Method and apparatus for predicting mesh point quantity
By combining the network point representation decision tree model and graph attention network, the problem of low accuracy in existing network point shipment prediction is solved, achieving more efficient shipment prediction and operational optimization, improving user experience and reducing costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SF TECH CO LTD
- Filing Date
- 2021-09-14
- Publication Date
- 2026-06-16
AI Technical Summary
Existing methods for predicting parcel volume at network points have low accuracy and cannot effectively capture outliers, leading to frequent occurrences of damage to user experience, such as parcel breakage during transportation.
A decision tree model for representing network nodes is used to represent network node feature information. By splicing the leaf node representation information of multiple decision trees with the network node feature information, and combining graph attention network and gradient boosting decision tree model, the prediction accuracy is improved.
It improves the accuracy of parcel volume forecasting at network points, enabling better matching of personnel and traffic, optimizing network operations, reducing transportation costs, and minimizing impact on user experience.
Smart Images
Figure CN115809712B_ABST
Abstract
Description
Technical Field
[0001] This application mainly relates to the field of big data technology, specifically to a method and apparatus for predicting the number of items at a network point. Background Technology
[0002] The express delivery industry is a competition between user experience and cost. User experience is mainly reflected in the user's perception of speed and undamaged delivery. In the continuous pursuit of speed, fluctuations in parcel volume can easily lead to damaged packages and other behaviors that harm the user experience. Cost is mainly the labor cost on the transportation routes. How to reasonably match the traffic pressure of each branch with the personnel shifts has become an optimization direction. However, traditional parcel volume forecasting usually uses time series models, which are more inclined to capture cyclical patterns and smooth operations within the movement cycle. They are not sensitive to outliers and have low accuracy.
[0003] In other words, the existing methods for predicting the number of items at network outlets have low accuracy. Summary of the Invention
[0004] This application provides a method and apparatus for predicting the quantity of items at a network point, aiming to solve the problem of low prediction accuracy in existing methods for predicting the quantity of items at a network point.
[0005] Firstly, this application provides a method for predicting the quantity of mail at a distribution point, the method comprising:
[0006] Obtain the target network's characteristic information within the first historical time period;
[0007] The network feature information is represented based on the network representation decision tree model to obtain the leaf node representation information of multiple decision trees;
[0008] The leaf node representation information of multiple decision trees and the network feature information are concatenated to obtain the network representation information;
[0009] Based on the network representation information, the predicted quantity of the target network node is determined.
[0010] Optionally, the network feature information includes basic feature information and feature information to be expanded;
[0011] The network feature information is represented by the network feature-based decision tree model to obtain the leaf node representation information of multiple decision trees, including:
[0012] The feature information to be expanded is expanded to obtain the expanded feature information;
[0013] Obtain the network representation decision tree model;
[0014] The expanded feature information and the basic feature information are input into the network node representation decision tree model to obtain the leaf node representation information of multiple decision trees.
[0015] Optionally, the step of extending the feature information to be extended to obtain extended feature information includes:
[0016] Obtain historical routes that passed through the target network point within the second historical time period but did not pass through the target network point within the first historical time period, and current train information from the network point feature information;
[0017] The current train information is expanded based on the historical number of trains on the historical route and the current train information in the network feature information to obtain expanded train information;
[0018] Based on the expanded train information, the feature information to be expanded is expanded to obtain the expanded feature information.
[0019] Optionally, the acquisition of the network node representation decision tree model includes:
[0020] Obtain a network point feature training set, wherein the network point feature training set includes multiple network point feature information samples and corresponding sample labels within a third historical time period;
[0021] The preset decision tree model is trained based on the network feature training set to obtain the network representation decision tree model.
[0022] Optionally, training a preset decision tree model based on the dot feature training set to obtain the dot representation decision tree model includes:
[0023] The network feature training set is input into the preset decision tree model to obtain the leaf node features of the decision tree;
[0024] The leaf node features of the decision tree are input into a graph attention network to obtain the attention coefficients between each node.
[0025] Based on the attention coefficient, the leaf node features of the decision tree are updated through the back gradient, and the preset decision tree model is iterated and updated multiple times to obtain the dot representation decision tree model.
[0026] Optionally, the preset decision tree model is a gradient boosting decision tree model.
[0027] Optionally, determining the predicted quantity of the target network node based on the network node characterization information includes:
[0028] The network node representation information is input into the network node quantity prediction model to obtain the network node quantity of the target network node. The network node quantity prediction model is obtained by training a first preset linear model with a network node representation training set. The network node representation training set includes multiple network node representation information samples and corresponding sample labels.
[0029] Secondly, this application provides a device for predicting the quantity of items at network points, the device comprising:
[0030] The acquisition unit is used to acquire the network feature information of the target network point within the first historical time period.
[0031] The representation unit is used to represent the feature information of the network nodes based on the network node representation decision tree model, and obtain the leaf node representation information of multiple decision trees;
[0032] The splicing unit is used to splice the leaf node representation information of multiple decision trees and the network feature information to obtain network representation information;
[0033] The determining unit is used to determine the predicted quantity information of the target network point based on the network point characterization information.
[0034] Optionally, the network feature information includes basic feature information and feature information to be expanded;
[0035] The characterization unit is used for:
[0036] The feature information to be expanded is expanded to obtain the expanded feature information;
[0037] Obtain the network representation decision tree model;
[0038] The expanded feature information and the basic feature information are input into the network node representation decision tree model to obtain the leaf node representation information of multiple decision trees.
[0039] Optionally, the characterization unit is used for:
[0040] Obtain historical routes that passed through the target network point within the second historical time period but did not pass through the target network point within the first historical time period, and current train information from the network point feature information;
[0041] The current train information is expanded based on the historical number of trains on the historical route and the current train information in the network feature information to obtain expanded train information;
[0042] Based on the expanded train information, the feature information to be expanded is expanded to obtain the expanded feature information.
[0043] Optionally, the characterization unit is used for:
[0044] Obtain a network point feature training set, wherein the network point feature training set includes multiple network point feature information samples and corresponding sample labels within a third historical time period;
[0045] The preset decision tree model is trained based on the network feature training set to obtain the network representation decision tree model.
[0046] Optionally, the characterization unit is used for:
[0047] The network feature training set is input into the preset decision tree model to obtain the leaf node features of the decision tree;
[0048] The leaf node features of the decision tree are input into a graph attention network to obtain the attention coefficients between each node.
[0049] Based on the attention coefficient, the leaf node features of the decision tree are updated through the back gradient, and the preset decision tree model is iterated and updated multiple times to obtain the dot representation decision tree model.
[0050] Optionally, the preset decision tree model is a gradient boosting decision tree model.
[0051] Optionally, the determining unit is configured to:
[0052] The network node representation information is input into the network node quantity prediction model to obtain the network node quantity of the target network node. The network node quantity prediction model is obtained by training a first preset linear model with a network node representation training set. The network node representation training set includes multiple network node representation information samples and corresponding sample labels.
[0053] Thirdly, this application provides a computer device, the computer device comprising:
[0054] One or more processors;
[0055] Memory; and
[0056] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the method for predicting the number of network points as described in any of the first aspects.
[0057] Fourthly, this application provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to perform the steps in the dot quantity prediction method described in any one of the first aspects.
[0058] This application provides a method and apparatus for predicting the quantity of items at a network point. The method includes: acquiring network point feature information of a target network point within a first historical time period; representing the network point feature information based on a network point representation decision tree model to obtain leaf node representation information of multiple decision trees; concatenating the leaf node representation information of the multiple decision trees with the network point feature information to obtain network point representation information; and determining the predicted quantity of items at the target network point based on the network point representation information. After acquiring the target network point feature information, this application represents the network point feature information according to a preset network point representation decision tree model to obtain leaf node representation information of multiple decision trees. Then, it concatenates the network point feature information with the leaf node representation information of the multiple decision trees represented by the tree model as network point representation information for network point quantity prediction. This allows for the extraction of more information from the network point feature information and improves the accuracy of network point quantity prediction. Attached Figure Description
[0059] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0060] Figure 1 This is a schematic diagram of a scenario for the prediction system for the number of items at a network point provided in an embodiment of this application;
[0061] Figure 2 This is a schematic flowchart of an embodiment of the method for predicting the number of items at a network point provided in this application.
[0062] Figure 3 This is a schematic diagram of an embodiment of the device for predicting the quantity of network items provided in this application.
[0063] Figure 4 This is a schematic diagram of an embodiment of the computer device provided in this application. Detailed Implementation
[0064] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0065] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more features. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.
[0066] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.
[0067] This application provides a method and apparatus for predicting the quantity of items at network points, which will be described in detail below.
[0068] Please see Figure 1 , Figure 1 This is a schematic diagram of a scenario for a network item quantity prediction system provided in an embodiment of this application. The network item quantity prediction system may include a computer device 100, which integrates a network item quantity prediction device.
[0069] In this embodiment, the computer device 100 can be a standalone server, a server network, or a server cluster. For example, the computer device 100 described in this embodiment includes, but is not limited to, a computer, a network host, a single network server, a set of multiple network servers, or a cloud server composed of multiple servers. The cloud server is composed of a large number of computers or network servers based on cloud computing.
[0070] In this embodiment, the computer device 100 described above can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device 100 can be a desktop computer, a portable computer, a network server, a handheld computer (Personal Digital Assistant, PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, an embedded device, etc. This embodiment does not limit the type of computer device 100.
[0071] Those skilled in the art will understand that Figure 1 The application environment shown is merely one application scenario of the solution in this application and does not constitute a limitation on the application scenario of the solution in this application. Other application environments may include more than one application scenario. Figure 1 The number of computer devices shown is more or less, for example Figure 1 Only one computer device is shown in the diagram. It is understood that the network item volume prediction system may also include one or more other computer devices capable of processing data, which are not specifically limited here.
[0072] In addition, such as Figure 1 As shown, the network item quantity prediction system may also include a memory 200 for storing data.
[0073] It should be noted that, Figure 1 The schematic diagram of the network point shipment volume prediction system shown is merely an example. The network point shipment volume prediction system and scenario described in this application embodiment are for the purpose of more clearly illustrating the technical solutions of this application embodiment and do not constitute a limitation on the technical solutions provided in this application embodiment. As those skilled in the art will know, with the evolution of the network point shipment volume prediction system and the emergence of new business scenarios, the technical solutions provided in this application embodiment are also applicable to similar technical problems.
[0074] First, this application provides a method for predicting the number of items at a network point. The method includes: obtaining network point feature information of the target network point within a first historical time period; representing the network point feature information based on a network point representation decision tree model to obtain the leaf node representation information of multiple decision trees; concatenating the leaf node representation information of the multiple decision trees with the network point feature information to obtain network point representation information; and determining the predicted number of items at the target network point based on the network point representation information.
[0075] like Figure 2 As shown, Figure 2 This is a schematic flowchart of an embodiment of the method for predicting the quantity of items at network outlets in this application. The method for predicting the quantity of items at network outlets includes the following steps S201 to S204:
[0076] S201. Obtain the network feature information of the target network point within the first historical time period.
[0077] The target point can be any point in the logistics system, such as a transit point, sales point, distribution center, airport, cold storage, etc. The first historical time period can be 7 days, 5 days, etc., depending on the specific situation.
[0078] Branch network characteristic information can include basic characteristic information and characteristic information to be expanded. Basic and expanded characteristic information can be categorized based on specific circumstances. For example, basic characteristic information may include: branch type, the time difference between the latest arrival time and the end time of the shift, whether packages are picked up or delivered, the proportion of express delivery products (next-day delivery, same-day delivery, etc.), and shift duration. Branch type can be a transit point, sales point, distribution center, airport terminal, cold storage, same-city point, or large-item consolidation point. Expanded characteristic information includes: branch order volume, number of employees, employee efficiency, space efficiency, estimated delivery time, and the number of orders transferred from the previous shift.
[0079] S202. Based on the network node representation decision tree model, the network node feature information is represented to obtain the leaf node representation information of multiple decision trees.
[0080] In this embodiment of the application, the network feature information is represented based on the network representation decision tree model to obtain the leaf node representation information of multiple decision trees. This may include: performing feature expansion on the feature information to be expanded to obtain expanded feature information; obtaining the network representation decision tree model; and inputting the expanded feature information and basic feature information into the network representation decision tree model to obtain the leaf node representation information of multiple decision trees.
[0081] In a specific embodiment, extending the feature information to be extended to obtain extended feature information may include:
[0082] (1) Obtain the current train information from the historical routes and network feature information that passed through the target network point in the second historical time period and did not pass through the target network point in the first historical time period.
[0083] The second historical time period refers to the time preceding the first historical time period. For example, the second historical time period is the number of Fibonacci sequence days preceding the first historical time period. For example, the second historical time period could be 1, 2, 3, 5, 8, 13, or 21 days prior to the first historical time period. The current train information in the network feature information includes the number of trains and their origin and destination. For example, if the target network point is A, historical routes that passed through the target network point within the second historical time period but did not pass through the target network point within the first historical time period include: within 1 day, there are 3 historical routes: BA, CA, and DA; within 3 days, there are 2 historical routes: EA and FA. The current train information contains 100 trains.
[0084] (2) Based on the historical number of trains on the historical route and the current train information in the network feature information, the current train information is expanded to obtain expanded train information.
[0085] Specifically, the process involves obtaining the historical number of trips for each historical route and the total number of historical trips. It also involves obtaining the percentage of historical trips for each historical route relative to the total number of historical trips. Based on this percentage and the current number of trips, expanded trip information is determined. For example, if the target destination is A, and the historical routes include BA, CA, and DA within one day, with route BA accounting for 3% of the total number of trips, and the current trip information contains 100 trips, then 3 routes BA are added to the current trip information to obtain expanded trip information.
[0086] (3) Expand the feature information to be expanded based on the expanded train information to obtain the expanded feature information.
[0087] Specifically, the order volume of each train service corresponding to the network points is obtained from the expanded train information and merged with the feature information to be expanded to obtain the expanded feature information. By expanding the features in the feature information to be expanded, the number of features can be increased, more possible situations can be covered, and the stability of the model can be improved. For example, for the route from Shenzhen to Beijing, there are sampling schemes such as Shenzhen-Zhengzhou-Shijiazhuang-Beijing, Shenzhen-Shijiazhuang-Beijing, and Shenzhen-Beijing. Moreover, there will be logistics anomalies such as Shenzhen-Shijiazhuang-Zhengzhou-Shijiazhuang-Beijing (cross-selling phenomenon when different logistics products are connected to the network). Expanding the number of features covers more possible situations and improves the stability of the model.
[0088] In a specific embodiment, obtaining a network node representation decision tree model may include: obtaining a network node feature training set, wherein the network node feature training set includes multiple network node feature information samples and corresponding sample labels within a third historical time period; and training a preset decision tree model based on the network node feature training set to obtain a network node representation decision tree model.
[0089] The third historical time period is longer than the first historical time period. For example, the third historical time period is one month, and the first historical time period is a 7-day period within the third historical time period. This allows multiple network feature information samples and corresponding sample labels to be obtained from the third historical time period.
[0090] The branch location feature information sample and its corresponding sample label can be the number of packages delivered to the branch within one day after the first historical time period; or it can be the ratio of the number of packages delivered to the branch within one day after the first historical time period to the number of packages delivered to the branch on the last day of the first historical time period. For example, the sample label corresponding to the branch location feature information sample can be one of three categories: (0, 0.8], (0.8, 1.3], or (1.3, +∞]. For example, a branch location feature information sample is the branch location feature information from August 1, 2021 to August 8, 2021, and the corresponding sample label is the ratio of the number of packages delivered on August 9, 2021 to August 8, 2021.
[0091] Optionally, the preset decision tree model is a Gradient Boosting Decision Tree (GBDT) model. Specifically, the dot feature training set is input into the preset decision tree model. The preset decision tree model builds a decision tree on the dot feature training set. Then, the target variable for each sample in the dot feature training set is adjusted to the residual between the predicted value and the actual value of the first decision tree, resulting in a new dot feature training set. A second decision tree is built under the new dot feature training set, and this process is iterated n times to obtain n decision trees and a dot representation decision tree model. During prediction, the data to be predicted is input into the dot representation decision tree model to obtain the leaf node representation information of the n decision trees.
[0092] In a specific embodiment, a preset decision tree model is trained based on a dot feature training set to obtain a dot representation decision tree model, including:
[0093] (1) Input the network feature training set into the preset decision tree model to obtain the leaf node features of the decision tree.
[0094] (2) Input the leaf node features of the decision tree into the graph attention network to obtain the attention coefficients between each node.
[0095] Graph Attention Networks (GAT) are essentially feature extractors, much like Graph Convolutional Networks (CNNs), but designed for graph data. GAT ingeniously designs a method for extracting features from graph data, allowing us to use these features for node classification, graph classification, edge prediction, and incidentally, graph embeddings. The biggest difference between GAT and CNNs is the introduction of an attention mechanism, assigning greater weight to more important nodes. In an end-to-end framework, attention weights and neural network parameters are learned together. In GAT, each node in the graph can be assigned different weights based on the features of its neighbors. Another advantage of GAT is that, after introducing the attention mechanism, it only concerns adjacent nodes, i.e., nodes sharing edges, without needing information about the entire graph.
[0096] Graph attention networks implement a self-attention mechanism for each node in the decision tree, resulting in attention coefficients. These attention coefficients represent the importance of neighboring nodes to the node.
[0097] (3) Based on the attention coefficient, the leaf node features of the decision tree are updated through the reverse gradient, and the preset decision tree model is iterated and updated multiple times to obtain the network representation decision tree model.
[0098] Specifically, through n iterations, the representation information of n decision trees and the node representation decision tree model are obtained. On the one hand, the graph attention network extracts attention coefficients that can reflect the hidden relationships between nodes, and then combines the attention coefficients of the graph structure and the leaf node features of the tree model as representation information for subsequent prediction, which is more accurate than the graph model and tree model alone.
[0099] S203. The leaf node representation information and network feature information of multiple decision trees are spliced together to obtain the network representation information.
[0100] Specifically, the concat function is used to concatenate the leaf node representation information, basic feature information, and extended feature information of multiple decision trees to obtain the network node representation information.
[0101] S204. Determine the predicted quantity of target network nodes based on network node characterization information.
[0102] In a specific embodiment, determining the predicted quantity information of a target network node based on network node representation information includes: inputting the network node representation information into a network node quantity prediction model to obtain the predicted quantity information of the target network node, wherein the network node quantity prediction model is obtained by training a first preset linear model with a network node representation training set, and the network node representation training set includes multiple network node representation information samples and corresponding sample labels.
[0103] The sample label corresponding to the branch representation information sample can be the number of packages at the branch within one day after the first historical time period; or it can be the ratio of the number of packages at the branch within one day after the first historical time period to the number of packages at the branch on the last day of the first historical time period. Correspondingly, the predicted package volume information for the branch can be either the number of packages at the branch or the ratio of the number of packages.
[0104] The first preset linear model can be the Logistic Regression (LR) model. Logistic Regression is a classification model in machine learning. The algorithm is simple and efficient, and using a linear model similar to LR can meet practical needs under massive data.
[0105] Furthermore, personnel feature information of multiple individuals is extracted from the network feature information. This personnel feature information is then input into a second preset linear model to obtain the scheduling probability of each individual. Individuals with a scheduling probability greater than a preset value are identified as scheduling personnel. The personnel feature information may include personnel efficiency indicators, personnel work time arrangement indicators, and personnel performance indicators. The second preset linear model can be a pre-trained LR model.
[0106] Furthermore, the output includes the quantity ratio and the shift scheduler. For example, the output [1.3; AA, BB] indicates that the predicted quantity for the day is 1.3 times that of yesterday, and the corresponding shift schedulers are AA and BB.
[0107] To better implement the method for predicting the quantity of items at network points in the embodiments of this application, based on the method for predicting the quantity of items at network points, the embodiments of this application also provide a device for predicting the quantity of items at network points, such as... Figure 3 As shown, the dot quantity prediction device 300 includes:
[0108] The acquisition unit 301 is used to acquire the network feature information of the target network point within the first historical time period.
[0109] The representation unit 302 is used to represent the feature information of the network nodes based on the network node representation decision tree model, and obtain the leaf node representation information of multiple decision trees.
[0110] The splicing unit 303 is used to splice the leaf node representation information and network feature information of multiple decision trees to obtain network representation information;
[0111] The determining unit 304 is used to determine the predicted quantity information of the target network point based on the network point characterization information.
[0112] Optionally, the branch network feature information includes basic feature information and feature information to be expanded;
[0113] Characterization unit 302 is used for:
[0114] The feature information to be expanded is expanded to obtain the expanded feature information;
[0115] Obtain the network representation decision tree model;
[0116] The extended feature information and basic feature information are input into the network node representation decision tree model to obtain the leaf node representation information of multiple decision trees.
[0117] Optionally, the characterization unit 302 is used for:
[0118] Obtain the current train information from the historical routes and network feature information that passed through the target network point in the second historical time period but did not pass through the target network point in the first historical time period;
[0119] The current train information is expanded based on the historical train number and network feature information of the historical route to obtain expanded train information;
[0120] Based on the expanded train information, the feature information to be expanded is expanded to obtain the expanded feature information.
[0121] Optionally, the characterization unit 302 is used for:
[0122] Obtain a training set of network point features, which includes multiple network point feature information samples and corresponding sample labels within the third historical time period;
[0123] The pre-defined decision tree model is trained based on the network feature training set to obtain the network representation decision tree model.
[0124] Optionally, the characterization unit 302 is used for:
[0125] Input the network feature training set into the preset decision tree model to obtain the leaf node features of the decision tree;
[0126] The leaf node features of the decision tree are input into the graph attention network to obtain the attention coefficients between the nodes.
[0127] Based on the attention coefficient, the leaf node features of the decision tree are updated through the back gradient, and the preset decision tree model is iterated and updated multiple times to obtain the network representation decision tree model.
[0128] Optionally, the default decision tree model is a gradient boosting decision tree model.
[0129] Optionally, determining unit 304 is used for:
[0130] The network node representation information is input into the network node item quantity prediction model to obtain the network node item quantity of the target network node. The network node item quantity prediction model is obtained by training the first preset linear model with the network node representation training set. The network node representation training set includes multiple network node representation information samples and corresponding sample labels.
[0131] This application also provides a computer device that integrates any of the network item quantity prediction devices provided in this application. The computer device includes:
[0132] One or more processors;
[0133] Memory; and
[0134] One or more applications, wherein the applications are stored in memory and configured to be executed by a processor as steps of the dot quantity prediction method in any of the embodiments described above.
[0135] like Figure 4 As shown, it illustrates a structural schematic diagram of the computer device involved in the embodiments of this application, specifically:
[0136] The computer device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that the computer device structure shown in the figures does not constitute a limitation on the computer device, and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:
[0137] Processor 401 is the control center of the computer device, connecting various parts of the computer device through various interfaces and routes. It performs various functions and processes data by running or executing software programs and / or modules stored in memory 402, and by calling data stored in memory 402, thereby providing overall monitoring of the computer device. Optionally, processor 401 may include one or more processing cores; processor 401 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. Preferably, processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the aforementioned modem processor may not be integrated into processor 401.
[0138] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.
[0139] The computer device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.
[0140] The computer device may also include an input unit 404, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.
[0141] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the computer device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402 to realize various functions, as follows:
[0142] Obtain the target network's characteristic information within the first historical time period;
[0143] The network feature information is represented based on the network representation decision tree model, and the leaf node representation information of multiple decision trees is obtained.
[0144] The node representation information is obtained by concatenating the leaf node representation information and node feature information of multiple decision trees;
[0145] Based on the network representation information, the predicted quantity of items for the target network is determined.
[0146] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.
[0147] Therefore, embodiments of this application provide a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), a disk, or an optical disk, etc. A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in any of the dot quantity prediction methods provided in embodiments of this application. For example, the computer program loaded by the processor can execute the following steps:
[0148] Obtain the target network's characteristic information within the first historical time period;
[0149] The network feature information is represented based on the network representation decision tree model, and the leaf node representation information of multiple decision trees is obtained.
[0150] The node representation information is obtained by concatenating the leaf node representation information and node feature information of multiple decision trees;
[0151] Based on the network representation information, the predicted quantity of items for the target network is determined.
[0152] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the detailed descriptions of other embodiments above, which will not be repeated here.
[0153] In practice, each of the above units or structures can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units or structures, please refer to the previous method embodiments, which will not be repeated here.
[0154] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.
[0155] The above provides a detailed description of a method and apparatus for predicting the quantity of network items provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for predicting the quantity of items at a network point, characterized in that, The method for predicting the quantity of items at network points includes: Obtain the target network's characteristic information within the first historical time period; The network feature information is represented based on the network representation decision tree model to obtain the leaf node representation information of multiple decision trees; The leaf node representation information of multiple decision trees and the network feature information are concatenated to obtain the network representation information; Based on the network representation information, the predicted quantity of parts for the target network is determined; The network point feature information includes basic feature information and feature information to be expanded. The process of representing the network point feature information based on the network point representation decision tree model to obtain leaf node representation information of multiple decision trees includes: obtaining historical routes that passed through the target network point within a second historical time period but did not pass through the target network point within the first historical time period, and current train information in the network point feature information; expanding the current train information based on the historical train count of the historical routes and the current train information in the network point feature information to obtain expanded train information; expanding the feature information to be expanded based on the expanded train information to obtain expanded feature information; obtaining the network point representation decision tree model; and inputting the expanded feature information and the basic feature information into the network point representation decision tree model to obtain leaf node representation information of multiple decision trees.
2. The method for predicting the quantity of items at network outlets according to claim 1, characterized in that, The network node representation decision tree model includes: Obtain a network point feature training set, wherein the network point feature training set includes multiple network point feature information samples and corresponding sample labels within a third historical time period; The preset decision tree model is trained based on the network feature training set to obtain the network representation decision tree model.
3. The method for predicting the quantity of items at network outlets according to claim 2, characterized in that, The step of training a preset decision tree model based on the network feature training set to obtain the network representation decision tree model includes: The network feature training set is input into the preset decision tree model to obtain the leaf node features of the decision tree; The leaf node features of the decision tree are input into a graph attention network to obtain the attention coefficients between each node. Based on the attention coefficient, the leaf node features of the decision tree are updated through the back gradient, and the preset decision tree model is iterated and updated multiple times to obtain the dot representation decision tree model.
4. The method for predicting the quantity of items at network outlets according to claim 3, characterized in that, The preset decision tree model is a gradient boosting decision tree model.
5. The method for predicting the quantity of items at network outlets according to claim 1, characterized in that, The step of determining the predicted quantity of the target network node based on the network node characterization information includes: The network node representation information is input into the network node quantity prediction model to obtain the network node quantity of the target network node. The network node quantity prediction model is obtained by training a first preset linear model with a network node representation training set. The network node representation training set includes multiple network node representation information samples and corresponding sample labels.
6. A device for predicting the quantity of dots, characterized in that, The device for predicting the quantity of network items includes: The acquisition unit is used to acquire the network feature information of the target network point within the first historical time period. The representation unit is used to represent the feature information of the network nodes based on the network node representation decision tree model, and obtain the leaf node representation information of multiple decision trees. The splicing unit is used to splice the leaf node representation information of multiple decision trees and the network feature information to obtain network representation information; The determining unit is used to determine the predicted quantity information of the target network point based on the network point characterization information; The network point feature information includes basic feature information and feature information to be expanded. The representation unit is further configured to: acquire historical routes that passed through the target network point within a second historical time period but did not pass through the target network point within the first historical time period, and current train information in the network point feature information; expand the current train information based on the historical number of trains on the historical routes and the current train information in the network point feature information to obtain expanded train information; expand the feature information to be expanded based on the expanded train information to obtain expanded feature information; acquire a network point representation decision tree model; and input the expanded feature information and the basic feature information into the network point representation decision tree model to obtain leaf node representation information of multiple decision trees.
7. A computer device, characterized in that, The computer device includes: One or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the method for predicting the number of network points as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, It stores a computer program, which is loaded by a processor to execute the steps of the method for predicting the quantity of network items according to any one of claims 1 to 5.