Data processing methods, apparatus, computer-readable storage media, and computer equipment

Abstract

This invention discloses a data processing method, apparatus, computer-readable storage medium, and computer device. The method involves acquiring statistical data in at least two dimensions related to GDP information for each grid cell; generating at least two graph structures based on the positional relationships between grid cells and the statistical data; constructing a graph neural network model based on the at least two graph structures; determining the feature vector corresponding to each grid cell based on the graph neural network model; training a preset regression model based on the first feature vectors corresponding to a preset number of first grid cells and first GDP information; and calculating the target GDP corresponding to the target region based on the trained regression model. Thus, by employing machine learning methods, training a network model based on grid cells with partially known GDP, and then using the trained network model to predict the GDP of other grid cells, the efficiency of assessing the level of regional economic development can be improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically to a data processing method, apparatus, computer-readable storage medium, and computer equipment. Background Technology

[0002] Assessing the economic development level of a country or region is an important part of government work. Accurate assessments of economic development help governments make better economic plans and decisions.

[0003] Currently, the assessment of regional economic development levels still relies on economic censuses conducted annually or every five years, which consumes a lot of manpower and time and is very inefficient. Summary of the Invention

[0004] This application provides a data processing method, apparatus, storage medium, and computer equipment. This method can improve the efficiency of data processing, and thus improve the calculation efficiency of GDP information for a target region.

[0005] The first aspect of this application provides a data processing method, the method comprising:

[0006] Obtain statistical data in at least two dimensions related to GDP information for each of multiple grid cells, wherein the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical region according to preset rules;

[0007] Generate at least two graph structures with the plurality of grid cells as nodes based on the positional relationships between grid cells and the statistical data of the at least two dimensions;

[0008] A graph neural network model is constructed based on the at least two graph structures, and a representation learning is performed on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell.

[0009] Obtain the first gross domestic product information corresponding to a preset number of first grid cells in the plurality of grid cells, and train a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first gross domestic product information to obtain the trained regression model;

[0010] Based on the trained regression model, the target GDP information corresponding to the target area in the preset geographical region is determined.

[0011] Accordingly, a second aspect of this application provides a data processing apparatus, the apparatus comprising:

[0012] The acquisition unit is used to acquire statistical data of at least two dimensions related to gross domestic product information for each of the multiple grid cells, wherein the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical region according to preset rules;

[0013] A generation unit is used to generate at least two graph structures with the plurality of grid cells as nodes based on the positional relationship between grid cells and the statistical data of the at least two dimensions;

[0014] A learning unit is used to construct a graph neural network model based on the at least two graph structures, and to perform representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell.

[0015] The training unit is used to acquire the first gross domestic product information corresponding to a preset number of first grid cells in the plurality of grid cells, and to train a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first gross domestic product information to obtain the trained regression model.

[0016] The calculation unit is used to calculate the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model.

[0017] In some embodiments, the generating unit includes:

[0018] The first generating sub-unit is used to acquire positional relationship data between grid units and generate a first graph structure between grid units based on the positional relationship data;

[0019] The second generation subunit is used to extract functional attribute relationship data between grid units from the statistical data of the at least two dimensions, and generate a second graph structure between grid units based on the functional attribute relationship data;

[0020] The third generation subunit is used to extract interaction relationship data between grid units from the statistical data of the at least two dimensions, and generate a third graph structure between grid units based on the interaction relationship data;

[0021] The learning unit is also used for:

[0022] A graph neural network is constructed based on the first graph structure, the second graph structure, and the third graph structure.

[0023] In some embodiments, the first generating subunit includes:

[0024] The determination module is used to determine the adjacency relationship between grid cells based on the positional relationship data;

[0025] The generation module is used to generate an adjacency matrix based on the adjacency relationship;

[0026] A construction module is used to construct a first graph structure between the grid cells, with the plurality of grid cells as nodes and the adjacency matrix as edges.

[0027] In some embodiments, the learning unit includes:

[0028] The learning sub-unit is used to perform unsupervised learning training on multiple sub-graph neural network models contained in the graph neural network based on a preset loss function, so as to obtain multiple trained sub-graph neural network models.

[0029] The input sub-unit is used to input statistical data of at least two dimensions associated with the gross domestic product information corresponding to each grid unit into the trained multiple subgraph neural network model to obtain multiple sub-feature vectors corresponding to each grid unit;

[0030] The first defined sub-unit is used to determine the feature vector corresponding to each grid unit based on the multiple sub-feature vectors corresponding to each grid unit.

[0031] In some embodiments, the determining subunit includes:

[0032] The acquisition module is used to obtain the weight coefficients corresponding to each sub-feature vector;

[0033] The summation module is used to perform a weighted summation of the multiple sub-feature vectors based on the weight coefficients corresponding to each sub-feature vector, so as to obtain the feature vector corresponding to each grid cell.

[0034] In some embodiments, the acquiring unit includes:

[0035] The first acquisition subunit is used to acquire first remote sensing image data, first population flow data, first infrastructure statistics data, and first road network data of a preset geographical area;

[0036] The first sub-unit is used to divide the geographical area based on preset rules to obtain multiple grid units;

[0037] The second sub-unit is used to divide the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data based on the preset rules, so as to obtain the second remote sensing image data, the second population flow data, the second infrastructure statistics data, and the second road network data corresponding to each grid unit.

[0038] The generation unit is further configured to:

[0039] Based on the positional relationships between grid cells and the second remote sensing image data, second population flow data, second infrastructure statistics data, and second road network data corresponding to each grid cell, at least two graph structures with the plurality of grid cells as nodes are generated.

[0040] In some embodiments, the computing unit includes:

[0041] The second acquisition subunit is used to acquire a test data set, the test data set including second gross production value information corresponding to a second preset number of second grid units in the plurality of grid units;

[0042] The test subunit is used to perform accuracy testing on the trained regression model using the test dataset.

[0043] The first calculation subunit is used to calculate the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model when the result index value of the accuracy test is greater than a preset threshold.

[0044] In some embodiments, the computing unit includes:

[0045] The second determining subunit is used to determine a plurality of third grid units corresponding to the target region among the plurality of grid units and to determine a third feature vector corresponding to each third grid unit;

[0046] The second calculation subunit is used to input each third feature vector into the trained regression model to obtain the third gross product information corresponding to each third grid cell, and to calculate the target gross product information of the target region based on the third gross product information.

[0047] A third aspect of this application also provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to perform the steps of the data processing method provided in the first aspect of this application.

[0048] A fourth aspect of this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the data processing method provided in the first aspect of this application.

[0049] A fifth aspect of this application provides a computer program product or computer program, the computer program product or computer program including computer instructions stored in a storage medium. A processor of a computer device reads the computer instructions from the storage medium, and the processor executes the computer instructions, causing the computer device to perform the steps of the data processing method provided in the first aspect.

[0050] The data processing method provided in this application involves obtaining statistical data in at least two dimensions related to GDP information for each of multiple grid cells, where the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical area according to preset rules; generating at least two graph structures with multiple grid cells as nodes based on the positional relationship between the grid cells and the statistical data in at least two dimensions; constructing a graph neural network model based on the at least two graph structures, and performing representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; obtaining first GDP information corresponding to a first preset number of first grid cells in the multiple grid cells, and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; and calculating the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model. In this way, by dividing a geographical region into multiple grid units, and constructing a graph neural network based on the statistical data associated with GDP information for each grid unit and the positional relationships between grid units, a representation learning is performed on each grid unit. Then, the vectors obtained from the representation learning of each grid unit are used to train a regression model with known label data. In this way, the trained regression model can automatically predict the GDP information corresponding to each grid unit in the geographical region, thereby improving the prediction efficiency of regional GDP information and thus improving the efficiency of assessing the level of regional economic development. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of a data processing scenario in this application;

[0053] Figure 2 This is a flowchart illustrating the data processing method provided in this application;

[0054] Figure 3 This is another flowchart illustrating the data processing method provided in this application;

[0055] Figure 4 This is a schematic diagram of the structure of the multi-view graph neural network model of this application;

[0056] Figure 5 This is a schematic diagram of the data processing apparatus provided in this application;

[0057] Figure 6 This is a schematic diagram of the structure of the terminal provided in this application;

[0058] Figure 7 This is a schematic diagram of the server structure provided in this application. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0060] This invention provides a data processing method, apparatus, computer-readable storage medium, and computer device. The data processing method can be used in a data processing apparatus. The data processing apparatus can be integrated into a computer device, which can be a terminal or a server. The terminal can be a mobile phone, tablet computer, laptop computer, smart TV, wearable smart device, personal computer (PC), etc. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0061] Please see Figure 1 This is a schematic diagram of a data processing scenario provided in this application. As shown in the figure, when computer device A acquires statistical data of at least two dimensions related to GDP information for each grid cell in a preset geographical area, where each grid cell is a sub-region obtained by dividing the preset geographical area according to preset rules; at least two graph structures with multiple grid cells as nodes are generated based on the positional relationship between grid cells and the statistical data of at least two dimensions; a graph neural network model is constructed based on the at least two graph structures, and representation learning is performed on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; the first GDP information corresponding to a first preset number of first grid cells in the multiple grid cells is acquired, and a preset regression model is trained based on the first feature vector corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; the target GDP information corresponding to the target area in the preset geographical area is calculated based on the trained regression model.

[0062] It should be noted that, Figure 1 The illustrated data processing scenario diagram is merely an example. The data processing scenarios described in the embodiments of this application are intended to more clearly illustrate the technical solutions of this application and do not constitute a limitation on the technical solutions provided in this application. Those skilled in the art will understand that, with the evolution of data processing and the emergence of new business scenarios, the technical solutions provided in this application are equally applicable to similar technical problems.

[0063] The implementation scenarios described above will be explained in detail below.

[0064] This application will describe embodiments from the perspective of a data processing device, which can be integrated into a computer device. The computer device can be a terminal, such as a mobile phone, tablet computer, laptop computer, smart TV, wearable smart device, or personal computer (PC). Figure 2 The diagram shown is a flowchart of the data processing method provided in this application, which includes:

[0065] Step 101: Obtain statistical data for at least two dimensions associated with the gross domestic product information for each of the multiple grid cells.

[0066] In related technologies, the assessment of a region's economic development level generally uses the region's Gross Domestic Product (GDP) as an evaluation indicator. GDP is the final result of the production activities of all resident units in a country or region within a certain period. It is a core indicator of national economic accounting and an important indicator for measuring a country or region's economic status and development level. In the embodiments of this application, GDP can be used as an indicator to assess the regional economic development level to provide a detailed explanation of the solution provided in this application. Generally, government departments need to regularly collect GDP data for a country or region in order to make better policy decisions and development plans. However, currently, the collection of GDP data still relies on manual economic censuses, which is very inefficient. Therefore, this application provides a data processing method that can improve the efficiency of calculating regional GDP, thereby improving the efficiency of assessing the level of economic development. The data processing method provided in this application will be further described in detail below.

[0067] First, before assessing a region's GDP, a pre-defined geographical area encompassing that region should be established. For example, if assessing the economic development level of Shenzhen, the pre-defined geographical area could be Guangdong Province, or other pre-defined geographical areas encompassing Shenzhen, such as the whole of China or globally. Generally, a pre-defined geographical area slightly larger than the region to be assessed is sufficient; there is no need to set it too large. Specifically, the pre-defined geographical area can be determined based on statistical data associated with GDP information within that area, ensuring convenient data collection.

[0068] After determining the preset geographical area, it can be divided according to preset rules to obtain multiple grid units, which are essentially plots of land. The preset geographical area can be divided using a regular grid, resulting in multiple grid units of the same shape; or it can be divided using road network data or administrative boundary data, resulting in multiple grid units of irregular shapes. The regular grid can use kilometer grids, hexagonal grids, etc.; the road network data can include roads at various levels, such as highways, national roads, provincial roads, and urban arterial roads, resulting in irregular grid units that conform to the geographical spatial pattern and land function distribution.

[0069] After dividing the preset geographical area into multiple grid cells, statistical data associated with GDP for each grid cell are acquired. This GDP-related statistical data includes, but is not limited to, satellite remote sensing imagery data, spatiotemporal trajectory data, Point of Interest (POI) data, and road network data. In this embodiment, statistical data from at least two of these dimensions can be acquired to avoid inaccurate estimations from a single data source.

[0070] In some embodiments, statistical data in at least two dimensions associated with GDP information are obtained for each of a plurality of grid cells, including:

[0071] 1. Acquire the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data for the preset geographical area;

[0072] 2. Divide the geographical area based on preset rules to obtain multiple grid units;

[0073] 3. Based on preset rules, the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data are divided to obtain the second remote sensing image data, the second population flow data, the second infrastructure statistics data, and the second road network data corresponding to each grid unit.

[0074] In some embodiments, when the grid cells are obtained by dividing the data into regular grids, it is inconvenient to statistically analyze the GDP-related statistics within a single regular grid, as some statistical data associated with GDP information are often compiled based on administrative divisions. In this case, statistical data for the entire preset geographical area can be obtained, and then these statistical data can be divided according to preset division rules to obtain the statistical data corresponding to each grid cell. The GDP-related statistical data may include remote sensing image data, population flow data, infrastructure statistics, and road network data. Remote sensing image data includes, but is not limited to, multispectral imagery and nighttime light remote sensing imagery; population flow data includes, but is not limited to, population movement data based on mobile phone positioning and public transportation flow data; infrastructure statistics include, but are not limited to, statistical data on infrastructure in categories such as shopping, food, living services, medical care, and education; and road network data includes, but is not limited to, data on road traffic facility construction density and bus stop density.

[0075] Specifically, we can first acquire the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data corresponding to the preset geographical area. Then, based on the preset division rules, we divide the statistics data corresponding to the preset geographical area to obtain the second remote sensing image data, the second population flow data, the second infrastructure statistics data, and the second road network data corresponding to each grid unit.

[0076] In some embodiments, statistical data of at least two dimensions associated with the gross domestic product information for each grid cell can also be obtained directly.

[0077] Step 102: Generate at least two graph structures with multiple grid cells as nodes based on the positional relationships between grid cells and statistical data in at least two dimensions.

[0078] After obtaining statistical data in at least two dimensions related to GDP information for each grid cell, the connections between different grid cells can be further determined based on these statistical data. For example, the functional attribute similarity between grid cells can be determined based on infrastructure statistical data corresponding to different grid cells. Specifically, if two grid cells both have a large number of factories, then the functional attribute of these two grid cells is likely to be industrial area; if two grid cells both have a large number of supermarkets, then the functional attribute of these two grid cells is likely to be residential area. Therefore, if both are industrial areas or both are residential areas, the functional attribute similarity between these two grid cells is relatively high; if they belong to different functional areas, then their functional attribute similarity is relatively low.

[0079] After determining the functional attribute similarity relationships between each grid cell, a functional attribute similarity diagram can be generated based on these relationships. In addition, a positional relationship diagram between grid cells can be generated based on their positional relationships; a spatial interaction diagram can also be generated based on the flow and interaction of people between grid cells. In this embodiment, at least a positional relationship diagram can be generated based on the positional relationships between grid cells, and another relationship diagram between grid cells can be generated based on statistical data from at least two dimensions, resulting in at least two graph structures with multiple grid cells as nodes. These at least two graph structures constitute a multi-view graph structure.

[0080] In some embodiments, at least two graph structures with multiple grid cells as nodes are generated based on the positional relationships between grid cells and statistical data in at least two dimensions, including:

[0081] 1. Obtain the positional relationship data between grid cells, and generate the first graph structure between grid cells based on the positional relationship data;

[0082] 2. Extract functional attribute relationship data between grid cells from statistical data in at least two dimensions, and generate a second graph structure between grid cells based on the functional attribute relationship data;

[0083] 3. Extract interaction relationship data between grid cells from statistical data in at least two dimensions, and generate a third graph structure between grid cells based on the interaction relationship data.

[0084] In this embodiment, the positional relationships between the multiple grid cells obtained from the division can be acquired. These positional relationships may include adjacency relationships and distance relationships. Then, a first graph structure is generated based on the positional relationships between the grid cells, with grid cells as nodes and positional relationships as edges. This first graph structure is the aforementioned positional relationship graph.

[0085] Then, functional attribute relationship data between grid cells can be extracted from statistical data in at least two dimensions associated with GDP information corresponding to each grid cell. Specifically, the functional attribute relationships between grid cells can be extracted from remote sensing image data, population flow data, infrastructure statistics data, and road network data corresponding to each grid cell. As mentioned earlier, the functional attribute relationship data between grid cells can be determined based on the infrastructure statistics data corresponding to each grid cell. After extracting the functional attribute relationship data between grid cells, a second graph structure can be generated with grid cells as nodes and the functional attribute relationships between grid cells as edges. This second graph structure is the aforementioned functional attribute similarity relationship graph.

[0086] Furthermore, interaction data between grid units can be extracted from statistical data of at least two dimensions associated with GDP information corresponding to each grid unit. Specifically, interaction data between grid units can be determined from population flow data corresponding to each grid unit. Then, a third structural graph between grid units can be generated, with grid units as nodes and interaction data between grid units as edges. This third structural graph is the aforementioned spatial interaction graph.

[0087] In some embodiments, generating a first graph structure between grid cells based on positional relationship data includes:

[0088] 1.1 Determine the adjacency relationships between grid cells based on positional relationship data;

[0089] 1.2 Generate an adjacency matrix based on adjacency relationships;

[0090] 1.3 Construct the first graph structure between grid cells using multiple grid cells as nodes and the adjacency relation matrix as edges.

[0091] After obtaining the positional relationships between grid cells, the adjacency relationships between them can be specifically determined, i.e., whether two grid cells are adjacent. Once the adjacency relationships are determined, an adjacency matrix, also known as an adjacency matrix, can be generated. Furthermore, multiple grid cells can be used as nodes, and the adjacency matrix can represent the set of edges between nodes, thereby generating the first graph structure between grid cells.

[0092] Accordingly, in some embodiments, a similarity matrix can be generated based on the similarity relationship between grid cells, and then the grid cells are used as nodes to represent the set of edges between nodes using the similarity matrix, thereby generating a second graph structure between grid cells.

[0093] Similarly, in some embodiments, an interaction matrix can be generated based on the interaction relationships between grid cells, and then the interaction matrix can be used to represent the set of edges between the nodes, with the grid cells as nodes, to generate a third graph structure between the grid cells.

[0094] Step 103: Construct a graph neural network model based on at least two graph structures, and perform representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell.

[0095] Specifically, after generating at least two graph structures with the aforementioned multiple grid units as nodes based on the positional relationships between grid units and statistical data of at least two dimensions associated with GDP information corresponding to each grid unit, a graph neural network model can be constructed based on these at least two graph structures. Specifically, a multi-view graph neural network model can be constructed. This multi-view graph neural network model can be composed of multiple separate graph neural network models, with each graph structure corresponding to a separate graph neural network model. These separate graph neural network models can also be referred to as sub-graph neural network models.

[0096] In some embodiments, when three graph structures are generated based on the positional relationships between grid cells and the statistical data associated with GDP corresponding to each grid cell, the graph neural network model constructed based on these three graph structures also includes three sub-graph neural network models. Each sub-graph neural network model corresponds to a graph structure; here, the sub-graph neural network model can be a graph convolutional neural network. Graph convolutional neural networks extend convolution operations from traditional data (e.g., image data) to graph data. Its core idea is to learn a function mapping through which nodes in the graph can aggregate their own features and the features of their neighbors to generate new representations of the nodes. Graph convolutional neural networks are the foundation of many complex graph neural network models.

[0097] After constructing a graph neural network model based on at least two graph structures, the graph neural network model is trained unsupervised to obtain a trained graph neural network model. Then, the features corresponding to each grid cell are input into the graph neural network model one by one to obtain the feature vector corresponding to each grid cell.

[0098] In some embodiments, representation learning is performed on each grid cell based on a graph neural network model to obtain a feature vector corresponding to each grid cell, including:

[0099] 1. Based on a preset loss function, perform unsupervised learning training on multiple sub-graph neural network models contained in the graph neural network to obtain multiple trained sub-graph neural network models;

[0100] 2. Input the statistical data of at least two dimensions related to the gross domestic product information corresponding to each grid cell into the trained subgraph neural network model to obtain multiple sub-feature vectors corresponding to each grid cell;

[0101] 3. Determine the feature vector corresponding to each grid cell based on the multiple sub-feature vectors corresponding to each grid cell.

[0102] After constructing the graph neural network model, a loss function can be designed to train the model. In this embodiment, the loss function of the graph neural network model can be determined by weighted summation of various feature reconstruction losses, different grid cell comparison losses, and different modality classification losses according to certain weights.

[0103] After determining the loss function for training the graph neural network model, the model is trained based on this loss function to obtain the trained graph neural network model. The trained graph neural network model contains multiple sub-graph neural network models; here, the sub-graph neural network models refer to the trained sub-graph neural network models. For any target sub-graph neural network model, the sub-feature vector obtained after processing each grid cell can be determined. Therefore, for any grid cell, multiple sub-feature vectors output by each trained sub-graph neural network can be determined. By weighted summing of these multiple sub-feature vectors corresponding to that grid cell, the feature vector corresponding to each grid cell can be obtained.

[0104] In some embodiments, determining the feature vector corresponding to each grid cell based on multiple sub-feature vectors corresponding to each grid cell includes:

[0105] 3.1 Obtain the weight coefficients corresponding to each sub-feature vector;

[0106] 3.2. Based on the weight coefficients corresponding to each sub-feature vector, the multiple sub-feature vectors are weighted and summed to obtain the feature vector corresponding to each grid cell.

[0107] After obtaining multiple sub-feature vectors corresponding to each grid cell, the weight coefficients for each sub-feature vector can be further obtained. Specifically, the weight coefficients for each sub-feature vector can be determined based on the correlation between the data corresponding to each graph structure and the gross product. For example, when the correlation ratio between the location graph, functional attribute similarity graph, and spatial interaction graph and the gross product is 2:3:5, then the output weight of the subgraph neural network model corresponding to the location graph can be determined to be 20%, the output weight of the subgraph neural network model corresponding to the functional attribute similarity graph to be 30%, and the output weight of the subgraph neural network model corresponding to the spatial interaction graph to be 50%. This further determines the weight coefficients for the sub-feature vectors output by these three subgraph neural network models. After obtaining the weight coefficients for each sub-feature vector, a weighted sum of multiple sub-feature vectors can be performed based on the weight coefficients for each sub-feature vector to obtain the feature vector corresponding to each grid cell.

[0108] Step 104: Obtain the first gross domestic product information corresponding to a first preset number of first grid cells in multiple grid cells, and train a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first gross domestic product information to obtain the trained regression model.

[0109] After determining the feature vector corresponding to each grid cell, a first preset number of first grid cells are selected from multiple grid cells, and the first gross domestic product (GDP) information corresponding to these first preset number of grid cells is obtained. Specifically, for example, after dividing a preset geographical area into 10,000 grid cells, 100 representative grid cells can be selected as first grid cells, and the first GDP information of these 100 first grid cells is obtained. The GDP information of these first preset number of first grid cells can be obtained based on statistical yearbook data corresponding to these grid cells, or it can be obtained through on-site manual statistics of these areas. That is, the first GDP information corresponding to these first grid cells is the actual GDP information of these first grid cells. Further, since the feature vector corresponding to each grid cell has been determined in step 103, the first feature vector corresponding to the aforementioned first preset number of first grid cells can be determined accordingly. Further, the first feature vector and the first GDP information can be used as training samples to train a preset regression model, resulting in the trained regression model. The first feature vector is the input of the preset regression model, and the first gross domestic product information is the label corresponding to the output of the preset regression model.

[0110] Step 105: Calculate the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model.

[0111] After training a preset regression model to obtain the trained regression model, this trained regression model can be deployed to the application scenario for GDP estimation. In this application, since the trained regression model can calculate the GDP information of any grid cell in a preset geographical area, it can also calculate the target GDP information of a target area within the preset geographical area. Specifically, for example, if the preset geographical area is Guangdong Province, the aforementioned preset regression model can be used to calculate the GDP information of any target area within Guangdong Province.

[0112] In some embodiments, the target GDP information corresponding to a target region within a preset geographical area is calculated based on a trained regression model, including:

[0113] 1. Determine multiple third grid cells corresponding to the target region within multiple grid cells, and determine the third feature vector corresponding to each third grid cell;

[0114] 2. Input each third feature vector into the trained regression model to obtain the third GDP information corresponding to each third grid cell, and calculate the target GDP information of the target region based on the third GDP information.

[0115] When calculating the GDP information of a target region using a trained regression model, the specific grid units contained within the target region can be determined first. Then, the GDP information of each grid unit within the target region is obtained individually, and finally, the GDP information of the target region is calculated based on the GDP information corresponding to the grid units. Specifically, for example, if the geographical region is preset to Guangdong Province and divided into 10,000 grid units, then if the GDP information of Zhongshan City needs to be calculated, multiple target grid units contained in the corresponding area of ​​Zhongshan City can be determined from the 10,000 grid units corresponding to Guangdong Province; these can be referred to as third grid units. Assuming that Zhongshan City contains 500 third grid units, then the third feature vector corresponding to these 500 third grid units can be obtained from the feature vector corresponding to each grid unit determined in step 103. These 500 third feature vectors are input one by one into the trained regression model to obtain 500 third GDP information output by the trained regression model. Finally, the GDP information of Zhongshan City is obtained by summing these 500 third GDP information.

[0116] In some embodiments, the target GDP information corresponding to a target region within a preset geographical area is calculated based on a trained regression model, including:

[0117] A. Obtain a test data set, which includes the second gross production value information corresponding to a second preset number of second grid cells in multiple grid cells;

[0118] B. Use the test dataset to test the accuracy of the trained regression model;

[0119] C. When the result of the accuracy test is greater than the preset threshold, the target GDP information corresponding to the target area in the preset geographical area is calculated based on the trained regression model.

[0120] In this embodiment, after training a preset regression model to obtain the trained regression model, a test dataset can be used to test the accuracy of the regression model. The test dataset can consist of multiple grid cells that differ from the grid cells in the training dataset and their corresponding GDP information. For example, when training the preset regression model, 100 representative grid cells and their GDP information from 10,000 grid cells are used as training samples. Then, when testing the accuracy of the trained regression model, other grid cells besides these 100 grid cells can be used, for example, 50. For these 50 grid cells, their corresponding feature vectors and GDP information are obtained. The GDP information for these 50 grid cells can be obtained from statistical yearbook data corresponding to these grid cells, or it can be obtained through on-site manual statistics of the regions corresponding to these grid cells.

[0121] As described above, the data processing method provided in this application involves obtaining statistical data in at least two dimensions related to GDP information for each of multiple grid cells, where the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical region according to preset rules; generating at least two graph structures with multiple grid cells as nodes based on the positional relationship between grid cells and the statistical data in at least two dimensions; constructing a graph neural network model based on the at least two graph structures, and performing representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; obtaining first GDP information corresponding to a first preset number of first grid cells in the multiple grid cells, and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; and calculating the target GDP information corresponding to the target region in the preset geographical region based on the trained regression model. In this way, by dividing a geographical region into multiple grid units, and constructing a graph neural network based on the statistical data associated with GDP information for each grid unit and the positional relationships between grid units, a representation learning is performed on each grid unit. Then, the vectors obtained from the representation learning of each grid unit are used to train a regression model with known label data. In this way, the trained regression model can automatically predict the GDP information corresponding to each grid unit in the geographical region, thereby improving the prediction efficiency of regional GDP information and thus improving the efficiency of assessing the level of regional economic development.

[0122] Accordingly, this application will further describe the data processing method provided in this application from the perspective of a computer device, wherein the computer device can be a terminal or a server. Figure 3The diagram shown is another flowchart illustrating the data processing method provided in this application, which includes:

[0123] Step 201: The computer device divides the preset geographical area into multiple plot units.

[0124] The preset geographical region can be determined based on the target region for economic development level assessment. The preset geographical region can be any geographical region that includes the target region. Specifically, when assessing the economic development level of various cities and districts in Guangdong Province, the preset geographical region can be Guangdong Province, or it can be China or the entire world. It is understandable that a larger preset geographical region allows for a wider scope of economic development level assessment, thus expanding the applicability of the method provided in this application; however, a larger preset geographical region also leads to a greater computational burden of acquiring and processing data, resulting in decreased assessment efficiency.

[0125] To assess the level of regional economic development, several evaluation indicators for regional economic development can be used. These indicators include, but are not limited to, GDP and the Engel coefficient. In this embodiment, GDP can be used as an example to illustrate the data processing method provided in this application.

[0126] After determining the preset geographical area, it can be divided according to preset rules to obtain multiple land parcel units. These division rules can be based on grid lines, road network data, or administrative boundary data. It is understood that the above method for dividing the preset geographical area into land parcels is merely an example; other preset rules can also be used to divide the preset geographical area, which are not limited here. The resulting multiple land parcel units can be represented by a set of land parcel units. .in, represents the i-th plot unit, and N represents the total number of plot units.

[0127] Step 202: The computer equipment acquires remote sensing image data, spatiotemporal trajectory data, point of interest data, and road network data for each plot unit.

[0128] After dividing a predefined geographical area into multiple plot units, resulting in a set of plot units, remote sensing imagery, spatiotemporal trajectory data, points of interest (POIs) data, and road network data for each plot unit can be further acquired. The remote sensing imagery can be satellite imagery, specifically including multispectral imagery and nighttime light remote sensing imagery. After acquiring the remote sensing imagery, spectral, texture, and shape features can be extracted. Spatiotemporal trajectory data can include population movement data based on mobile phone positioning and public transportation flow data, which can be statistically analyzed at the plot unit scale to obtain the intensity of population activity within each plot unit and the intensity of personnel interaction between plot units. POIs data can include infrastructure data within the plot unit, including but not limited to shopping malls, food courts, community service centers, medical institutions, financial institutions, and educational institutions. The road network includes the construction density and accessibility of road facilities within the plot unit. For example, facility construction density can be characterized by indicators such as road construction density and public transportation stop density, while accessibility can be measured by the accessibility of transportation stops within the area.

[0129] The aforementioned remote sensing impact data, spatiotemporal trajectory data, point-of-interest data, and road network data are all statistical data from different dimensions associated with the GDP information of the land parcels. It is understood that the statistical data associated with the GDP information of the land parcels may include other dimensions besides these; this is merely an example and not intended to limit the scope of this application. In this application, statistical data from one dimension or multiple dimensions can be used. Using statistical data from multiple dimensions avoids inaccurate estimations from a single data source, thereby improving the accuracy of the GDP prediction for each land parcel.

[0130] Step 203: The computer equipment constructs a multi-view graph structure based on the remote sensing image data, spatiotemporal trajectory data, point of interest data, and road network data of each plot unit.

[0131] After acquiring remote sensing imagery, spatiotemporal trajectory data, points of interest (POIs), and road network data for each land parcel unit, the relationships between these land parcel units can be extracted from these statistical data. These relationships can be evaluated from multiple dimensions, resulting in multi-dimensional relationships between land parcel units. Then, based on these relationships, a graph structure with land parcel units as nodes is constructed, resulting in multiple graph structures, i.e., a multi-view structure.

[0132] First, the adjacency relationships between land parcels can be determined based on the division of land parcel units, and a spatial adjacency graph can be constructed based on these relationships. The spatial adjacency graph can be represented as follows:

[0133]

[0134] in, V represents the spatial adjacency graph; V represents the set of nodes in the spatial adjacency graph, corresponding to the set of land parcel units R; The set of edges in a spatial adjacency graph is represented by an adjacency matrix. The adjacency matrix is ​​represented as follows. This can be expressed as follows:

[0135] .

[0136] in, Representing the adjacency matrix The value in the i-th row and j-th column, This represents the i-th element in set R. Let represent the j-th element in set R.

[0137] Then, based on the statistical data of the points of interest for each plot unit, the functional similarity between any two plot units can be calculated, thus obtaining the functional similarity relationship between the plot units. For example, if a plot is designated for industrial, residential, or commercial use, the more similar the functions, the higher the functional similarity between the two plots. Furthermore, a functional similarity graph can be constructed based on the functional similarity relationship between the plot units.

[0138]

[0139] in, V represents the similarity graph; V represents the set of nodes in the similarity graph, corresponding to the set of land parcel units R; The set of edges in a similarity graph is represented by a similarity matrix. It means that, among them:

[0140]

[0141] in, Representing similarity matrices The value in the i-th row and j-th column, For the i-th plot unit in the set of plot units, Let be the j-th plot unit in the set of plot units. SIM(a, b) represents calculating the similarity between a and b. Specifically, it can be the cosine similarity of the word vectors corresponding to the main attributes of the two plot units.

[0142] Furthermore, the interaction relationships between land parcel units can be determined based on spatiotemporal trajectory data, and a spatial interaction diagram can be constructed based on these relationships.

[0143]

[0144] in, V represents the spatial interaction diagram; V represents the set of nodes in the spatial interaction diagram, corresponding to the set of land parcel units R; The set of edges representing a spatial interaction graph is represented by an interaction matrix. It means that, among them:

[0145]

[0146] in, Representing the interaction matrix The value in the i-th row and j-th column, Representing land parcel units and The strength of interaction between them The expansion factor is used to highlight the changes in spatial effects.

[0147] Among them, the spatial adjacency graph, the similarity graph, and the spatial interaction graph together constitute the multi-view graph structure G, where:

[0148]

[0149] Step 204: The computer device constructs a multi-view graph neural network model based on the multi-view graph structure and trains the multi-view graph neural network model to obtain the trained multi-view graph neural network model.

[0150] After generating a multi-view graph structure G based on the acquired statistical data, a multi-view graph neural network (MV-GNN) g can be further constructed based on the multi-view graph structure G. The multi-view graph neural network g also includes multiple single-view graph neural networks g1, g2, and g3. For each single-view graph neural network, its core network can be a graph convolutional network (GCN) or a variant of GCN, such as a graph transformer network (GTN).

[0151] For any given land parcel, its statistical data can be merged to determine the parcel's attribute feature X. Then, X is input into each individual view graph neural network to obtain the hidden vector output by each individual view graph neural network.

[0152]

[0153] in, This is the feature vector output by the k-th single-view graph neural network. This is the relation matrix between the k-th single-view graph neural network and the corresponding view. Specifically, After obtaining the latent vectors of the plot unit and the corresponding outputs of the neural network for each individual view, the global latent vector corresponding to that plot unit can be obtained through the cross-view fusion module.

[0154]

[0155] Where h is the global latent vector corresponding to the land parcel unit. Let n be the weight corresponding to the k-th view, and n be the total number of views, where n=3.

[0156] Then, we further design the loss function for training the multi-view image neural network model. Specifically, we can design a feature reconstruction loss. Comparison of losses between different land parcel units and classification loss for different modalities And so on, and obtain the final loss by weighted summation:

[0157]

[0158] in, and for and The corresponding weights.

[0159] The multi-view graph neural network model is trained using the loss function described above, so that the relationship between the hidden vectors output by the plot units gradually approaches the relationship between the features X corresponding to the plot units, thus completing the training of the multi-view graph neural network model.

[0160] like Figure 4 The diagram shows the structure of the multi-view graph neural network model of this application. As shown, the input data of the multi-view graph neural network is the attribute feature X of each plot unit. The attribute feature X is generated by combining statistical data from each plot unit, including satellite imagery data, POI data, spatiotemporal trajectory data, and road network data. The learning objective of the multi-view graph neural network model is defined by a loss function composed of feature reconstruction loss, plot comparison loss, and modality classification loss. The specific structure of the multi-view graph neural network consists of three single-view GNNs corresponding to three graph structures. Each single-view GNN performs convolution processing on the attribute features of the input grid unit, resulting in a corresponding latent vector. These latent vectors are then fused by the cross-view fusion module to obtain the feature vector output by the multi-view graph neural network. Based on the aforementioned loss function, the multi-view graph neural network model is trained to determine the optimal model parameters corresponding to the minimum loss function, thus completing the training of the multi-view graph neural network.

[0161] Step 205: The computer device determines the feature vector corresponding to each plot unit based on the trained multi-view graph neural network model.

[0162] Among them, after completing the training of the multi-view graph neural network model g, the representation learning of each plot unit is completed, and the sub-feature vectors corresponding to each plot unit and each single-view graph neural network model are obtained.

[0163] Then, further calculate the feature vector output by the multi-view graph neural network model for each plot unit based on the sub-feature vectors corresponding to each plot unit and each single-view graph neural network model and the weight coefficients corresponding to each single-view graph neural network model, so as to obtain the feature vector corresponding to each plot unit.

[0164] Step 206: The computer device obtains the GOP data corresponding to a preset number of plot units in the plot unit set, and trains a preset regression model according to the feature vectors corresponding to the preset number of plot units and the GDP data to obtain the trained regression model.

[0165] E Among them, after determining the feature vector corresponding to each plot unit, a preset number of first plot units can be determined from the plot unit set. These plot units should be relatively evenly distributed in the preset geographical area, and then the first feature vectors corresponding to these first plot units can be determined from the previously determined feature vectors corresponding to each plot unit, so as to obtain a preset number of first feature vectors.

[0166] Then, obtain the GDP data of these preset number of plot units. Among them, the GDP data of these preset number of plot units can be obtained from social and economic statistical data or statistical yearbooks, and these GDP data form a data set , where is the GDP data corresponding to the i-th first plot unit, n is the preset number, where n < N. Then the GDP data corresponding to each first feature vector can be obtained, and each first feature vector and its corresponding GDP data form a set of training data, and a total of n sets of training sample data of training data can be obtained. Then use the training sample data containing n sets of training data to train the preset regression model. Specifically, use each first feature vector as the input of the model, and use the GDP data corresponding to the first feature vector as the label of the model output to train the model to obtain the trained regression model.

[0167] Step 207: The computer device determines the target plot units included in the target area.

[0168] When assessing the economic development level of a target region, it is necessary to first determine the target land parcels contained within that region. As mentioned earlier, since the preset geographical region includes the target region, the target region must be composed of a portion of the land parcels in the set of land parcels. Therefore, after determining the target region for which the economic development level assessment is to be conducted, this target region can be identified within the preset geographical region, and then the target number of target land parcels within that region can be further determined. For example, if Guangdong Province is divided into 10,000 land parcels, then when assessing the economic development level of Shenzhen, the 800 target land parcels corresponding to Shenzhen can be identified from these 10,000 land parcels.

[0169] Step 208: The computer equipment determines the target feature vector corresponding to each target plot unit.

[0170] After determining the target number of target land parcels corresponding to the target area in the set of land parcel units, the target feature vector corresponding to each target land parcel unit can be further determined. Since the feature vector corresponding to each land parcel unit has already been determined in step 205, it is only necessary to determine the target feature vector corresponding to each target land parcel unit from the mapping relationship between land parcel units and feature vectors, thus obtaining the target number of target feature vectors.

[0171] Step 209: The computer device inputs the target feature vector into the trained regression model to obtain the output target GDP value, and calculates the target GDP of the target area based on the target GDP value.

[0172] After determining the target number of target feature vectors corresponding to the target region, these feature vectors can be input into the trained regression model to obtain the GDP value corresponding to each target feature vector, resulting in a total of the target number of GDP values. Furthermore, these target number of GDP values ​​can be summed to obtain the target GDP corresponding to the target region. Then, the economic development level of the target region can be evaluated based on its target GDP.

[0173] As described above, the data processing method provided in this application involves obtaining statistical data in at least two dimensions related to GDP information for each of multiple grid cells, where the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical region according to preset rules; generating at least two graph structures with multiple grid cells as nodes based on the positional relationship between grid cells and the statistical data in at least two dimensions; constructing a graph neural network model based on the at least two graph structures, and performing representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; obtaining first GDP information corresponding to a first preset number of first grid cells in the multiple grid cells, and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; and calculating the target GDP information corresponding to the target region in the preset geographical region based on the trained regression model. In this way, by dividing a geographical region into multiple grid units, and constructing a graph neural network based on the statistical data associated with GDP information for each grid unit and the positional relationships between grid units, a representation learning is performed on each grid unit. Then, the vectors obtained from the representation learning of each grid unit are used to train a regression model with known label data. In this way, the trained regression model can automatically predict the GDP information corresponding to each grid unit in the geographical region, thereby improving the prediction efficiency of regional GDP information and thus improving the efficiency of assessing the level of regional economic development.

[0174] To better implement the above methods, embodiments of the present invention also provide a data processing device that can be integrated into a terminal.

[0175] For example, such as Figure 5 The diagram shown is a schematic representation of the data processing apparatus provided in an embodiment of this application. The data processing apparatus may include an acquisition unit 301, a generation unit 302, a learning unit 303, a training unit 304, and a calculation unit 305, as follows:

[0176] The acquisition unit 301 is used to acquire statistical data of at least two dimensions related to the gross domestic product information for each of the multiple grid cells. The multiple grid cells are multiple sub-regions obtained by dividing a preset geographical area according to preset rules.

[0177] The generation unit 302 is used to generate at least two graph structures with multiple grid cells as nodes based on the positional relationship between grid cells and statistical data in at least two dimensions.

[0178] Learning unit 303 is used to construct a graph neural network model based on at least two graph structures, and to perform representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell.

[0179] Training unit 304 is used to obtain the first gross production value information corresponding to a preset number of first grid cells in multiple grid cells, and to train a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first gross production value information to obtain the trained regression model.

[0180] The calculation unit 305 is used to calculate the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model.

[0181] In some embodiments, the generating unit includes:

[0182] The first generating sub-unit is used to obtain the positional relationship data between grid units and generate the first graph structure between grid units based on the positional relationship data;

[0183] The second generating sub-unit is used to extract functional attribute relationship data between grid units from statistical data in at least two dimensions, and generate a second graph structure between grid units based on the functional attribute relationship data;

[0184] The third generation sub-unit is used to extract interaction relationship data between grid units from statistical data in at least two dimensions, and generate a third graph structure between grid units based on the interaction relationship data;

[0185] The learning unit is also used for:

[0186] Construct a graph neural network based on the first graph structure, the second graph structure, and the third graph structure.

[0187] In some embodiments, the first generating subunit includes:

[0188] The determination module is used to determine the adjacency relationship between grid cells based on the positional relationship data;

[0189] The generation module is used to generate an adjacency matrix based on adjacency relationships.

[0190] The building module is used to construct the first graph structure between grid cells, with multiple grid cells as nodes and an adjacency matrix as edges.

[0191] In some embodiments, the learning unit includes:

[0192] The learning sub-unit is used to perform unsupervised learning training on multiple sub-graph neural network models contained in the graph neural network based on a preset loss function, so as to obtain multiple trained sub-graph neural network models.

[0193] The input sub-unit is used to input statistical data of at least two dimensions associated with the gross product information corresponding to each grid unit into the trained subgraph neural network model to obtain multiple sub-feature vectors corresponding to each grid unit.

[0194] The first defined sub-unit is used to determine the feature vector corresponding to each grid unit based on the multiple sub-feature vectors corresponding to each grid unit.

[0195] In some embodiments, determining the subunit includes:

[0196] The acquisition module is used to obtain the weight coefficients corresponding to each sub-feature vector;

[0197] The summation module is used to perform a weighted summation of multiple sub-feature vectors based on the weight coefficients corresponding to each sub-feature vector, so as to obtain the feature vector corresponding to each grid cell.

[0198] In some embodiments, the acquiring unit includes:

[0199] The first acquisition subunit is used to acquire first remote sensing image data, first population flow data, first infrastructure statistics data, and first road network data of a preset geographical area;

[0200] The first sub-unit is used to divide a geographic area based on preset rules, resulting in multiple grid units;

[0201] The second sub-unit is used to divide the first remote sensing image data, the first population flow data, the first infrastructure statistics data and the first road network data according to preset rules, so as to obtain the second remote sensing image data, the second population flow data, the second infrastructure statistics data and the second road network data corresponding to each grid unit.

[0202] The generating unit is also used for:

[0203] Based on the positional relationships between grid cells and the second remote sensing image data, second population flow data, second infrastructure statistics data, and second road network data corresponding to each grid cell, at least two graph structures with multiple grid cells as nodes are generated.

[0204] In some embodiments, the computing unit includes:

[0205] The second acquisition subunit is used to acquire a test data set, which includes second gross production value information corresponding to a second preset number of second grid units in multiple grid units;

[0206] The test subunit is used to test the accuracy of the trained regression model using a test dataset.

[0207] The first calculation subunit is used to calculate the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model when the result index value of the accuracy test is greater than the preset threshold.

[0208] In some embodiments, the computing unit includes:

[0209] The second determining sub-unit is used to determine multiple third grid units corresponding to the target region among multiple grid units and to determine the third feature vector corresponding to each third grid unit;

[0210] The second calculation subunit is used to input each third feature vector into the trained regression model to obtain the third GDP information corresponding to each third grid cell, and to calculate the target GDP information of the target area based on the third GDP information.

[0211] In practice, each of the above units 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, please refer to the previous method embodiments, which will not be repeated here.

[0212] As described above, the data processing method provided in this application embodiment involves: an acquisition unit 301 acquiring statistical data of at least two dimensions associated with GDP information for each of multiple grid units, wherein the multiple grid units are multiple sub-regions obtained by dividing a preset geographical region according to preset rules; a generation unit 302 generating at least two graph structures with multiple grid units as nodes based on the positional relationship between grid units and the statistical data of at least two dimensions; a learning unit 303 constructing a graph neural network model based on the at least two graph structures and performing representation learning on each grid unit based on the graph neural network model to obtain the feature vector corresponding to each grid unit; a training unit 304 acquiring first GDP information corresponding to a first preset number of first grid units in the multiple grid units and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid units and the first GDP information to obtain the trained regression model; and a calculation unit 305 calculating the target GDP information corresponding to the target region in the preset geographical region based on the trained regression model. In this way, by dividing a geographical region into multiple grid units, and constructing a graph neural network based on the statistical data associated with GDP information for each grid unit and the positional relationships between grid units, a representation learning is performed on each grid unit. Then, the vectors obtained from the representation learning of each grid unit are used to train a regression model with known label data. In this way, the trained regression model can automatically predict the GDP information corresponding to each grid unit in the geographical region, thereby improving the prediction efficiency of regional GDP information and thus improving the efficiency of assessing the level of regional economic development.

[0213] This application also provides a computer device, which can be a terminal, such as... Figure 6 As shown, the terminal may include a radio frequency (RF) circuit 401, a memory 402 including one or more computer-readable storage media, an input unit 403, a display unit 404, a sensor 405, an audio circuit 406, a wireless Fidelity (WiFi) module 407, a processor 408 including one or more processing cores, and a power supply 409, among other components. Those skilled in the art will understand that... Figure 6 The terminal structure shown does not constitute a limitation on the terminal and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0214] RF circuit 401 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink information from the base station and hands it over to one or more processors 408 for processing; additionally, it transmits uplink data to the base station. Typically, RF circuit 401 includes, but is not limited to, an antenna, at least one amplifier, a tuner, one or more oscillators, a Subscriber Identity Module (SIM) card, a transceiver, a coupler, a low-noise amplifier (LNA), a duplexer, etc. Furthermore, RF circuit 401 can also communicate wirelessly with networks and other devices. Wireless communication can use any communication standard or protocol, including but not limited to GSM, GPRS, CDMA, WCDMA, LTE, email, and SMS.

[0215] The memory 402 can be used to store software programs and modules. The processor 408 executes various functional applications and information interactions 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 terminal (such as audio data, phone book, 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 access to the memory 402 for the processor 408 and the input unit 403.

[0216] Input unit 403 can be used to receive input digital or character information, and to generate keyboard, mouse, joystick, optical, or trackball signal inputs related to user settings and function control. Specifically, in one embodiment, input unit 403 may include a touch-sensitive surface and other input devices. A touch-sensitive surface, also known as a touch display or touchpad, can collect user touch operations on or near it (e.g., user operations using fingers, styluses, or any suitable object or accessory on or near the touch-sensitive surface) and drive corresponding connection devices according to a pre-set program. Optionally, the touch-sensitive surface may include a touch detection device and a touch controller. The touch detection device detects the user's touch orientation and the signal generated by the touch operation, transmitting the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, sends it to the processor 408, and can receive and execute commands from the processor 408. Furthermore, various types of touch-sensitive surfaces, such as resistive, capacitive, infrared, and surface acoustic wave, can be used. In addition to the touch-sensitive surface, input unit 403 may also include other input devices. Specifically, other input devices may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0217] Display unit 404 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the terminal. These graphical user interfaces can be composed of graphics, text, icons, video, and any combination thereof. Display unit 404 may include a display panel, optionally configured as a liquid crystal display (LCD), organic light-emitting diode (OLED), or similar form. Furthermore, a touch-sensitive surface may cover the display panel. When the touch-sensitive surface detects a touch operation on or near it, it transmits the information to processor 408 to determine the type of touch event. Subsequently, processor 408 provides corresponding visual output on the display panel according to the type of touch event. Although in Figure 6 In this context, the touch-sensitive surface and the display panel are two separate components for implementing input and output functions. However, in some embodiments, the touch-sensitive surface and the display panel can be integrated to achieve both input and output functions.

[0218] The terminal may also include at least one sensor 405, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display panel according to the ambient light level, and the proximity sensor can turn off the display panel and / or backlight when the terminal is moved to the ear. As a type of motion sensor, a gravity acceleration sensor can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity and can be used for applications that recognize the phone's posture (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition-related functions (such as pedometer, tapping), etc. Other sensors that the terminal may also be equipped with, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be described in detail here.

[0219] Audio circuitry 406, a speaker, and a microphone provide an audio interface between the user and the terminal. Audio circuitry 406 converts received audio data into electrical signals, transmits them to the speaker, and the speaker converts them into sound signals for output. Conversely, the microphone converts collected sound signals into electrical signals, which are then received by audio circuitry 406, converted back into audio data, and processed by processor 408. The processed data is then transmitted via RF circuitry 401 to, for example, another terminal, or output to memory 402 for further processing. Audio circuitry 406 may also include an earphone jack to facilitate communication between a peripheral headset and the terminal.

[0220] WiFi is a short-range wireless transmission technology. Terminals using the WiFi module 407 can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 6 WiFi module 407 is shown, but it is understood that it is not an essential component of the terminal and can be omitted as needed without changing the essence of the invention.

[0221] The processor 408 is the control center of the terminal, connecting various parts of the phone via various interfaces and lines. It executes software programs and / or modules stored in the memory 402, and calls data stored in the memory 402 to perform various functions and process data, thereby performing overall detection of the phone. Optionally, the processor 408 may include one or more processing cores; preferably, the processor 408 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 modem processor may not be integrated into the processor 408.

[0222] The terminal also includes a power supply 409 (such as a battery) to power various components. Preferably, the power supply can be logically connected to the processor 408 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 409 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.

[0223] Although not shown, the terminal may also include a camera, Bluetooth module, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 408 in the terminal will load the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 408 will run the applications stored in the memory 402 to realize various functions:

[0224] The process involves: acquiring statistical data in at least two dimensions related to GDP information for each grid cell in a set of grid cells, where the grid cells are sub-regions obtained by dividing a set geographical region according to preset rules; generating at least two graph structures with the grid cells as nodes based on the positional relationships between the grid cells and the statistical data in at least two dimensions; constructing a graph neural network model based on the at least two graph structures, and performing representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; acquiring first GDP information corresponding to a first preset number of first grid cells in the set of grid cells, and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; and calculating the target GDP information corresponding to the target region in the set geographical region based on the trained regression model.

[0225] It should be noted that the computer device provided in this application embodiment and the method in the above embodiment belong to the same concept. The specific implementation of each of the above operations can be found in the previous embodiments, and will not be repeated here.

[0226] This application also provides a computer device, which can be a server, such as... Figure 7 The diagram shown is a structural schematic of the computer device provided in this application. Specifically:

[0227] The computer device may include components such as a processing unit 501 with one or more processing cores, a storage unit 502 with one or more storage media, a power module 503, and an input module 504. Those skilled in the art will understand that... Figure 7The computer device structure shown 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:

[0228] The processing unit 501 is the control center of the computer device. It connects various parts of the computer device via various interfaces and lines. By running or executing software programs and / or modules stored in the storage unit 502, and by calling data stored in the storage unit 502, it performs various functions of the computer device and processes data, thereby performing overall detection of the computer device. Optionally, the processing unit 501 may include one or more processing cores; preferably, the processing unit 501 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 modem processor may not be integrated into the processing unit 501.

[0229] Storage unit 502 can be used to store software programs and modules. Processing unit 501 executes various functional applications and data processing by running the software programs and modules stored in storage unit 502. Storage unit 502 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback, image playback, and web page access), etc.; the data storage area may store data created based on the use of the computer device. In addition, storage unit 502 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, storage unit 502 may also include a memory controller to provide processing unit 501 with access to storage unit 502.

[0230] The computer equipment also includes a power supply module 503 that supplies power to various components. Preferably, the power supply module 503 can be logically connected to the processing unit 501 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply module 503 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.

[0231] The computer device may also include an input module 504, which can be used to receive input numeric or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0232] 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 processing unit 501 in the computer device loads the executable files corresponding to the processes of one or more applications into the storage unit 502 according to the following instructions, and the processing unit 501 runs the applications stored in the storage unit 502 to realize various functions, as follows:

[0233] The process involves: acquiring statistical data in at least two dimensions related to GDP information for each grid cell in a set of grid cells, where the grid cells are sub-regions obtained by dividing a set geographical region according to preset rules; generating at least two graph structures with the grid cells as nodes based on the positional relationships between the grid cells and the statistical data in at least two dimensions; constructing a graph neural network model based on the at least two graph structures, and performing representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; acquiring first GDP information corresponding to a first preset number of first grid cells in the set of grid cells, and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; and calculating the target GDP information corresponding to the target region in the set geographical region based on the trained regression model.

[0234] It should be noted that the computer device provided in this application embodiment and the method in the above embodiment belong to the same concept. The specific implementation of each of the above operations can be found in the previous embodiments, and will not be repeated here.

[0235] 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.

[0236] Therefore, embodiments of the present invention provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the methods provided in the embodiments of the present invention. For example, the instructions can execute the following steps:

[0237] The process involves: acquiring statistical data in at least two dimensions related to GDP information for each grid cell in a set of grid cells, where the grid cells are sub-regions obtained by dividing a set geographical region according to preset rules; generating at least two graph structures with the grid cells as nodes based on the positional relationships between the grid cells and the statistical data in at least two dimensions; constructing a graph neural network model based on the at least two graph structures, and performing representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell; acquiring first GDP information corresponding to a first preset number of first grid cells in the set of grid cells, and training a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first GDP information to obtain the trained regression model; and calculating the target GDP information corresponding to the target region in the set geographical region based on the trained regression model.

[0238] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0239] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0240] Since the instructions stored in the computer-readable storage medium can execute the steps of any of the methods provided in the embodiments of the present invention, the beneficial effects that any of the methods provided in the embodiments of the present invention can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0241] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a storage medium. A processor of a computer device reads the computer instructions from the storage medium and executes the computer instructions, causing the computer device to perform the aforementioned actions. Figure 2 or Figure 3 The methods provided in various optional implementations.

[0242] The data processing method, apparatus, computer-readable storage medium, and computer device provided by the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. 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 the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A data processing method, characterized in that, The method includes: Obtain statistical data in at least two dimensions related to GDP information for each of multiple grid cells, wherein the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical region according to preset rules; Generating at least two graph structures with the plurality of grid cells as nodes based on the positional relationships between grid cells and the statistical data of at least two dimensions includes: acquiring positional relationship data between grid cells and generating a first graph structure between grid cells based on the positional relationship data; extracting functional attribute relationship data between grid cells from the statistical data of at least two dimensions and generating a second graph structure between grid cells based on the functional attribute relationship data; and extracting interaction relationship data between grid cells from the statistical data of at least two dimensions and generating a third graph structure between grid cells based on the interaction relationship data. Constructing a graph neural network model based on the at least two graph structures includes: constructing a graph neural network based on the first graph structure, the second graph structure, and the third graph structure; Based on the graph neural network model, representation learning is performed on each grid cell to obtain the feature vector corresponding to each grid cell; Obtain the first gross domestic product information corresponding to a first preset number of first grid cells in the plurality of grid cells, and train a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first gross domestic product information to obtain the trained regression model; The target GDP information corresponding to the target area in the preset geographical region is calculated based on the trained regression model.

2. The method according to claim 1, characterized in that, The step of generating a first graph structure between grid cells based on the positional relationship data includes: The adjacency relationship between grid cells is determined based on the positional relationship data; Generate an adjacency matrix based on the adjacency relationships; A first graph structure is constructed between the grid cells using the plurality of grid cells as nodes and the adjacency matrix as edges.

3. The method according to claim 1, characterized in that, The process of learning the representation of each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell includes: Unsupervised learning training is performed on multiple sub-graph neural network models contained in the graph neural network based on a preset loss function to obtain multiple trained sub-graph neural network models. The statistical data of at least two dimensions associated with the gross domestic product information corresponding to each grid cell are input into the trained subgraph neural network model to obtain multiple sub-feature vectors corresponding to each grid cell. The feature vector corresponding to each grid cell is determined based on the multiple sub-feature vectors corresponding to each grid cell.

4. The method according to claim 3, characterized in that, The step of determining the feature vector corresponding to each grid cell based on multiple sub-feature vectors corresponding to each grid cell includes: Obtain the weight coefficients corresponding to each sub-feature vector; The multiple sub-feature vectors are weighted and summed based on the weight coefficients corresponding to each sub-feature vector to obtain the feature vector corresponding to each grid cell.

5. The method according to claim 1, characterized in that, The acquisition of statistical data in at least two dimensions related to GDP information for each of the multiple grid cells includes: Acquire the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data for the preset geographical area; The geographical region is divided into multiple grid units based on preset rules; Based on the preset rules, the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data are divided to obtain the second remote sensing image data, the second population flow data, the second infrastructure statistics data, and the second road network data corresponding to each grid unit. The generation of at least two graph structures with the plurality of grid cells as nodes based on the positional relationships between grid cells and the statistical data of the at least two dimensions includes: Based on the positional relationships between grid cells and the second remote sensing image data, second population flow data, second infrastructure statistics data, and second road network data corresponding to each grid cell, at least two graph structures with the plurality of grid cells as nodes are generated.

6. The method according to claim 1, characterized in that, The calculation of the target GDP information corresponding to the target area in the preset geographical region based on the trained regression model includes: Obtain a test data set, the test data set including second gross production value information corresponding to a second preset number of second grid cells in the plurality of grid cells; The accuracy of the trained regression model was tested using the test dataset. When the result of the accuracy test is greater than a preset threshold, the target GDP information corresponding to the target area in the preset geographical region is calculated based on the trained regression model.

7. The method according to claim 1, characterized in that, The calculation of the target GDP information corresponding to the target area in the preset geographical region based on the trained regression model includes: In the plurality of grid cells, a plurality of third grid cells corresponding to the target region are determined, and a third feature vector corresponding to each third grid cell is determined; Each third feature vector is input into the trained regression model to obtain the third GDP information corresponding to each third grid cell, and the target GDP information of the target region is calculated based on the third GDP information.

8. A data processing apparatus, characterized in that, The device includes: The acquisition unit is used to acquire statistical data of at least two dimensions related to gross domestic product information for each of the multiple grid cells, wherein the multiple grid cells are multiple sub-regions obtained by dividing a preset geographical region according to preset rules; A generation unit is used to generate at least two graph structures with the plurality of grid cells as nodes based on the positional relationship between grid cells and the statistical data of the at least two dimensions; A learning unit is used to construct a graph neural network model based on the at least two graph structures, and to perform representation learning on each grid cell based on the graph neural network model to obtain the feature vector corresponding to each grid cell. The training unit is used to acquire the first gross domestic product information corresponding to a preset number of first grid cells in the plurality of grid cells, and to train a preset regression model based on the first feature vectors corresponding to the preset number of first grid cells and the first gross domestic product information to obtain the trained regression model. The calculation unit is used to calculate the target gross domestic product information corresponding to the target area in the preset geographical area based on the trained regression model. The generation unit includes: The first generating sub-unit is used to acquire positional relationship data between grid units and generate a first graph structure between grid units based on the positional relationship data; The second generation subunit is used to extract functional attribute relationship data between grid units from the statistical data of the at least two dimensions, and generate a second graph structure between grid units based on the functional attribute relationship data; The third generation subunit is used to extract interaction relationship data between grid units from the statistical data of the at least two dimensions, and generate a third graph structure between grid units based on the interaction relationship data; The learning unit is also used for: A graph neural network is constructed based on the first graph structure, the second graph structure, and the third graph structure.

9. The apparatus according to claim 8, characterized in that, The first generating subunit includes: The determination module is used to determine the adjacency relationship between grid cells based on the positional relationship data; The generation module is used to generate an adjacency matrix based on the adjacency relationship; A construction module is used to construct a first graph structure between the grid cells, with the plurality of grid cells as nodes and the adjacency matrix as edges.

10. The apparatus according to claim 9, characterized in that, The learning unit includes: The learning sub-unit is used to perform unsupervised learning training on multiple sub-graph neural network models contained in the graph neural network based on a preset loss function, so as to obtain multiple trained sub-graph neural network models. The input sub-unit is used to input statistical data of at least two dimensions associated with the gross domestic product information corresponding to each grid unit into the trained multiple subgraph neural network model to obtain multiple sub-feature vectors corresponding to each grid unit; Determine sub-cells, which are used to determine the feature vector corresponding to each grid cell based on multiple sub-feature vectors corresponding to each grid cell.

11. The apparatus according to claim 10, characterized in that, The determined subunit includes: The acquisition module is used to obtain the weight coefficients corresponding to each sub-feature vector; The summation module is used to perform a weighted summation of the multiple sub-feature vectors based on the weight coefficients corresponding to each sub-feature vector, so as to obtain the feature vector corresponding to each grid cell.

12. The apparatus according to claim 8, characterized in that, The acquisition unit includes: The first acquisition subunit is used to acquire first remote sensing image data, first population flow data, first infrastructure statistics data, and first road network data of a preset geographical area; The first sub-unit is used to divide the geographical area based on preset rules to obtain multiple grid units; The second sub-unit is used to divide the first remote sensing image data, the first population flow data, the first infrastructure statistics data, and the first road network data based on the preset rules, so as to obtain the second remote sensing image data, the second population flow data, the second infrastructure statistics data, and the second road network data corresponding to each grid unit. The generation unit is further configured to: Based on the positional relationships between grid cells and the second remote sensing image data, second population flow data, second infrastructure statistics data, and second road network data corresponding to each grid cell, at least two graph structures with the plurality of grid cells as nodes are generated.

13. The apparatus according to claim 8, characterized in that, The computing unit includes: The second acquisition subunit is used to acquire a test data set, the test data set including second gross production value information corresponding to a second preset number of second grid units in the plurality of grid units; The test subunit is used to perform accuracy testing on the trained regression model using the test dataset. The first calculation subunit is used to calculate the target GDP information corresponding to the target area in the preset geographical area based on the trained regression model when the result index value of the accuracy test is greater than a preset threshold.

14. The apparatus according to claim 8, characterized in that, The computing unit includes: The second determining subunit is used to determine a plurality of third grid units corresponding to the target region among the plurality of grid units and to determine a third feature vector corresponding to each third grid unit; The second calculation subunit is used to input each third feature vector into the trained regression model to obtain the third gross product information corresponding to each third grid cell, and to calculate the target gross product information of the target region based on the third gross product information.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the steps of the data processing method according to any one of claims 1 to 7.

16. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the data processing method according to any one of claims 1 to 7.

17. A computer program, characterized in that, The computer program includes computer instructions stored in a storage medium. A processor of the computer device reads the computer instructions from the storage medium and executes the computer instructions, causing the computer device to perform the steps of the data processing method according to any one of claims 1 to 7.

18. A computer program product, characterized in that, The computer program product includes computer instructions stored in a storage medium; the processor of the computer device reads the computer instructions from the storage medium and executes the computer instructions, causing the computer device to perform the steps of the data processing method according to any one of claims 1 to 7.

Images