A method and device for generating origin-destination population flow data

Abstract

Embodiments of the present application provide a kind of starting point and end point population flow data generation method and device. Among them, the method comprises: the graph model corresponding to the target area is constructed, and graph neural network model is established according to the graph model;Using the historical population distribution data and the starting point and end point population flow data of each region in the target area, the graph neural network model is trained;Using the graph neural network model obtained after training, the population flow data of the target area is generated.Relative to prior art, the model of the embodiments of the present application can generate higher accuracy population flow data based on relatively easy to collect population distribution data, easy to implement engineering, with the advantages of small amount of calculation, lower implementation cost and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of population flow analysis, and in particular, relates to a method and device for generating origin-destination population flow data. BACKGROUND

[0002] Origin-destination (OD) population flow data (also referred to as OD flow data herein) is a variety of short-term, repeated or periodic movements of population between geographical areas. The OD population flow data is widely used in the fields of road traffic management and planning, route recommendation and public safety services.

[0003] There are two main ways to obtain OD flow data in the prior art, including:

[0004] One way is to obtain the data by using a gravity model. The gravity model is a widely used trip distribution model, and is named after its similar form to Newton's law of gravity. The gravity model analogizes the number of populations in two regions to the mass of two objects in the law of gravity, and analogizes the population flow between the two regions to the attraction between the two objects in the law of gravity, and determines that the number of trips Tp between the two regions is proportional to the trip generation number of the departure region and proportional to the square of the traffic impedance between the two regions. The gravity model uses simple universal gravitation to analogize the complex population flow between regions, and therefore the expression ability of the model is not enough, and the precision is low, which is difficult to meet the requirements of precision in actual applications.

[0005] Another way to obtain OD flow data in the prior art is to generate OD flow data based on trajectory data of population movement. Specifically, the actual movement position information of people and vehicles is collected by using satellite positioning and public wireless network positioning technology, and then the movement trajectories of all individuals over time are generated using the collected actual movement position information, and the aggregated population flow information between regions is obtained by statistics. The above-mentioned method can obtain high-precision flow data, but the input data of the model is strictly required, and the conditions are harsh. It is necessary to obtain the movement trajectories of most individuals at all times to obtain the flow data, and therefore the model implementation cost is high, and it is not easy to implement engineering.

[0006] Due to the above-mentioned shortcomings of the prior art, there is an urgent need for a scheme capable of generating population flow data with relatively high precision and easy to implement engineering. SUMMARY

[0007] The technical problem to be solved by the embodiments of the present application is to provide a method and device for generating origin-destination population flow data, so as to improve the quality of the generated population flow data at a low implementation cost.

[0008] To solve the above technical problems, according to one aspect of the present application, a method for generating origin-destination population flow is provided, comprising:

[0009] constructing a graph model corresponding to a target region, and establishing a graph neural network model according to the graph model, wherein each region in the target region corresponds to a node in the graph model, and the physical accessibility between two regions is represented by a weighted edge between the nodes corresponding to the two regions;

[0010] training the graph neural network model using historical population distribution data and origin-destination population flow data of each region in the target region, wherein the graph neural network model generates directed population flow data between regions corresponding to two nodes by fusing the abstract features of the regions corresponding to the two nodes; the abstract feature of the node is an abstract expression of the population quantity change information of the node and adjacent nodes fused after graph neural network calculation;

[0011] generating population flow data of the target region using the trained graph neural network model

[0012] In addition, according to at least one embodiment of the present application, the feature input of each node is the historical population quantity of multiple time periods, and a convolutional neural network based on attention mechanism is used for feature extraction to obtain historical population quantity change information.

[0013] In addition, according to at least one embodiment of the present application, each node collects historical population quantity change information of adjacent nodes in the process of graph neural network calculation, so that the abstract feature of each node contains topological information of the graph.

[0014] In addition, according to at least one embodiment of the present application, the step of fusing the abstract features of the regions corresponding to two nodes includes:

[0015] performing vector inner product calculation on the abstract features of the two nodes; or,

[0016] performing vector splicing on the abstract features of the two nodes, and mapping the spliced vector to one dimension.

[0017] In addition, according to at least one embodiment of the present application, the step of training the graph neural network model using historical population distribution data and origin-destination population flow data of each region in the target region includes:

[0018] The following training is repeatedly performed using training data until the graph neural network model meets a preset convergence condition, the training data including a plurality of training samples, each training sample including historical population distribution data and corresponding population flow data of regions in the target area:

[0019] The historical population distribution data in the sample is input into the graph neural network model to obtain population flow data output by the graph neural network model;

[0020] An error between the population flow data in the sample and the population flow data generated by the graph neural network model is calculated, and the model parameters of the graph neural network model are adjusted using a gradient descent algorithm according to the error.

[0021] In addition, according to at least one embodiment of the present application, before the step of generating population flow data of the target area using the graph neural network model obtained after training, the method further comprises:

[0022] The graph neural network model obtained after training is tested using a test sample including historical population distribution data and corresponding population flow data of regions in the target area to determine whether the graph neural network model obtained after training meets a preset performance index requirement.

[0023] According to another aspect of the present application, a starting point and ending point population flow generation device is also provided, comprising:

[0024] A modeling unit is configured to construct a graph model corresponding to a target area and establish a graph neural network model according to the graph model, wherein each region in the target area corresponds to a node in the graph model, and physical accessibility between two regions is represented by a weighted edge between nodes corresponding to the two regions;

[0025] A training unit is configured to train the graph neural network model using historical population distribution data and starting point and ending point population flow data of regions in the target area, wherein the graph neural network model generates directed population flow data between regions corresponding to two nodes by fusing abstract features of the two regions, and the abstract features of the nodes are abstract expressions of population quantity change information of the nodes and adjacent nodes fused after graph neural network calculation;

[0026] A generation unit is configured to generate population flow data of the target area using the graph neural network model obtained after training.

[0027] In addition, according to at least one embodiment of the present application, the feature input of each node is the historical population quantity of a plurality of time periods, and a convolutional neural network based on an attention mechanism is used for feature extraction to obtain historical population quantity change information.

[0028] In addition, according to at least one embodiment of the present application, each node collects historical population quantity change information of adjacent nodes in the process of graph neural network calculation, so that the abstract features of each node contain topological information of the graph.

[0029] In addition, according to at least one embodiment of the present application, the graph neural network model comprises a fusion unit for fusing the abstract features of the corresponding regions of the two nodes in the following manner:

[0030] performing vector inner product calculation on the abstract features of the two nodes; or,

[0031] performing vector splicing on the abstract features of the two nodes, and mapping the spliced vector to one dimension.

[0032] In addition, according to at least one embodiment of the present application, the training unit is further configured to:

[0033] repeatedly performing the following training using training data until the graph neural network model meets a preset convergence condition, the training data comprising a plurality of training samples, each training sample comprising historical population distribution data and corresponding population flow data of each region in the target area:

[0034] inputting the historical population distribution data in the sample into the graph neural network model to obtain population flow data output by the graph neural network model;

[0035] calculating the error between the population flow data in the sample and the population flow data generated by the graph neural network model, and adjusting the model parameters of the graph neural network model using a gradient descent algorithm according to the error.

[0036] In addition, according to at least one embodiment of the present application, the generation device further comprises:

[0037] a verification unit configured to test the graph neural network model obtained after training using a test sample comprising historical population distribution data and corresponding population flow data of each region in the target area, and determine whether the graph neural network model obtained after training meets a preset performance index requirement.

[0038] The embodiment of the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the population flow data generation method described above.

[0039] Compared with the prior art, the population flow data generation method and device provided by the embodiment of the present application has at least the following beneficial effects: the generation method of the embodiment of the present application can generate population flow data with higher accuracy based on population distribution data which is relatively easy to collect, is easy to implement in engineering, has the advantages of smaller calculation amount and lower implementation cost, etc. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0041] Figure 1 A system block diagram of the OD population flow data generation device of the embodiment of the present application;

[0042] Figure 2 A schematic diagram of an algorithm framework used by the embodiment of the present application;

[0043] Figure 3 A schematic diagram of the graph neural network model training process of the embodiment of the present application;

[0044] Figure 4 A schematic diagram of the embodiment of the present application for generating population flow data using a graph neural network model;

[0045] Figure 5 A structural schematic diagram of the OD population flow generation device provided by the embodiment of the present application;

[0046] Figure 6 Another structural schematic diagram of the OD population flow generation device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the technical problems, technical solutions and advantages of the present application clearer, the following will be described in detail in conjunction with the drawings and specific embodiments. In the following description, specific details such as specific configurations and components are provided only to help a comprehensive understanding of the embodiments of the present application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, in order to be clear and concise, the description of known functions and structures is omitted.

[0048] It should be understood that the reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described is included in at least one embodiment of the application. Therefore, appearances of "in one embodiment" or "in an embodiment" at various places throughout the specification are not necessarily referring to the same embodiment. Furthermore, various particular features, structures, or characteristics can be combined in one or more embodiments.

[0049] In various embodiments of the present application, it should be understood that the size of the serial number of the following processes does not mean the order of execution, and the execution order of the processes should be determined by their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0050] As described in the background, the existing population flow generation method has the problem that the solution with low implementation cost has poor generation result accuracy, and the solution with high accuracy has high implementation cost. In order to solve at least one of the above problems, the embodiments of the present application provide a population flow generation method, which can improve the accuracy of the population flow generation result and reduce the implementation cost of the population flow data generation by making a trade-off between the implementation cost and the accuracy of the generation result.

[0051] The population flow generation method provided by the embodiments of the present application can generate OD population flow data between each region in a target area. Here, the target area can be a plurality of geographic ranges under the jurisdiction of a certain city, and the region is a subset of the geographic range under the jurisdiction of the city, and there is usually no overlapping part between each region. The area and boundary of the region can be divided according to the needs of population flow statistics. In the embodiments of the present application, the target area can also be a larger geographic range including a plurality of cities, and the region can be the geographic range under the jurisdiction of each city. At this time, the population flow reflects the flow change rule of the population between each city. Please refer to Figure 1 The OD population flow generation method provided by the embodiments of the present application includes:

[0052] Step 11, constructing a graph model corresponding to the target area, and establishing a graph neural network model according to the graph model, wherein each region in the target area corresponds to a node in the graph model, and the physical accessibility between two regions is represented by the weighted edge between the nodes corresponding to the two regions.

[0053] Here, the embodiment of the present application introduces a graph model in the population flow data generation, takes each region as a node in the graph model, and constructs a graph neural network model based on the graph model, which includes a graph neural network (GNN, Graph Neural Networks). In the graph model, the information of the weighted edge (edge with weight value) between any two nodes represents the physical accessibility between regions. The accessibility is related to the distance between regions and the traffic network between regions. Generally, the shorter the distance and the denser the traffic network, the better the accessibility.

[0054] In step 12, the graph neural network model is trained by using the historical population distribution data of each region in the target area and the origin-destination population flow data.

[0055] Here, the graph neural network model generates the directed population flow data between the regions corresponding to the two nodes by fusing the abstract features of the regions corresponding to the two nodes. The abstract feature of the node is an abstract expression (vector expression) of the population quantity change information of the node and adjacent nodes fused after graph neural network calculation. Before the graph neural network calculation, the feature input of each node is the historical population quantity of a plurality of historical time periods (also referred to as time slices) of the corresponding region. The embodiment of the present application can use a convolutional neural network based on an attention mechanism to perform feature extraction to obtain the historical population quantity change information.

[0056] In the graph neural network calculation, each node collects the historical population quantity change information of adjacent nodes in the process of graph neural network calculation, so that the abstract feature of each node contains the topological information of the graph. In this way, each node generates a new feature expression (i.e., the abstract feature of the region corresponding to the node) by integrating the historical population quantity change information of the node and adjacent nodes, and the feature expression contains the population quantity change information of the region corresponding to the node and the population quantity change information of other regions corresponding to adjacent nodes. The information of the weighted edge between nodes is the physical accessibility. The higher the accessibility, the greater the weight, and the greater the influence between adjacent nodes connected by the edge during calculation.

[0057] In step 12, the embodiment of the present application uses the historical population distribution data of each region in the target area and the OD population flow data collected in advance to perform supervised training on the graph neural network model. Here, the training data includes a plurality of samples of the target area collected in history, each sample includes historical population distribution data of each region in the target area and corresponding population flow data in a certain historical period, and the corresponding population flow data can be obtained by manual statistics or other means.

[0058] Specifically, in step 12, the embodiment of the present application repeatedly performs the following training using the training data until the graph neural network model meets the preset convergence condition:

[0059] A) inputting historical population distribution data in the sample into the graph neural network model to obtain population flow data output by the graph neural network model;

[0060] B) calculating an error between the population flow data in the sample and the population flow data generated by the graph neural network model, and adjusting model parameters of the graph neural network model using a gradient descent algorithm according to the error

[0061] Step 13, using the graph neural network model obtained after training, generating population flow data of the target region.

[0062] Here, when generating population flow data using the graph neural network model obtained after training, population distribution data of the target region collected in a target time period (such as the last time) can be input into the graph neural network model, and then the generation result of the population flow data can be obtained according to the output of the graph neural network model.

[0063] Through the above steps, the embodiment of the present application introduces a graph model and a graph neural network model into population flow data generation, and trains the model based on historical population distribution data and population flow data of the region in the model algorithm implementation. Since the model training does not require related data of individual moving trajectories, the embodiment of the present application can simplify the acquisition of training data, reduce the requirements for training data, and thus reduce the implementation cost of the model. Moreover, compared with the traditional gravity model, the graph model and the graph neural network model of the embodiment of the present application introduce the attribute features of the nodes and the connection features between the nodes to train the model, optimize the expression ability of the image to the population flow parameters, and can improve the quality of the model human flow data generation.

[0064] According to at least one embodiment of the present application, between steps 12 and 13 described above, the embodiment of the present application can also test the graph neural network model obtained after training using a test sample including historical population distribution data and corresponding population flow data of each region in the target region, to determine whether the graph neural network model obtained after training meets the preset performance index requirement. If the preset performance index requirement is not met, the model can be further trained using more training data.

[0065] Figure 2 An example of an algorithm framework used by the population flow generation device of the embodiment of the present application is provided. As shown in FIG. 6, the algorithm framework includes the following steps: Figure 2As shown, the storage unit stores historical population distribution data and OD population flow data of each region in the target region, part of which is provided as training data for model training, and the other part can be used as test data to test and verify the performance of the trained model. As shown in Figure 2 As shown, the embodiment of the present application constructs a fully connected graph model based on the traffic network between regions in the target region and the distance between regions, wherein each node represents a region, the feature input of each node is the historical population number of multiple time periods, and a convolutional neural network based on attention mechanism is used for feature extraction to obtain historical population number change information. In addition, each node collects historical population number change information of adjacent nodes in the process of graph neural network calculation, so that the abstract feature of each node contains the topological information of the graph. In this way, the abstract feature of each node is an abstract expression that integrates the population number change information of the node and adjacent nodes after graph neural network calculation.

[0066] Figure 2 As shown in the middle, the abstract feature (Region embedding) of each node is generated based on historical population distribution data of multiple historical periods, such as t, t-1 and t-2, etc. Then, by traversing, the abstract features between each two nodes are fused, such as vector inner product calculation of the abstract features of the two nodes; or vector splicing of the abstract features of the two nodes, and mapping the spliced vector to one dimension, so that OD pairs can be obtained, each OD pair can reflect the directed population flow between the regions corresponding to the two nodes.

[0067] Continuing to refer to Figure 2 After the graph neural network model is trained, the embodiment of the present application can test the above-mentioned model by using test samples, that is, input the historical population distribution data in the test samples into the model, and then compare the error between the model output result and the population flow data in the test samples to verify the performance of the model.

[0068] Please refer to Figure 3 When the embodiment of the present application trains the model using the training data, the specific process includes:

[0069] 301, model initialization, including constructing a graph model and establishing a graph neural network model, and initializing model parameters.

[0070] 302-303, sample historical population distribution data x and its corresponding OD flow data y from the input training data;

[0071] 304-305, input x into the graph neural network model to obtain the OD flow data generated by the model Calculate the output of the model error loss with respect to the real data y, which can be referred to as Figure 2 the bottom right corner of the algorithm framework diagram.

[0072] 306, based on the error loss, adjusting the neural network related parameters of the graph neural network model using a gradient descent algorithm.

[0073] 307, determining whether the error loss converges, if not, repeating the steps 304-306 until the error loss is low to a certain extent and converges.

[0074] 308, saving the trained graph neural network model.

[0075] In addition, before step 308, the embodiment of the present application can also input the historical population change data in the test data into the graph neural network model after the error loss converges, and determine the error between the output of the model and the real OD flow data corresponding to the input test data, to determine whether the performance of the model meets the requirements.

[0076] Figure 4 Further, the steps of generating population flow data using the trained model are given, which specifically include:

[0077] 401, selecting a graph neural network model corresponding to a target area.

[0078] 402, inputting population distribution data into the above model.

[0079] 403, recovering OD data from the above model and outputting an OD matrix, the OD matrix being used to represent directional population flow data between regions.

[0080] As can be seen from the above, the embodiment of the present application can obtain a relatively accurate population flow data generation result based on relatively loose training data, is easy to implement in engineering, has the advantages of small amount of calculation, low implementation cost, etc.

[0081] Based on the above population flow generation method, the embodiment of the present application also provides a device for implementing the above method.

[0082] Please refer to Figure 5 The embodiment of the present application provides a starting point and end point population flow generation device 50, which comprises:

[0083] The modeling unit 51 is used to construct a graph model corresponding to a target area, and to establish a graph neural network model according to the graph model, wherein each region in the target area corresponds to a node in the graph model, and the physical accessibility between two regions is represented by the weighted edge between the nodes corresponding to the two regions.

[0084] The training unit 52 is configured to train the graph neural network model by using historical population distribution data and origin-destination population flow data of each region in the target region, wherein the graph neural network model generates directional population flow data between two nodes corresponding regions by fusing abstract features of the two nodes corresponding regions; and the abstract feature of the node is an abstract expression of population quantity change information of the node and adjacent nodes after graph neural network calculation.

[0085] The generating unit 53 is configured to generate population flow data of the target region by using the trained graph neural network model.

[0086] According to the above units, the embodiment of the present application can generate population flow generation results more accurately based on training data which is relatively easy to obtain.

[0087] According to at least one embodiment of the present application, the feature input of each node is historical population quantity of multiple time periods, and a convolutional neural network based on an attention mechanism is used for feature extraction to obtain historical population quantity change information.

[0088] According to at least one embodiment of the present application, each node collects historical population quantity change information of adjacent nodes in the process of graph neural network calculation, so that the abstract feature of each node contains topological information of the graph.

[0089] According to at least one embodiment of the present application, the graph neural network model comprises a fusion unit configured to fuse abstract features of two nodes corresponding regions in the following manner:

[0090] performing vector inner product calculation on the abstract features of the two nodes; or,

[0091] performing vector splicing on the abstract features of the two nodes, and mapping the spliced vector to one dimension.

[0092] According to at least one embodiment of the present application, the training unit is further configured to:

[0093] repeatedly perform the following training by using training data until the graph neural network model meets a preset convergence condition, wherein the training data comprises multiple training samples, and each training sample comprises historical population distribution data and corresponding population flow data of each region in the target region:

[0094] inputting the historical population distribution data in the sample into the graph neural network model to obtain population flow data output by the graph neural network model;

[0095] Calculate an error between the population flow data in the sample and the population flow data generated by the graph neural network model, and adjust model parameters of the graph neural network model by using a gradient descent algorithm according to the error.

[0096] According to at least one embodiment of the present application, the generation device further comprises:

[0097] A verification unit is configured to test the trained graph neural network model by using a test sample comprising historical population distribution data and corresponding population flow data of each region in the target area, and determine whether the trained graph neural network model meets a preset performance index requirement.

[0098] As Figure 6 shown, the embodiment of the present application further provides another population flow data generation device 60, which specifically comprises a processor 61, a memory 62, a bus system 63, a receiver 64 and a transmitter 65. The processor 61, the memory 62, the receiver 64 and the transmitter 65 are connected through the bus system 63. The memory 62 is configured to store instructions, and the processor 61 is configured to execute the instructions stored in the memory 62 to control the receiver 64 to receive signals and control the transmitter 65 to transmit signals.

[0099] The processor 61 is configured to read programs in the memory and execute the following processes:

[0100] A graph model corresponding to the target area is constructed, and a graph neural network model is established according to the graph model. Each region in the target area corresponds to a node in the graph model, and the physical accessibility between two regions is represented by a weighted edge between nodes corresponding to the two regions.

[0101] The graph neural network model is trained by using historical population distribution data of each region in the target area and origin-destination population flow data. The graph neural network model generates directed population flow data between regions corresponding to two nodes by fusing abstract features of the regions corresponding to the two nodes. The abstract feature of the node is an abstract expression of population quantity change information of the node and adjacent nodes after graph neural network calculation.

[0102] The graph neural network model is trained by using historical population distribution data of each region in the target area and origin-destination population flow data. The graph neural network model generates directed population flow data between regions corresponding to two nodes by fusing abstract features of the regions corresponding to the two nodes. The abstract feature of the node is an abstract expression of population quantity change information of the node and adjacent nodes after graph neural network calculation.

[0103] It can be understood that, when the program is executed by the processor 61, the processes of the method embodiment shown in Figure 1 can be realized, and the same technical effects can be achieved. To avoid repetition, details are not described here.

[0104] It should be understood that the processor 61 can be a central processing unit (CPU) in the embodiments of the present application, and can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), ready-to-program gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or can also be any conventional processor.

[0105] The memory 62 can include read-only memory and random access memory, and provide instructions and data to the processor 61. Part of the memory 62 can also include non-volatile random access memory. For example, the memory 62 can also store device type information.

[0106] The bus system 63 can include not only a data bus, but also a power bus, a control bus, a status signal bus, etc. However, for the sake of clarity, all the buses are marked as the bus system 63 in the figure.

[0107] In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 61 or the instructions in the form of software. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as hardware processor execution completion, or executed by a combination of hardware and software modules in the processor. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, or other mature storage media in the art. The storage medium is located in the memory 62, and the processor 61 reads the information in the memory 62 and combines the hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0108] In some embodiments of the present application, a computer readable storage medium is also provided, which stores a program that can implement the following steps when executed by a processor:

[0109] A graph model corresponding to the target region is constructed, and a graph neural network model is established according to the graph model, wherein each region in the target region corresponds to a node in the graph model, and the physical accessibility between two regions is represented by the weighted edge between the nodes corresponding to the two regions.

[0110] The graph neural network model is trained by using historical population distribution data and origin-destination population flow data of each region in the target region, wherein the graph neural network model generates directed population flow data between two regions corresponding to two nodes by fusing abstract features of the two regions; and the abstract feature of the node is an abstract expression of population quantity change information of the node and adjacent nodes obtained by graph neural network calculation.

[0111] The graph neural network model obtained after training is used to generate population flow data of the target region.

[0112] The program is executed by the processor to realize Figure 1 All implementation manners in the population flow generation method shown in the description and can achieve the same technical effects. To avoid repetition, it will not be described here.

[0113] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0114] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0115] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division, and actual implementation can have another division manner. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0116] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the present application.

[0117] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0118] If the functions are realized in the form of software function units and sold or used as independent products, the functions can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0119] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for generating population flow data from origin to destination, characterized in that, include: Construct a graph model corresponding to the target region, and establish a graph neural network model based on the graph model, wherein each region in the target region corresponds to a node in the graph model, and the physical reachability between two regions is represented by the weighted edge between the nodes corresponding to the two regions; The graph neural network model is trained using historical population distribution data and origin-endpoint population flow data of each region within the target area. The graph neural network model generates directed population flow data between the regions corresponding to two nodes by fusing the abstract features of the regions corresponding to the two nodes. The abstract features of a node are abstract expressions that integrate population change information of the current node and its neighboring nodes, obtained after calculation by the graph neural network. Using the trained graph neural network model, population flow data for the target region is generated; The feature input for each node is the historical population size over multiple time periods, and a convolutional neural network based on an attention mechanism is used to extract features to obtain information on changes in the historical population size. During the computation of the graph neural network, each node collects historical population change information of its neighboring nodes so that the abstract features of each node contain the topological information of the graph.

2. The generation method as described in claim 1, characterized in that, The steps for fusing the abstract features of corresponding regions of two nodes include: Calculate the vector inner product of the abstract features of the two nodes; or, The abstract features of the two nodes are concatenated into vectors, and the concatenated vectors are mapped to one dimension.

3. The generation method as described in claim 1, characterized in that, The steps for training the graph neural network model using historical population distribution data and origin-destination population flow data for each region within the target area include: The following training process is repeated using the training data until the graph neural network model meets the preset convergence condition. The training data includes multiple training samples, each of which includes historical population distribution data and corresponding population flow data for each region within the target area: The historical population distribution data in the sample is input into the graph neural network model to obtain the population flow data output by the graph neural network model; The error between the population flow data in the sample and the population flow data generated by the graph neural network model is calculated, and the model parameters of the graph neural network model are adjusted using the gradient descent algorithm based on the error.

4. The generation method as described in claim 1, characterized in that, Before the step of generating population flow data for the target region using the trained graph neural network model, the method further includes: Using test samples including historical population distribution data and corresponding population flow data of each region within the target area, the trained graph neural network model is tested to determine whether the trained graph neural network model meets the preset performance index requirements.

5. A device for generating population flow data, characterized in that, include: A modeling unit is used to construct a graph model corresponding to the target region and establish a graph neural network model based on the graph model. Each region in the target region corresponds to a node in the graph model, and the physical reachability between two regions is represented by a weighted edge between the nodes corresponding to the two regions. The training unit is used to train the graph neural network model using historical population distribution data and origin-endpoint population flow data of each region within the target area. The graph neural network model generates directed population flow data between the regions corresponding to the two nodes by fusing the abstract features of the regions corresponding to the two nodes. The abstract features of the nodes are abstract expressions that fuse the population change information of the current node and its neighboring nodes after being calculated by the graph neural network. The generation unit is used to generate population flow data of the target area using the trained graph neural network model; The feature input for each node is the historical population size over multiple time periods, and a convolutional neural network based on an attention mechanism is used to extract features to obtain information on changes in the historical population size. During the computation of the graph neural network, each node collects historical population change information of its neighboring nodes so that the abstract features of each node contain the topological information of the graph.

6. The generating apparatus as described in claim 5, characterized in that, The graph neural network model includes a fusion unit for fusing abstract features of corresponding regions of two nodes in the following manner: Calculate the vector inner product of the abstract features of the two nodes; or, The abstract features of the two nodes are concatenated into vectors, and the concatenated vectors are mapped to one dimension.

7. The generating apparatus as described in claim 5, characterized in that, The training unit is also used for: The following training process is repeated using the training data until the graph neural network model meets the preset convergence condition. The training data includes multiple training samples, each of which includes historical population distribution data and corresponding population flow data for each region within the target area: The historical population distribution data in the sample is input into the graph neural network model to obtain the population flow data output by the graph neural network model; The error between the population flow data in the sample and the population flow data generated by the graph neural network model is calculated, and the model parameters of the graph neural network model are adjusted using the gradient descent algorithm based on the error.

8. The generating apparatus as described in claim 5, characterized in that, Also includes: The verification unit is used to test the trained graph neural network model using test samples including historical population distribution data and corresponding population flow data of each region in the target area, and to determine whether the trained graph neural network model meets the preset performance index requirements.

Images