objects in the historical time, the quantity of orders in the historical time, the delivery overtime risk rate in the historical time and the like corresponding to the resource object provider from the waybill sample data; wherein the resource object provider dimension may be understood as a merchant or store dimension, i.e. the resource object provider is a merchant or store. Taking a takeout application scenario as an example, the real-time unfinished quantity of the resource objects can be understood as the real-time unfinished order quantity of merchants; the completion duration of the resource object in the historical time can be understood as the meal serving duration of a merchant within 7 days in the historical time; the order quantity in the historical time can be understood as the order quantity in 7 days in the history; the delivery timeout risk rate over the historical time may be understood as the sum of the values of the historical 7-day timeout tags divided by the number of waybills over the historical 7-day period.
Extracting at least one characteristic data of the type of a freight transportation line, the number of delivery parties, the number of unfinished delivery freight notes, a delivery pressure value, a delivery overtime risk rate in historical time and the like corresponding to the freight transportation dimension from the freight transportation sample data according to the freight transportation dimension; in this embodiment, the type of the power line may be the type of the distributor, for example: crowd-sourced delivery or team delivery, etc. The number of the distribution parties can be understood as the number of the distribution parties in the business circle range of the waybill; the number of the unfinished delivery waybills can be understood as the number of the unfinished delivery waybills in the range of the business district; the distribution pressure value can be understood as the number of unfinished distribution waybills in the business district range divided by the number of distributors in the business district; the delivery timeout risk rate in the historical time may be understood as the sum of the values of the timeout tags in the 7 days of the business district divided by the number of waybills in the 7 days of the business district history.
And extracting at least one characteristic data of a weather grade, a distribution path road condition, a distribution time point and the like corresponding to the environmental dimension from the waybill sample data according to the distribution environmental dimension.
In the embodiment, the dimensionality is provided, the dimensionality of a distribution party can be added according to actual requirements, the selection of specific dimensionality is not limited, and the dimensionality of sample data related to estimation of the waybill timeout range can be met.
The above is a description of the specific implementation process of step S101 in the takeout application scenario provided in this embodiment.
Step S102: determining a real label value corresponding to the sample data according to a defined label value of the estimated waybill distribution overtime level;
the purpose of step S102 is to determine a real tag value corresponding to the sample data.
The specific implementation process of step S102 may include:
step S102-1: defining a label value of the estimated waybill distribution overtime level;
step S102-2: and determining the real label value corresponding to the sample data according to the label value.
The tag value in step S102-1 is used to describe a timeout estimate value of the waybill, and the timeout estimate value can reflect a timeout risk, that is, the timeout estimate value and the timeout risk may correspond to each other.
In this embodiment, the defined label values of the estimated waybill distribution timeout levels may be as shown in table 1 below:
table 1:
Figure GDA0003603493470000101
from table 1 above, it can be seen that the larger the value of the Label (Label) is, the larger the risk of timeout is. In this embodiment, the user T represents the expected delivery duration of the committed user. As can be seen from table 1, the label value for the estimated waybill delivery timeout level relates not only to time but also to the case of canceling an order, which may include at least two of user active cancellation and system cancellation.
In step S102-2, the real tag value may be determined according to the actual delivery time data of the sample data, for example: if the delivery time of the sample data A is within the expected delivery time range of the promising user, the sample data A corresponds to a label value of 0; and if the delivery time exceeds the expected delivery time by 15-20 min, the corresponding label value is 4, and the corresponding label value is the real label value.
In the step S102, the real tag value corresponding to the sample data is determined by the defined tag predicting the waybill distribution overtime level, that is, the overtime level of the sample data can be obtained, and the definition of the tag value shows that the greater the tag value, the more the overtime is.
Step S103: inputting the real label value and the characteristic data as a sample data set into a constructed waybill distribution timeout estimation model for training to obtain a predicted value of the waybill sample data timeout;
the waybill distribution timeout estimation model in the step S103 may be a neural network including a 5-layer structure. The first layer is an input layer, and the number of nodes can be adjusted according to the characteristic data, so that the number of the nodes is matched with the number of the characteristic data. The second layer is a hidden layer and may include 512 nodes. The third layer is an implied layer and can comprise 256 nodes. The fourth layer is a hidden layer and may include 128 nodes. The fifth layer is an output layer and may include 1 node.
In order to avoid overfitting of the schedule delivery timeout estimation model, in this embodiment, a Batch norm structural layer (also referred to as a BN layer, that is, a Batch Normalization structural layer) is added to any at least one layer from the first layer to the fifth layer of the schedule delivery timeout estimation model, and parameters are initialized randomly, so as to avoid overfitting of the model. In this embodiment, a BatchNorm structure may be added to each layer, so that the inputs of each layer of the neural network can be kept the same during the model training process.
The purpose of step S103 is to obtain an estimated value of the waybill sample data timeout through training of the waybill distribution timeout estimation model. In order to determine the difference between the estimated value and the real tag value, a loss function is required to determine, the larger the value of the loss function is, the larger the difference between the estimated value and the real tag value is, and the more inaccurate the estimation of the estimated value is, so that the difference between the estimated value and the real tag value is determined by using at least two loss functions in the embodiment, so that the waybill distribution timeout estimation model can perform parameter adjustment according to the loss function, the estimated value output by the waybill distribution timeout estimation model is closer to the real tag value corresponding to the waybill sample data, and the output effect of the waybill distribution timeout estimation model is improved. Therefore, it can be realized by executing the contents of step S104.
Step S104: calculating the loss between the estimated value and the real label value according to a first loss function and a second loss function to obtain a loss parameter value;
the first loss function in the step S104 may be a mean square error loss function (mse loss); the second loss function may be a rank loss function (rank loss). The specific implementation process of step S104 may include:
step S104-1: calculating a mean square error loss function according to the real label value and the estimated value;
step S104-2: calculating a ranking loss function according to the real label value and the estimated value;
step S104-3: calculating the loss parameter value using the mean square error loss function as the first loss function and the ordering loss function as the second loss function.
The calculation formula of the mean square error loss function in step S104-1 may be as follows:
Figure GDA0003603493470000121
the calculation formula of the ranking loss function in step S104-2 may be as follows:
Figure GDA0003603493470000122
in one of the above-mentioned two formulas,
Figure GDA0003603493470000123
as authentic labelsThe value of the one or more of the one,
Figure GDA0003603493470000124
is an estimated value.
The loss parameter value (loss) may be determined as a sum of a quotient of the mean square error loss function (mse _ loss) and a set reference value and the rank loss function (rank _ loss), and may be specifically calculated as:
loss＝rank_loss+mse_loss/M；
the reference value M may be 4000 in this embodiment, and the reference value M may be a measured reference value or an empirical value predicted from previous data.
Step S105: and adjusting the waybill distribution timeout pre-estimation model according to the loss parameter value, and determining the adjusted waybill distribution timeout pre-estimation model as a target waybill distribution timeout pre-estimation model.
The specific implementation process of step S105 may be to adjust and update parameters of the waybill delivery timeout estimation model according to the loss parameter values, so as to obtain a target waybill delivery timeout estimation model, so that a more accurate estimated value can be obtained after waybill data is input into the target waybill delivery timeout estimation model, so that a subsequent scheduling system performs scheduling assignment on the waybill, and the efficiency of the scheduling system is improved. When the loss parameter value tends to 0, the estimated value is close to the real label value, the estimation of the estimated value is accurate, and the model training is finished.
According to the training method for the waybill delivery timeout estimation model, the first loss function and the second loss function are fused, optimization updating of the model is achieved, estimated timeout risks of the trained model are more accurate when the trained model is used subsequently, stable and reliable timeout risk reference data are not provided by a scheduling system, and efficiency and quality of scheduling work are improved.
The above is a detailed description of an embodiment of a training method for a waybill distribution timeout estimation model provided in the present application, and corresponds to the aforementioned embodiment of the training method for a waybill distribution timeout estimation model, the present application further discloses an embodiment of a training apparatus for a waybill distribution timeout estimation model, please refer to fig. 2, since the apparatus embodiment is basically similar to the method embodiment, the description is simpler, and related points can be found in part of the description of the method embodiment. The device embodiments described below are merely illustrative.
As shown in fig. 2, fig. 2 is a structural diagram of an embodiment of a training apparatus for a waybill distribution timeout estimation model provided in the present application, where the embodiment of the training apparatus includes:
the extraction unit 201 is configured to extract feature data in the collected waybill sample data;
a tag determining unit 202, configured to determine, according to a defined tag value of the estimated waybill distribution timeout level, a real tag value corresponding to the sample data;
the training unit 203 is configured to input the real label value and the feature data as a sample data set to a constructed waybill distribution timeout estimation model for training, and obtain a predicted value of the waybill sample data timeout;
an obtaining unit 204, configured to calculate a loss between the predicted value and the real tag value according to a first loss function and a second loss function, and obtain a loss parameter value;
a target determining unit 205, configured to adjust the waybill distribution timeout estimation model according to the loss parameter value, and determine the adjusted waybill distribution timeout estimation model as a target waybill distribution timeout estimation model.
In this embodiment, the extracting unit 201 may include: a collecting subunit and an extracting subunit; the acquisition subunit is used for acquiring the waybill sample data. The extraction subunit is configured to extract the feature data according to the waybill sample data. Wherein the acquisition subunit may include: the range determining subunit is used for determining the acquisition range of the waybill sample data according to the time from the creation of the waybill to the time from the delivery party to receive the waybill; the collection subunit is specifically configured to collect the waybill sample data within the collection range according to set collection time. The extraction subunit may be specifically configured to extract the feature data from the waybill sample data according to a set extraction dimension.
In this embodiment, the extracting the sub-unit may include extracting feature data from at least one of the following dimensions:
extracting at least one characteristic data of the unit price of passengers, the distribution distance, the number of resource objects and the duration time of the waybill corresponding to the waybill dimension from the waybill sample data according to the waybill dimension;
extracting at least one characteristic data of the real-time unfinished quantity of the resource objects, the finished quantity of the resource objects in historical time, the quantity of orders in historical time and the delivery overtime risk rate in historical time corresponding to the resource object provider from the waybill sample data according to the dimension of the resource object provider;
extracting at least one characteristic data of the type of the freight line, the number of delivery parties, the number of unfinished delivery freight notes, a delivery pressure value and a delivery overtime risk rate in historical time corresponding to the freight note dimension from the freight note sample data according to the freight dimension;
and extracting at least one characteristic data of the weather grade, the distribution path road condition and the distribution time point corresponding to the environment dimension from the waybill sample data according to the distribution environment dimension.
For specific content of the extracting unit 201, reference may be made to the description of step S101 in the above training method embodiment, and details are not repeated here.
In this embodiment, the tag determining unit 202 may include: the system comprises a definition subunit and a determination subunit, wherein the definition subunit is used for defining the label value of the estimated waybill distribution overtime level; the determining subunit is configured to determine, according to the tag value, the real tag value corresponding to the sample data.
For the specific implementation process of the label determination unit 202, reference may be made to the specific content of step S102 in the above training method embodiment, and details are not repeated here.
In this embodiment, the waybill distribution timeout estimation model in the training unit 203 may be a neural network including a 5-layer structure, the first layer is an input layer, and the number of nodes may be adjusted according to the feature data, so that the number of nodes matches the number of the feature data. The second layer is a hidden layer and may include 512 nodes. The third layer is an implied layer and can comprise 256 nodes. The fourth layer is a hidden layer and may include 128 nodes. The fifth layer is an output layer and may include 1 node. In order to avoid overfitting of the waybill distribution timeout estimation model, in this embodiment, a Batch norm structural layer (also referred to as a BN layer, that is, a Batch Normalization structural layer) is added to any one of the first layer to the fifth layer of the waybill distribution timeout estimation model, and parameters are initialized randomly, so as to avoid overfitting of the model. In this embodiment, a BatchNorm structure may be added to each layer, so that the inputs of each layer of the neural network are kept distributed the same during the model training process.
For the specific implementation process of the training unit 203, reference may be made to step S203 in the above training method embodiment.
In this embodiment, the obtaining unit 204 may include: the system comprises a first calculation subunit, a second calculation subunit and a third calculation subunit; the first calculating subunit is configured to calculate a mean square error loss function according to the real tag value and the estimated value; the second calculating subunit is configured to calculate a sorting loss function according to the real tag value and the estimated value; the third calculation subunit is configured to calculate the loss parameter value using the mean square error loss function as the first loss function and using the ordering loss function as the second loss function. The third calculation subunit may be specifically configured to determine, as the loss parameter value, a sum of a quotient of the mean square error loss function and a set reference value and the ranking loss function.
For a specific implementation process of the obtaining unit 204, reference may be made to step S104 in the above training method embodiment, and details are not repeated here.
In this embodiment, the specific implementation process of the target determining unit 205 may be that parameters of the waybill delivery timeout estimation model are adjusted and updated according to the loss parameter values, so as to obtain a target waybill delivery timeout estimation model, so that a more accurate estimation value can be obtained after waybill data is input into the target waybill delivery timeout estimation model, so that a subsequent scheduling system performs scheduling assignment on the waybill, and the efficiency of the scheduling system is improved.
The above is a description of an embodiment of the training apparatus of the waybill distribution timeout estimation model provided in the present application, and for a specific implementation process of the embodiment of the training apparatus, reference may be made to the description of the above step S101 to step S105, and details of the same or repeated contents are not repeated here.
Based on the above, the present application further provides an estimation method for waybill distribution timeout, as shown in fig. 3, fig. 3 is a flowchart of an embodiment of the estimation method for waybill distribution timeout provided in the present application; the embodiment of the estimation method comprises the following steps:
step S301: inputting waybill data to be estimated into a waybill distribution timeout estimation model to obtain a timeout estimated value of the waybill data to be estimated; in this embodiment, the estimated waybill data may be a food delivery waybill or a drug delivery waybill generated in a takeaway application scenario, which certainly does not exclude the delivery waybill of the life service goods.
The waybill distribution timeout estimation model in step S301 needs to be trained in advance, and the training method may adopt the descriptions of step S101 to step S105.
Step S302: determining a called mapping relation table according to attribute information in the waybill data to be estimated; the mapping relation table is used for describing a corresponding relation between an actual sample estimated value of actual waybill sample data of the waybill distribution timeout estimation model and an actual timeout risk rate;
as shown in fig. 4, when executing step S302, the mapping relationship table needs to be established, which may specifically include:
step S302-1: determining a data generation area range and/or a time range of the actual waybill sample data based on waybill attributes of waybill data to be estimated, and selecting the actual waybill sample data; the time range may be a preset time range for selecting actual waybill sample data, or may be the acquisition time of the actual waybill sample data, or may be the creation time of the actual waybill sample data, and the like, and in this embodiment, the actual waybill sample data within 3 days of the history is used. The generation area range may be a region location to which actual waybill sample data belongs, for example: the city, or the business circle of the city, etc., may be the city range or a certain region range of the city.
Step S302-2: inputting the actual waybill sample data into the waybill distribution timeout estimation model to obtain the actual sample estimated value;
step S302-3: dividing the actual waybill sample data according to the data sub-buckets according to the sequencing of the actual sample pre-estimated values; wherein the ordering may be in a descending order or in an ascending order;
step S302-4: acquiring the actual sample estimated value interval range of the data sub-bucket;
step S302-5: determining the actual overtime risk rate of the data sub-bucket according to the actual sample estimated value interval;
step S302-6: and establishing the mapping relation table according to the corresponding relation between the sample estimated value interval and the actual overtime risk rate.
In order to facilitate obtaining the corresponding actual timeout risk value meaning through the actual sample pre-estimated value interval, the embodiment uses a mapping relationship table, where the mapping relationship table at least includes the actual sample pre-estimated value, the actual timeout risk value, and the corresponding relationship between the actual sample pre-estimated value and the actual timeout risk value, as shown in the following table:
Figure GDA0003603493470000161
the mapping relation table can be used for calibrating the timeout estimated value, so that the output value of the timeout estimated model delivered through the waybill has waybill meaning and is not influenced by model replacement.
Since the waybill data related to different application scenarios are different, and therefore, the expression manner of the estimated value is also different, the timeout estimated value may be numerically calibrated according to the waybill data form related to the waybill distribution timeout application scenario, and when the waybill data form related to the waybill distribution timeout application scenario is the number of the timeout waybill, the determining the timeout risk rate corresponding to the sample estimated value may include:
and determining the actual overtime risk rate according to the sum of the selected actual sample estimated values and the total singular number of the freight note in the selected range when the actual sample estimated values are selected.
For example: in the waybill data representation of the takeaway scenario, if the estimated value is 1, 10, 100, etc., it is generally understood from the magnitude of the value that 100 is much larger than 10, and the difference between 10 and 1 is smaller than the difference between 10 and 10, however, the true label corresponding to the estimated value may be 1, 2, 3, and therefore, the estimated value needs to be calibrated or converted, that is, the estimated value is converted into a risk probability value, for example: the estimated values 0, 1, 10, 100 may be calibrated or converted to 0, 0.1, 0.2, 0.4, i.e.: the timeout risk (the estimated value is 1) is 0.1, which means that 1 of 10 waybills becomes a long waybill, and the timeout risk (the estimated value is 10) is 0.2, which means that 2 of 10 waybills become a timeout waybill. Namely: the timeout risk factor is the sum/run number of the sample estimated value. In this embodiment, a corresponding relationship is established between the actual sample estimated value and the actual timeout risk rate, thereby completing the calibration of the numerical value.
As shown in fig. 4, in order to improve the display resolution of the target timeout risk data, the actual sample prediction value may be divided into 24 data sub-buckets from low to high, and of course, the data sub-buckets herein are only one expression form and are not used to limit the way of data division. The label (label) in the data sub-buckets is averaged, so that compared with the existing mse _ loss model and rank _ loss model, a rank + mse loss model (waybill distribution timeout estimation model) has more obvious high and low gradients, the abscissa in fig. 4 represents the number of the data sub-buckets (namely 24 data sub-buckets), the ordinate represents the timeout estimated value, the smaller the number of the data sub-buckets is, the lower the label mean value is, the larger the number of the data sub-buckets is, the higher the label mean value is, the more obvious discrimination is shown, and the actual timeout risk rate of actual sample data can be effectively discriminated.
Step S303: determining data sub-buckets in the mapping relation table corresponding to the timeout estimated value;
step S304: and determining the actual overtime risk data corresponding to the data sub-buckets as target overtime risk data of the waybill to be estimated.
The step S303 may include:
step S303-1: determining a target actual sample estimated value interval according to the actual sample estimated value interval in which the overtime estimated value falls;
step S303-2: and determining the target actual sample estimated value interval as a data sub-bucket in the mapping relation table corresponding to the overtime estimated value.
In this embodiment, the method may further include: and outputting the target overtime risk data. The output can be output to a display platform of the dispatching system or an electronic device of a distribution party. The specific output display form is not limited, and the overtime risk of the waybill can be known visually and conveniently.
The output form of the target timeout risk data may also include other forms, and this embodiment has been referred to fig. 4.
The foregoing is a detailed description of an embodiment of a method for estimating waybill distribution timeout provided in the present application, which corresponds to the foregoing embodiment of a method for estimating waybill distribution timeout provided in the present application, and the present application further discloses an embodiment of an apparatus for estimating waybill distribution timeout, please refer to fig. 5, since the apparatus embodiment is basically similar to the method embodiment, the description is relatively simple, and related points can be found in the partial description of the method embodiment. The device embodiments described below are merely illustrative.
As shown in fig. 5, the estimation apparatus embodiment includes:
an obtaining unit 501, configured to input waybill data to be estimated into a waybill distribution timeout estimation model, and obtain a timeout estimated value of the waybill data to be estimated;
a first determining unit 502, configured to determine a called mapping relationship table according to attribute information in the waybill data to be estimated; the mapping relation table is used for describing a corresponding relation between an actual sample estimated value of actual waybill sample data of the waybill distribution timeout estimation model and an actual timeout risk rate;
a second determining unit 503, configured to determine data buckets in the mapping table corresponding to the timeout estimated value;
a third determining unit 504, configured to determine the actual timeout risk data corresponding to the data sub-buckets as the target timeout risk data of the waybill to be pre-estimated. Further comprising:
the selection unit is used for selecting the actual waybill sample data according to the data generation area range and/or the time range of the actual waybill sample data determined by the waybill attribute of the waybill data to be estimated;
a first obtaining unit, configured to input the actual waybill distribution timeout estimation model with the actual waybill sample data, and obtain a predicted value of the actual sample;
the dividing unit is used for dividing the actual waybill sample data into buckets according to the sequence of the actual sample estimated values;
the second obtaining unit is used for obtaining the actual sample estimated value interval range of the data sub-bucket;
the risk rate determining unit is used for determining the actual overtime risk rate of the data sub-bucket according to the actual sample estimated value interval;
and the establishing unit is used for establishing the mapping relation table according to the corresponding relation between the actual sample estimated value interval and the actual overtime risk rate.
The risk rate determining unit is specifically configured to determine the actual timeout risk rate according to the sum of the selected actual sample pre-evaluation values and a total singular number of the waybill in a selected range when the actual sample pre-evaluation value is selected.
The first determining unit 502 may be specifically configured to invoke the mapping relationship table matched with the waybill generation area according to the waybill generation area in the waybill data to be estimated; or calling the mapping relation table matched with the waybill creation time information according to the waybill creation time in the waybill data to be estimated.
The second determining unit 503 may specifically include:
the interval determination subunit is used for determining a target actual sample estimated value interval according to the actual sample estimated value interval in which the timeout estimated value falls;
and the sub-bucket determining sub-unit is used for determining the target actual sample estimated value interval as the data sub-bucket in the mapping relation table corresponding to the overtime estimated value.
The method can also comprise the following steps: and the output unit is used for outputting the target timeout risk data.
For a specific implementation process of the estimation apparatus for waybill distribution timeout provided in the present application, reference may be made to steps S301 to S305 of the estimation method for waybill distribution timeout provided in the present application, and details are not repeated here.
Based on the above, the present application further provides an waybill scheduling method, as shown in fig. 6, fig. 6 is a flowchart of an embodiment of the waybill scheduling method provided in the present application; the embodiment of the scheduling method comprises the following steps:
step S601: acquiring waybill information to be estimated, inputting the waybill information to be estimated into a waybill distribution overtime estimation model, and determining target overtime risk data of the waybill to be estimated;
step S602: arranging the freight notes to be estimated in a descending order according to the risk data values in the target overtime risk data;
step S603: and taking the arranged freight notes to be estimated as freight notes to be scheduled, and scheduling according to the arrangement sequence.
Corresponding to the foregoing waybill scheduling method embodiment, the present application further provides a waybill scheduling apparatus, as shown in fig. 7, an embodiment of the waybill scheduling apparatus may include:
the acquiring unit 701 is used for acquiring target timeout risk data of the waybill to be estimated, wherein the waybill to be estimated is input into a waybill distribution timeout estimation model, and the target timeout risk data is determined;
a sorting unit 702, configured to sort the waybills to be pre-estimated in a descending order according to the risk data values in the target timeout risk data;
and the scheduling unit 703 is configured to schedule the ranked waybills to be pre-estimated as waybills to be scheduled according to the ranking order.
Based on the above, the present application further provides a computer storage medium, configured to store data generated by a network platform and a program for processing the data generated by the network platform;
when the program is read and executed by the processor, the method for training the waybill distribution timeout estimation model is executed; or, executing the estimation method of the waybill distribution overtime; alternatively, the steps of the waybill scheduling method described above are performed.
Based on the above, the present application further provides an electronic device, as shown in fig. 8, an embodiment of the electronic device includes:
a processor 801;
a memory 802 for storing a program for processing data generated by the network platform, wherein the program, when being read and executed by the processor, executes the steps of the training method of the waybill distribution timeout estimation model; or, executing the estimation method of the waybill distribution overtime; alternatively, the steps of the waybill scheduling method described above are performed.
In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
1. Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Disks (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (transient media), such as modulated data signals and carrier waves.
2. As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
Although the present application has been described with reference to the preferred embodiments, it is not intended to limit the present application, and those skilled in the art can make variations and modifications without departing from the spirit and scope of the present application, therefore, the scope of the present application should be determined by the claims that follow.

Claims (15)

1. A training method of a waybill distribution timeout estimation model is characterized by comprising the following steps:
extracting characteristic data in the collected waybill sample data;
determining a real label value corresponding to the waybill sample data according to a defined label value of the estimated waybill distribution overtime level;
inputting the real label value and the characteristic data as a sample data set into a constructed waybill distribution timeout estimation model for training to obtain a predicted value of the waybill sample data timeout;
calculating the loss between the estimated value and the real label value according to a mean square error function and a sequencing loss function to obtain a loss parameter value; wherein the ordering loss function is calculated by the following formula:
Figure 154888DEST_PATH_IMAGE001
said
Figure 665504DEST_PATH_IMAGE002
Is a true tag value, said
Figure 523870DEST_PATH_IMAGE003
Is a predicted value;
and adjusting the waybill distribution timeout estimation model according to the loss parameter value, and determining the adjusted waybill distribution timeout estimation model as a target waybill distribution timeout estimation model.
2. The method for training the waybill distribution timeout estimation model according to claim 1, wherein the extracting the feature data in the collected waybill sample data includes:
collecting the waybill sample data;
and extracting the characteristic data according to the waybill sample data.
3. The method for training the waybill distribution timeout estimation model according to claim 2, wherein the collecting the waybill sample data comprises:
determining the collection range of the waybill sample data according to the time from the creation of the waybill to the time from the receiving of the waybill by a distributor;
and collecting the waybill sample data in the collection range according to the set collection time.
4. The method for training the waybill distribution timeout estimation model according to claim 1, wherein the extracting the feature data according to the waybill sample data comprises:
and extracting the feature data from the waybill sample data according to the set extraction dimension.
5. The method for training the waybill distribution timeout estimation model according to claim 1, wherein the determining a true tag value corresponding to the sample data according to the defined tag of the estimated waybill distribution timeout level comprises:
defining a label value of the estimated waybill distribution overtime level;
and determining the real label value corresponding to the sample data according to the label value.
6. The method for training the waybill distribution timeout estimation model according to claim 5, wherein the calculating the loss between the estimated value and the real label according to a mean square error loss function and a rank order loss function to obtain a loss parameter value comprises:
calculating a mean square error loss function according to the real label value and the estimated value;
and calculating the loss parameter value according to the mean square error loss function and the sequencing loss function.
7. A method for training a waybill dispatch timeout estimation model according to claim 6, wherein said calculating the loss parameter value according to the mean square error loss function and the ordering loss function comprises:
and determining the sum of the quotient of the mean square error loss function and the set reference value and the sorting loss function as the loss parameter value.
8. A method for predicting delivery timeout of waybills, comprising:
inputting waybill data to be estimated into a waybill distribution timeout estimation model to obtain a timeout estimated value of the waybill data to be estimated; the waybill distribution timeout estimation model is obtained by adopting a training method of the waybill distribution timeout estimation model according to any one of claims 1 to 6;
determining a called mapping relation table according to attribute information in the waybill data to be estimated; the mapping relation table is used for describing a corresponding relation between an actual sample estimated value and an actual overtime risk rate of actual waybill sample data of the waybill distribution overtime estimation model;
determining data sub-buckets in the mapping relation table corresponding to the timeout estimated value;
and determining actual overtime risk data corresponding to the data sub-buckets as target overtime risk data of the waybill to be estimated.
9. The method of claim 8, further comprising:
selecting actual waybill transport order sample data according to a data generation area range and/or a time range of the actual waybill transport order sample data determined by waybill transport order attributes of waybill transport order data to be estimated;
inputting the actual waybill sample data into the waybill distribution timeout estimation model to obtain the actual sample estimated value;
dividing the actual waybill sample data according to the data sub-buckets according to the sequencing of the actual sample pre-estimated values;
acquiring the actual sample estimated value interval range of the data sub-bucket;
determining the actual overtime risk rate of the data sub-bucket according to the actual sample estimated value interval;
and establishing the mapping relation table according to the corresponding relation between the actual sample pre-evaluation value interval and the actual overtime risk rate.
10. The method of claim 9, wherein said determining an actual timeout risk rate waybill for said data buckets based on said actual sample budget interval comprises:
and determining the actual overtime risk rate according to the sum of the selected actual sample estimated values and the total singular number of the freight note in the selected range when the actual sample estimated values are selected.
11. The method as claimed in claim 8, wherein the determining the mapping relationship table to be invoked according to the attribute information in the waybill data to be estimated includes:
calling the mapping relation table matched with the waybill generation area according to the waybill generation area in the waybill data to be estimated; or calling the mapping relation table matched with the waybill creation time information according to the waybill creation time in the waybill data to be estimated.
12. The method of claim 8, wherein the determining the data buckets in the mapping table corresponding to the timeout estimate comprises:
determining a target actual sample estimated value interval according to the actual sample estimated value interval in which the overtime estimated value falls;
and determining the target actual sample estimated value interval as a data sub-bucket in the mapping relation table corresponding to the overtime estimated value.
13. A waybill scheduling method, comprising:
acquiring waybill information to be estimated, inputting the waybill information to be estimated into a waybill distribution overtime estimation model, and determining target overtime risk data of the waybill to be estimated; wherein, the waybill distribution timeout estimation model is obtained by adopting a training method of the waybill distribution timeout estimation model according to any one of the claims 8 to 12;
arranging the freight notes to be estimated in a descending order according to the risk data values in the target overtime risk data;
and taking the arranged freight notes to be estimated as freight notes to be scheduled, and scheduling according to the arrangement sequence.
14. A computer storage medium for storing network platform generated data and a program for processing the network platform generated data;
the program, when read and executed by a processor, performs the steps of the method of training a waybill dispatch timeout estimation model according to any of claims 1-7 above; or, performing the steps of the method of estimation of waybill distribution timeouts as claimed in any of the preceding claims 8-12; alternatively, the steps of the waybill scheduling method as set forth in claim 13 above are performed.
15. An electronic device, comprising:
a processor;
a memory for storing a program for processing data generated by a network platform, wherein the program, when read and executed by the processor, performs the steps of the training method of the waybill distribution timeout estimation model according to any one of claims 1 to 7; alternatively, the steps of performing the method of estimating waybill dispatch timeout according to any of claims 10-12 above; alternatively, the steps of the waybill scheduling method as set forth in claim 13 above are performed.
CN202210029170.7A 2022-01-11 2022-01-11 Training method of waybill distribution timeout estimation model, storage medium and electronic equipment Active CN114066105B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210029170.7A CN114066105B (en) 2022-01-11 2022-01-11 Training method of waybill distribution timeout estimation model, storage medium and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210029170.7A CN114066105B (en) 2022-01-11 2022-01-11 Training method of waybill distribution timeout estimation model, storage medium and electronic equipment

Publications (2)

Publication Number Publication Date
CN114066105A CN114066105A (en) 2022-02-18
CN114066105B true CN114066105B (en) 2022-09-27

Family

ID=80230796

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210029170.7A Active CN114066105B (en) 2022-01-11 2022-01-11 Training method of waybill distribution timeout estimation model, storage medium and electronic equipment

Country Status (1)

Country Link
CN (1) CN114066105B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110390503A (en) * 2019-07-03 2019-10-29 北京三快在线科技有限公司 Method and apparatus, storage medium and the electronic equipment that model training, distribution time determine
CN113821535A (en) * 2021-07-15 2021-12-21 腾讯科技（深圳）有限公司 Model training method and device, computer equipment and storage medium

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109685276B (en) * 2018-12-27 2021-01-01 拉扎斯网络科技（上海）有限公司 Order processing method and device, electronic equipment and computer readable storage medium
CN111798283A (en) * 2019-04-09 2020-10-20 北京嘀嘀无限科技发展有限公司 Order distribution method and device, electronic equipment and computer readable storage medium
CN110705934A (en) * 2019-09-25 2020-01-17 北京三快在线科技有限公司 Abnormal order identification method and device, readable storage medium and electronic equipment
CN113393284A (en) * 2020-03-13 2021-09-14 北京三快在线科技有限公司 Order state prompting method and device, storage medium and electronic equipment
CN111815059B (en) * 2020-07-13 2021-04-23 拉扎斯网络科技（上海）有限公司 Data processing method and device, electronic equipment and computer readable storage medium
CN112465017A (en) * 2020-11-26 2021-03-09 平安科技（深圳）有限公司 Classification model training method and device, terminal and storage medium
CN113159281A (en) * 2021-03-25 2021-07-23 拉扎斯网络科技（上海）有限公司 Data processing method and data processing device

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110390503A (en) * 2019-07-03 2019-10-29 北京三快在线科技有限公司 Method and apparatus, storage medium and the electronic equipment that model training, distribution time determine
CN113821535A (en) * 2021-07-15 2021-12-21 腾讯科技（深圳）有限公司 Model training method and device, computer equipment and storage medium

Also Published As

Publication number Publication date
CN114066105A (en) 2022-02-18

Similar Documents

Publication Publication Date Title
US8374903B2 (en) Information criterion-based systems and methods for constructing combining weights for multimodel forecasting and prediction
US20090187468A1 (en) Automated replenishment using an economic profit quantity
US20170206621A1 (en) Systems and Methods for Determining the Effectiveness of Warehousing
US20180308039A1 (en) System and Method for Dynamically Establishing A Regional Distribution Center Truck Flow Graph to Distribute Merchandise
US8768866B2 (en) Computer-implemented systems and methods for forecasting and estimation using grid regression
CN107767092B (en) Processing method, display method and device of commodity object information
CN111415193B (en) Method for calculating class sales promotion degree based on store similarity of associated commodities
US20220277331A1 (en) Systems and methods for procurement cost forecasting
CN111445133B (en) Material management method and device, computer equipment and storage medium
CN113204712A (en) Information pushing method, device, medium and program product based on community service
CN116109252A (en) Warehouse replenishment management method and device, warehouse management system and storage medium
van der Hagen et al. Machine learning–based feasibility checks for dynamic time slot management
CN114239977A (en) Method, device, equipment and storage medium for determining estimated delivery time length
CN113627847A (en) Method and device for generating replenishment list
CN113780913B (en) Method and device for generating safety stock information
CN114066105B (en) Training method of waybill distribution timeout estimation model, storage medium and electronic equipment
WO2017123750A1 (en) Systems and methods for determining the effectiveness of warehousing
CN111507779A (en) Method and device for grading commodities
CN116128408A (en) Intelligent replenishment method, system and computer readable medium
CN113627839B (en) Sectional take-out distribution method
CN113762842B (en) Warehouse scheduling method, server and system
CN112990948A (en) Service evaluation method, device, equipment, storage medium and computer program product
CN110414875B (en) Capacity data processing method and device, electronic equipment and computer readable medium
CN112990968A (en) Lubrication dividing system and method
CN110458345A (en) Determine the method, apparatus, equipment and storage medium of machine loss shipment amount

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant