CN111353625A

CN111353625A - Method, device, equipment and storage medium for predicting dot part quantity

Info

Publication number: CN111353625A
Application number: CN201811572809.6A
Authority: CN
Inventors: 刘曙铭; 王本玉; 湛长兰; 李凤; 肖沙沙; 吴敏礽; 金晶
Original assignee: SF Technology Co Ltd
Current assignee: SF Technology Co Ltd; SF Tech Co Ltd
Priority date: 2018-12-21
Filing date: 2018-12-21
Publication date: 2020-06-30

Abstract

The application discloses a method, a device and equipment for predicting the screen dot quality and a readable storage medium. The method comprises the following steps: constructing a time-consuming prediction model; acquiring real-time routing data and historical routing data corresponding to the waybill data of the network points; predicting the quantity of the parts in the fixed time period of each network point by combining the real-time routing data and the time-consuming model; and calculating the route missing rate according to historical route data, and determining the final quantity of the mesh points in a fixed time period. According to the method and the device, the problems of inaccurate detection results and long calculation period of a machine learning algorithm in the prior art are solved by constructing a new time-consuming model, and the accuracy of flood peak component prediction is improved.

Description

Method, device, equipment and storage medium for predicting dot part quantity

Technical Field

The present invention relates generally to the field of logistics technology, and more particularly to a method, an apparatus, a device and a storage medium for predicting network component quantity.

Background

In the field of logistics express delivery, how to better predict the delivery volume is always an important link for improving enterprise services. The basis for optimizing the express delivery scheme is the prediction of the daily delivery quantity of the express at the distribution point, because various factors influence the final arrival time of the express at the distribution point in the actual delivery process.

At present, a part quantity prediction model utilizes a machine learning algorithm to learn the part quantity growth trend through a mesh point dimension and a time dimension so as to perform time sequence prediction. The prediction of the time series is based on the identification of "long-term trends", "seasonal variations", "cyclic variations", while the prediction is poor for "irregular variations". For example, in the industry at present, a large error exists in prediction of peak values for certain special dates (such as 618, 11 pairs and the like), and the prediction has a large error mainly because of the fact that the peak values of the pieces exist on the dates, and the growing trend of the pieces is difficult to learn from historical data. In addition, the dispatch of each network point cannot well utilize routing information and historical time-consuming information, so that the flood peak of the quantity of the sensing parts cannot be timely sensed, and inaccurate prediction is also caused.

How to solve the phenomenon of inaccurate component peak flood prediction caused by special dates is a problem to be solved urgently.

Disclosure of Invention

In view of the above-mentioned drawbacks and deficiencies of the prior art, it is desirable to provide a method, an apparatus, a device and a storage medium for predicting halftone part quantity, so as to improve accuracy of halftone part quantity prediction.

In a first aspect, an embodiment of the present invention provides a method for predicting a halftone dot quality, where the method includes:

constructing a time-consuming prediction model, wherein the time-consuming prediction model is trained on the basis of an Xgboost model;

acquiring real-time routing data and historical routing data corresponding to the waybill data of the network points;

predicting the quantity of the parts in the fixed time period of each network point by combining the real-time routing data and the time-consuming model;

and calculating the route missing rate according to historical route data, and determining the final quantity of the mesh points in a fixed time period.

In a second aspect, an embodiment of the present invention provides a device for predicting a halftone dot quality, where the device includes:

the model construction module is used for constructing a time-consuming prediction model, and the time-consuming prediction model is trained on the basis of an Xgboost model;

the data acquisition module is used for acquiring real-time routing data and historical routing data corresponding to the waybill data of the network points;

the quantity prediction module is used for predicting the quantity of the network points in a fixed time period by combining the real-time routing data and the time-consuming model;

and the quantity determining module is used for calculating the route missing rate according to historical route data and determining the final quantity of the network points in the fixed time period.

In a third aspect, embodiments of the present application provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the method as described in embodiments of the present application when executing the program.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, the computer program being configured to:

which when executed by a processor implements a method as described in embodiments of the present application.

According to the method for predicting the flood peak component quantity, the problems that in the prior art, a machine learning algorithm is inaccurate in detection result and long in calculation period are solved by optimizing and constructing a new time-consuming model, and the accuracy of predicting the flood peak component quantity is improved.

Drawings

Other features, objects and advantages of the present application will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:

fig. 1 is a schematic flowchart of a method for predicting a halftone dot quality according to an embodiment of the present application;

fig. 2 is a schematic structural block diagram of a mesh point quantity prediction apparatus provided in the embodiment of the present application according to the embodiment of the present invention;

fig. 3 is a schematic structural diagram of a computer system suitable for implementing a terminal device according to an embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the following drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not restrictive of the invention. It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings.

It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.

The network component predicting method and system described in the embodiments of the present application may include, for example, a terminal device, a network, a server, and the like. The terminal device may be, for example, a computer device. The network is used for connecting communication between the terminal device and the server. Which may be a wireless, wired, or fiber optic transmission medium.

Referring to fig. 1, fig. 1 is a schematic flowchart illustrating a method for predicting a halftone dot quality according to an embodiment of the present disclosure. The method may be performed on the server side.

As shown in fig. 1, the method includes:

and step 110, constructing a time-consuming prediction model, wherein the time-consuming prediction model is trained on the basis of an Xgboost model. In the embodiment of the application, such a time-consuming prediction model is built by integrating a plurality of CART trees based on Xgboost, which is short for Extreme Gradient ascent (Extreme Gradient ascent), and is similar to a Gradient ascent framework but more efficient. It has both a linear model solver and a tree learning algorithm. The sample output of the classification tree is in the form of classes. In fact, classification and regression are model-based, except that the results of classification are discrete and regression is continuous.

The time-consuming model is constructed using the Spark calculation engine based Xgboost algorithm and the time-consuming table is updated on an hourly basis. The Spark is a distributed computing engine based on the memory, most of the computing process of the Spark is carried out in the memory, the intermediate result is also stored in the memory, and the reading and writing of a disk are not needed, so that the model training speed can be increased when the data volume is large, the frequency updating is increased, and the updating by hours is achieved. For the time-consuming model, the latest change trend can be obtained by acquiring new data, and the latest change trend can directly influence the accuracy of the time-consuming model, so that the time-consuming accuracy of calculation is improved by increasing the frequency updating speed of the model.

And step 120, acquiring real-time routing data and historical routing data corresponding to the waybill data of the network points.

In the embodiment of the application, the real-time routing table contains data of the whole life cycle of the waybill, the data are sent to a network point from the courier's house-in to the courier, then to each transfer site, land transportation, air transportation and the like, and then to the user, and each node in the whole process uploads routing data. By analyzing these routing data, we can obtain the time consumption between transitions from historical data. The routing data comprises a plurality of operations, such as an entrance receiving route, a network point reaching route, a departure route, a transfer route and a destination reaching route, and the time consumption situation of a certain distance can be obtained by screening out part of routing data. For example, if a dispatch takes one hour from transition a to transition B, we will take the time of arrival at transition B as the end time, and the time of arrival at transition B minus the end time is the time consumption from transition a to transition B.

And performing duplicate removal, cleaning and completion operation on the real-time routing data and the historical routing data corresponding to the waybill data of each network point, and using the real-time routing data and the historical routing data as data sources for model input. Problems such as data duplication, data deletion or abnormal values occur due to problems in the data acquisition process caused by misoperation, a service system and other human reasons in the routing table. For these problems, the most original data needs to be subjected to deduplication operation, so as to ensure the uniqueness of the data. And abnormal value judgment needs to be carried out on data, for example, routing data from the place A to the place B takes one hour under the conventional condition, while some data can be calculated for one day or even longer, the data which is much beyond the normal value is obviously abnormal value, and the abnormal value needs to be processed to ensure that the data source is stable and accurate.

And step 130, predicting the quantity of the parts in the fixed time period of each network point by combining the real-time routing data and the time-consuming model.

In the embodiment of the application, the mark routing data is selected from the real-time routing table and is cleaned and preprocessed, the routing data representing the accurate position of the waybill needs to be selected due to the huge data amount in the routing table, the main flow of data access is that various routing data are generated and then transmitted to the server in real time, and the real-time routing data are accessed to the big data processing platform through the kafka component. Kafka is a data buffering component. In a real big data application scene, data explosion growth occurs in a certain period of time, if the real-time data is directly stored in a big data application system, the system may not be in time to process the rapidly-growing data, and a buffer mechanism is needed, which can be used as a buffer zone to slowly transfer the suddenly-increased data to the big data system, thereby ensuring that the data is not lost.

Specifically, the kafka component is utilized to analyze the received real-time routing data in real time by using spark timing technology. Spark timing is a spark-based high-level component, primarily used for real-time computing. The main purpose of the component is to carry out batch processing operation on continuous data according to a time period of 1S at the minimum. Each time, the data in the shortest 1S is processed, then the processing and processing operation is carried out on the data in the 1S, and the data in the next second is processed continuously after the data in the second is processed, so that the system is a real-time system which continuously runs for 7 x 24 hours. The estimated arrival time of the waybill at the network point can be obtained by combining the waybill table and the time-consuming table, the predictable period is shortened according to the updated routing data each time, and the prediction accuracy is improved.

And step 140, calculating a route missing rate according to the historical route data, and determining the final quantity of the nodes in a fixed time period.

In the embodiment of the application, because the real-time routing data is missing, a certain deviation exists in the predicted component flood peak according to the missing routing data, and the predicted component is low. And calculating the obtained actual quantity of components according to the historical routing data and performing data analysis on the actual quantity of components obtained in the waybill table, so that the missing proportion of the historical routing data can be confirmed, and dividing the quantity data predicted by the real-time routing by the proportion according to the calculated proportion to obtain the final quantity flood peak data.

Optionally, constructing a time-consuming prediction model includes the following steps:

s11, the feature set is constructed based on the time dimension, business dimension and geographic dimension.

In particular, due to date, time of day. Factors such as the peak on duty, the number of people on duty, weather and road conditions are all related to the time-consuming model, so that the characteristics are extracted and a characteristic project is constructed. The feature engineering is mainly established from the following three dimensions, namely a time dimension, a geographical dimension and a business dimension.

S12, acquiring an XgBoost integration algorithm and establishing a time-consuming model by using a Spark calculation engine.

Specifically, the integration algorithm in the embodiment of the present application may be that Xgboost is mainly based on a binary tree of a CART regression tree, and continuously splits features, for example, a current tree node is split based on a jth feature value, a sample with the feature value smaller than S is divided into a left sub-tree, and a sample with the feature value larger than S is divided into a right sub-tree.

The CART regression tree is actually to divide the sample space in the feature dimension, and the optimization of the space division is an NP-hard problem, so that a heuristic method is used for solving in a decision tree model, and a regression tree is finally obtained in order to solve the optimal segmentation feature and segmentation point.

boost is ensemble learning, which means that a plurality of classifiers are constructed to predict data, and then the results predicted by the plurality of classifiers are integrated by a certain strategy to be used as the final prediction result. Xgboost is also just an ensemble learning of boosting. Xgboost is also a GBDT in nature, but strives to bring speed and efficiency to a premium. The core algorithm idea is to add trees continuously, and to grow a tree by continuously performing feature splitting, wherein each time a tree is added, a new function is learned to fit the residual error predicted last time. Xgboost is more robust to residual learning than the traditional decision tree algorithm. And optimizing the model by residual errors step by step. For example, if we predict the age of a person, and if we predict the age of the person to be 25 years old through the first tree, then the residual error is the actual value minus the predicted value, and then the residual error is-3, which proves that the value predicted by the first regression tree is low. When training the second lesson tree, the predicted value of the second lesson tree is the residual error-3, if the predicted value of the second lesson tree is 4, the combination of the first lesson tree and the second lesson tree is 29, and the predicted value is higher than the real value by 1. And then continuously learning the residual error, and training the model only when the tree of the tree reaches the maximum or the iteration times are finished. Compared with the traditional decision tree, the XGboost-based residual error learning mechanism has better accuracy.

The computing engine adopts Spark as an open source cluster computing framework for real-time processing, and Spark has a high-level component Spark timing which can well solve the problem of real-time computing in the aspect of real-time data analysis, and is the most common in all other solutions. The whole Spark cluster is divided into a Master node and a Worker node, wherein the Master node is resident with a Master daemon process and a Driver process, the Master is responsible for changing serial Tasks into task sets Tasks capable of being executed in parallel and is also responsible for error problem processing and the like, the Worker node is resident with a Worker daemon process, the Master node and the Worker node are different in division, the Master load manages all the Worker nodes, and the Worker node is responsible for executing Tasks. Since Spark runs tasks on a cluster basis, the Spark comprises two nodes, one is a Master node and the other is a worker node. Master is mainly responsible for allocating resources, and each time a task runs on the spark cluster, Master is responsible for calling the cluster resources to execute the task, similar to a leader. The Worker is more like the bottom-layer staff, and after a new task comes, the leader Master can distribute the task to the workers, so that the workers can perform corresponding calculation tasks. This is Master and Worker in Spark. Each time a spare task is started, a Driver process is started on the Master, and the process is responsible for the execution of the whole task.

And S13, inputting the feature set into the time-consuming model for training to form a time-consuming prediction model.

Specifically, the feature set mainly comprises three aspects of a natural dimension, a time dimension and a business dimension, the time dimension-based features mainly include feature values of year, month, day, hour, working day, non-working day, holiday and the like, and such influence factors can influence the timeliness of express transportation. Express mail from a place A to a place B can be influenced by time dimension, the influence of weekends and workdays on time consumption, the influence of holidays and festivals and the influence of different time periods, so the influence of different time dimensions on time consumption must be analyzed through historical data of last three months, and influence weight is obtained.

In the aspect of area and weather dimension, the main reason is that express mail arrives at the place B from the place A and is also influenced by the dimension between the places, the traffic conditions of different places are different, and the time consumption is reduced by convenient traffic. Weather factors also affect the time consumption of the express mail, for example, rainy days or extreme weather greatly affects the transportation of the express mail, thereby causing the time consumption to increase. By accessing an external data source or crawling data by using a crawler, data of the geographical dimensionality of the website can be acquired so as to be used as a feature construction weight.

The direction of business dimension and the transportation process of the express are influenced by human factors. The number of people in the transfer can directly influence the transfer and transportation efficiency of the express mail, so that the personnel scheduling condition is obtained, the business weight is constructed, and the consumed time is calculated.

Furthermore, the collected characteristic data is input into a time-consuming prediction model for training, so that the model has the capability of predicting the time of the express mail arriving at the key station.

Optionally, a time-consuming model is constructed by using an xgboost algorithm based on the spark calculation engine, and the time-consuming table is updated according to the hourly frequency, so that the trend of time-consuming change can be found in time, and the time-consuming accuracy is improved.

Optionally, the real-time routing data and the historical routing data corresponding to the network node waybill data are obtained, where the real-time routing data and the historical routing data include any one of the following items or any combination of two or more of the following items: the order number, the operation time, the operation code, the operation place, the operation type and the transit data.

Specifically, the real-time routing list contains data of the whole life cycle of the freight note, the data are sent to the internet from the express delivery personnel to the home delivery site, then to each transfer site, land transportation, air transportation and the like, and finally sent to the user, and each node in the whole process can upload routing data. Since the time-consuming state of the dispatch between such transitions can be derived from this type of data.

Further, duplication removal is carried out according to the waybill number, the operation code and the operation time, uniqueness of data is guaranteed, then the data are removed, dirty data mainly comprise that important fields are missing, when the waybill number, the operation time and other important fields are missing, the data lose values and need to be removed, and the other data are abnormal in data value and mainly comprise that the waybill time is too long and completely exceeds a reasonable range, the data need to be removed.

Optionally, predicting the quantity of the components in the fixed time period of each mesh point by combining the real-time routing data and the time-consuming model, including the following steps:

and S21, determining the network point position of the express according to the real-time routing data.

Specifically, the real-time location of the courier may be determined based on the real-time routing data, for example, during a double 11 promotion, because the courier volume is large and the traffic impact is severe, it may not be accurate to analyze or predict the arrival time of the courier at the terminal based only on the real-time routing data.

And S22, predicting the time of the express item of the network point position to the final network point by using the time-consuming model.

Specifically, based on a time-consuming model established by a machine learning model used in Xgboosing processing, influence characteristics influencing express delivery are collected, and the time of the express delivery of the network point reaching the final address is predicted by using the time-consuming model.

And S23, determining the total predicted part amount of each final mesh point in a fixed time period by combining the time of the final mesh point.

Specifically, data predicted by the time-consuming model and data of a real-time route are combined and compared, the position of the express mail is accurately positioned through the real-time route, a destination and a route are obtained by combining a freight note, and the arrival time of the quantity of the express mail is predicted through the time-consuming model, so that the peak of the quantity of the express mail at a network site can be predicted. By continuously updating the real-time route, the time of the express mail reaching the network point is more accurate.

Optionally, calculating a route missing rate according to historical route data, and determining a final quantity of the mesh points in a fixed time period, including:

and S31, analyzing the missing rate of the historical routing data of each network point by using a statistical method.

Specifically, because real-time routing data is missing, the routing missing rate is calculated according to historical routing data, and the network node part data is supplemented according to the proportion. And obtaining the estimated delivery quantity and the actual delivery quantity of the network points, and dividing the estimated delivery quantity by the actual delivery quantity to obtain the route loss rate.

And S32, complementing the numerical value of the predicted part quantity of each mesh point based on the deficiency rate.

Specifically, according to the calculated average loss rate, the predicted peak flood component of the mesh point is divided by the loss rate, so that the final predicted peak flood component of the mesh point can be obtained.

In a second aspect, an embodiment of the present application provides a halftone quality prediction apparatus, as shown in fig. 2, which shows a block diagram of a halftone quality prediction apparatus provided according to an embodiment of the present application, and the apparatus includes:

and a model building module 210, configured to build a time-consuming prediction model, where the time-consuming prediction model is trained based on an Xgboost model.

The data obtaining module 220 is configured to obtain real-time routing data and historical routing data corresponding to the waybill data of the website.

And a component predicting module 230, configured to predict a component in a fixed time period of each mesh point by combining the real-time routing data and the time-consuming model.

And the quantity determining module 240 is configured to calculate a route missing rate according to historical route data, and determine a final quantity in the fixed time period of the mesh point.

Optionally, the model building module 210 is specifically configured to:

and the feature set unit is used for constructing a feature set based on the time dimension, the service dimension and the geographic dimension.

And the time-consuming model unit is used for acquiring the xgboost integration algorithm and establishing the time-consuming model by using a Spark calculation engine.

Optionally, the component prediction module 230 is specifically configured to:

and the determining unit is used for determining the network point position of the express according to the real-time routing data.

And the first prediction unit is used for predicting the time of the express item at the network point position to the final network point by using the time-consuming model.

And the second prediction unit is used for determining the total predicted part amount of each final mesh point in a fixed time period by combining the time of the final mesh point.

It should be understood that the units or modules described in the above-described apparatus correspond to the individual steps in the method described with reference to fig. 1. Thus, the operations and features described above for the method are equally applicable to the above described apparatus and the units comprised therein, which are not further elucidated here.

Referring now to FIG. 3, shown is a block diagram of a computer system 600 suitable for use in implementing a server according to embodiments of the present application.

As shown in fig. 3, the computer system 600 includes a Central Processing Unit (CPU)601 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM)602 or a program loaded from a storage section 608 into a Random Access Memory (RAM) 603. In the RAM 603, various programs and data necessary for the operation of the system 600 are also stored. The CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.

The following components are connected to the I/O interface 605: an input portion 606 including a keyboard, a mouse, and the like; an output portion 607 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 608 including a hard disk and the like; and a communication section 609 including a network interface card such as a LAN card, a modem, or the like. The communication section 609 performs communication processing via a network such as the internet. The driver 610 is also connected to the I/O interface 605 as needed. A removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 610 as necessary, so that a computer program read out therefrom is mounted in the storage section 608 as necessary.

In particular, the process described above with reference to fig. 1 may be implemented as a computer software program, according to an embodiment of the present disclosure. For example, embodiments of the present disclosure include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program comprising program code for performing the method of fig. 1. In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 609, and/or installed from the removable medium 611.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units or modules described in the embodiments of the present application may be implemented by software or hardware. The described units or modules may also be provided in a processor.

As another aspect, the present application also provides a computer-readable storage medium, which may be the computer-readable medium contained in the apparatus in the above-described embodiments; or it may be a computer readable medium that exists separately and that is not incorporated into the device. The computer readable medium stores one or more programs, and when the one or more programs are executed by the electronic device, the electronic device is enabled to implement the network element quantity prediction method as described in the above embodiments.

The above description is only a preferred embodiment of the application and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention herein disclosed is not limited to the particular combination of features described above, but also encompasses other arrangements formed by any combination of the above features or their equivalents without departing from the spirit of the invention as defined above. For example, the above features may be interchanged with other features disclosed in this application (but not limited to) having similar functions

As can be seen from the above description: those skilled in the art can clearly understand that the present application must be implemented by means of hardware background. Based on this understanding, the technical solution of the present application can be essentially said to be embodied in the form of a development program of a computer, and includes several instructions to enable a computer device (personal computer, server, or network device, etc.) to execute the method described in some parts of the embodiments of the present application.

Claims

1. A method for predicting a halftone dot quality, the method comprising:

2. A method for predicting a halftone partem according to claim 1, wherein said building a time-consuming prediction model comprises:

constructing a feature set based on a time dimension, a business dimension and a geographic dimension;

acquiring an XgBoost integration algorithm and establishing a time-consuming model by using a Spark calculation engine;

inputting the feature set into the time-consuming model for training to form a time-consuming prediction model.

3. A method for predicting a halftone component according to claim 2, wherein said constructing a feature set based on a time dimension, a business dimension and a geographic dimension comprises:

the set of time dimension features includes: one or more of year, month, day, hour, working day, non-working day, holiday;

the business dimension feature set comprises: manual shift and/or vehicle scheduling;

the set of geography-dimension characteristics includes: weather conditions and/or intersection factors.

4. A dot product quantity prediction method according to claim 3, wherein the time-consuming prediction model is updated at regular intervals based on a newly constructed feature set, and outputs a time-consuming table.

5. The method for predicting the quantity of network points according to claim 4, wherein the real-time routing data and the historical routing data corresponding to the obtained network point waybill data include any one of or a combination of any two or more of the following items:

the order number, the operation time, the operation code, the operation place, the operation type and the transit data.

6. A mesh point quantity forecasting method according to claim 5, wherein the forecasting the quantity of the mesh point in a fixed time period by combining the real-time routing data and the time-consuming model comprises:

determining the position of a network point of the express according to the real-time routing data;

predicting the time of the express item at the network point position to the final network point by using a time-consuming model;

and determining the total predicted part amount of each final mesh point in a fixed time period by combining the time of the final mesh point.

7. A mesh point quantity forecasting method according to claim 6, wherein the calculating of the route missing rate according to the historical route data and the determining of the final quantity of the mesh point in the fixed time period comprise:

analyzing the missing rate of historical routing data of each network point by using a statistical method;

and completing the numerical value of the predicted part quantity of each network point based on the deficiency rate.

8. A halftone dot quality prediction apparatus, comprising:

9. A mesh point quantity prediction device according to claim 8, wherein the model construction module is specifically configured to:

the feature set unit is used for constructing a feature set based on a time dimension, a business dimension and a geographic dimension;

the time-consuming model unit is used for acquiring an XgBoost integration algorithm and establishing a time-consuming model by using a Spark calculation engine;

and the training unit is used for inputting the feature set into the time-consuming model for training so as to form a time-consuming prediction model.

10. A mesh point quantity prediction device according to claim 8, wherein the quantity prediction module is specifically configured to:

the determining unit is used for determining the network point position of the express according to the real-time routing data;

the first prediction unit is used for predicting the time of the express item at the network point position to the final network point by using a time-consuming model;

11. An apparatus, comprising: at least one processor, at least one memory, and computer program instructions stored in the memory that, when executed by the processor, implement the method of any of claims 1-7.

12. A computer-readable storage medium having stored thereon computer program instructions, which when executed by a processor, perform the method according to any of claims 1-7.