Method, device, equipment and system for traffic prediction

Abstract

A traffic prediction method, device, equipment and system are disclosed, and relate to the field of computers. The method comprises obtaining map data of a target area, obtaining spatial features of the target area according to the map data, and predicting traffic of the target area according to the spatial features of the target area. The map data is used to indicate objects and object features contained in the target area, and the spatial features are used to indicate distribution conditions of the objects in the target area. The traffic prediction method provided in the application uses easily obtained map data to construct spatial features, and predicts traffic for the target area according to the spatial features, thereby improving the accuracy of traffic prediction.

Description

Technical Field

[0001] This application relates to the field of computers, and more particularly to a method, apparatus, device, and system for traffic prediction. Background Technology

[0002] Currently, traffic flow prediction for a region is based on dynamic data, such as network traffic, traffic volume, and user traffic. Dynamic data includes location data and signal quality-related data, such as GPS and Measurement Report (MR) data. However, dynamic data involves user privacy, is difficult to obtain, and has limited use. Furthermore, dynamic data is updated frequently, requiring frequent collection, storage, and processing of large-scale data, resulting in high computational complexity. Therefore, predicting regional traffic flow based on dynamic data leads to relatively low accuracy. Summary of the Invention

[0003] This application provides a method, apparatus, device, and system for traffic forecasting, which improves the accuracy of traffic forecasting.

[0004] Firstly, a traffic flow prediction method is provided. This method includes acquiring map data of a target area, obtaining spatial features of the target area based on the map data, and predicting traffic flow in the target area based on the spatial features. The map data indicates the objects and their characteristics within the target area. The spatial features indicate the distribution of objects within the target area.

[0005] The method provided in this application uses map data to predict traffic flow. Since map data is readily available and fully represents the objects and relationships within the target area, analyzing the map data yields the distribution of objects in the target area, such as spatial information in multiple dimensions like points, lines, and polygons within the target area. Based on this information, the traffic flow in the target area can be predicted, which can effectively improve the accuracy of traffic flow prediction.

[0006] Furthermore, since the traffic prediction method provided in this application does not require the use of dynamic data (e.g., historical traffic data) to predict traffic in the target area, but instead uses readily available map data, the cold start problem is solved. In addition, the map data fully represents the objects and relationships within the target area, enabling the traffic prediction method provided in this application to adapt to the geographical differences in different regions and improving cross-regional migration capabilities.

[0007] In one possible implementation, obtaining spatial features of the target area based on map data includes: obtaining spatial features of multiple raster cells in the target area based on map data, wherein the spatial features of the raster cells indicate the distribution of objects within the raster cells. Predicting traffic in the target area based on the spatial features of the target area includes: predicting traffic in multiple raster cells based on the spatial features of the multiple raster cells, and obtaining the traffic in the target area based on the traffic in the multiple raster cells.

[0008] By dividing the target area into grid cells, spatial features of smaller-granular grid cells are constructed for the target area, thereby enabling traffic prediction of smaller-granular grid cells in the target area and further improving the accuracy of traffic prediction.

[0009] In another possible implementation, object characteristics include at least one of object relationships, object shape, and object size.

[0010] Since object features represent the characteristics of objects within a target area and the characteristics between objects, spatial features constructed from map data can express the distribution of objects within the target area, such as spatial information in multiple dimensions like points, lines, and polygons. Therefore, traffic flow prediction based on the distribution of objects within the target area effectively improves the accuracy of traffic flow prediction.

[0011] In another possible implementation, the spatial features include at least one of the density or coverage of objects contained in the target region.

[0012] The distribution of objects in a target area can be represented by quantifiable indicators such as density and coverage. Based on the distribution of objects in the target area, traffic prediction can be performed on the target area, which effectively improves the accuracy of traffic prediction and reduces the computational complexity of traffic prediction.

[0013] In another possible implementation, multiple grid cells are of the same size.

[0014] Dividing the target area into grid cells of equal size reduces errors caused by differences in grid size and improves the accuracy of traffic forecasting.

[0015] In another possible implementation, predicting the traffic flow in the target area based on the spatial characteristics of the target area includes: predicting the traffic flow in the target area during at least one time period based on the spatial characteristics of the target area.

[0016] Traffic flow forecasting is performed for different time periods in the target area to obtain the dynamic change pattern of traffic flow over time.

[0017] For example, predicting traffic flow in a target area during specific periods such as holidays can provide a basis for the scheduling of resources such as traffic lights.

[0018] In another possible implementation, predicting the traffic flow of the target area based on the spatial characteristics of the target area includes: inputting the spatial characteristics of the target area into a traffic prediction model and outputting the traffic flow of the target area.

[0019] Using models to predict traffic in target areas makes traffic prediction more intelligent, thereby effectively improving the accuracy of traffic prediction.

[0020] Secondly, a model training method is provided, including: acquiring map data and traffic data of a training area; obtaining spatial features of the training area based on the map data; and obtaining traffic and temporal features of the training area based on the traffic data. An artificial intelligence model is trained based on the spatial, temporal, and traffic features of the training area to obtain a traffic prediction model. Specifically, map data indicates the objects and object characteristics contained in the training area; spatial features indicate the distribution of objects in the training area; temporal features indicate the temporal distribution of traffic in the training area; traffic features indicate the distribution of traffic in the training area; and the traffic prediction model is used to predict the traffic in a target area.

[0021] The model training method provided in this application uses the temporal and spatial characteristics of the training area as input to the model, enabling the model to learn the relationship between traffic flow changes and spatiotemporal distribution, thus obtaining a traffic flow prediction model. Therefore, by training the traffic flow prediction model from both temporal and spatial dimensions, the model can effectively improve the accuracy of traffic flow prediction.

[0022] In another possible implementation, spatial features of the training area are obtained from map data, including: obtaining spatial features of multiple raster cells in the training area from map data, wherein the spatial features of the raster cells are used to indicate the distribution of objects in the raster cells, and the spatial features of the training area include the spatial features of multiple raster cells.

[0023] By dividing the training area into grid cells and constructing spatial features for smaller grid cells, the resulting traffic prediction model can predict traffic for even smaller grid cells.

[0024] In another possible implementation, the traffic characteristics and temporal characteristics of the training region are obtained based on traffic data, including: obtaining the temporal characteristics and traffic characteristics of multiple grid cells in the training region based on traffic data. The temporal characteristics of the grid cells indicate the temporal distribution of traffic within the grid cells, the traffic characteristics of the grid cells indicate the traffic distribution within the grid cells, the traffic characteristics of the training region include the traffic characteristics of multiple grid cells, and the temporal characteristics of the training region include the temporal characteristics of multiple grid cells.

[0025] By dividing the training area into grid cells, time features and flow features are constructed for smaller grid cells, and the resulting flow prediction model can predict flow for smaller grid cells.

[0026] In another possible implementation, the traffic data includes at least one of location data and signal quality-related data.

[0027] By using traffic data as the true value and the output of the traffic prediction model as the predicted value, a traffic prediction model with higher prediction accuracy is obtained by minimizing the error between the true value and the predicted value.

[0028] In another possible implementation, the spatial features include at least one of the density or coverage of objects contained in the training region.

[0029] The distribution of objects in the training area is represented by quantifiable metrics such as density and coverage. A traffic prediction model is then trained based on this distribution. This enables the traffic prediction model to make traffic predictions, effectively improving the accuracy of traffic prediction.

[0030] In another possible implementation, a traffic prediction model is obtained by training a model based on spatial features, temporal features, and traffic flow features. This includes: inputting spatial features and temporal features into an artificial intelligence model to obtain the predicted traffic flow for the training area; adjusting the artificial intelligence model based on the error between the traffic flow features and the predicted traffic flow until the artificial intelligence model converges to obtain the traffic prediction model.

[0031] During model training, the parameters of the artificial intelligence model are dynamically adjusted using the model's loss function. The loss function measures the error between traffic characteristics and predicted traffic. When the error is large, the parameters of the artificial intelligence model are adjusted to reduce the error. This process continues until the loss function is minimized, resulting in a traffic prediction model. This enables the traffic prediction model to make traffic predictions, effectively improving the accuracy of traffic prediction.

[0032] Thirdly, a data processing apparatus is provided, comprising modules for performing operational steps of the method in the first aspect or any possible implementation thereof. For example, the data processing apparatus includes a communication module and a processing module.

[0033] The communication module is used to acquire map data of the target area, which indicates the objects and object characteristics contained in the target area; the processing module is used to acquire spatial characteristics of the target area based on the map data, which indicates the distribution of objects in the target area; the processing module is also used to predict the traffic flow of the target area based on the spatial characteristics of the target area.

[0034] In one possible implementation, when the communication module obtains the spatial features of the target area based on map data, specifically: the processing module obtains the spatial features of multiple raster cells in the target area based on the map data, whereby the spatial features of the raster cells indicate the distribution of objects within the raster cells. When the processing module predicts the traffic flow of the target area based on the spatial features of the target area, specifically: the processing module predicts the traffic flow of multiple raster cells based on the spatial features of the multiple raster cells, and obtains the traffic flow of the target area based on the traffic flow of the multiple raster cells.

[0035] In another possible implementation, object characteristics include at least one of object relationships, object shape, and object size.

[0036] In another possible implementation, the spatial features include at least one of the density or coverage of objects contained in the target region.

[0037] In another possible implementation, multiple grid cells are of the same size.

[0038] In another possible implementation, when the processing module predicts the traffic flow of the target area based on the spatial characteristics of the target area, it is specifically used to: predict the traffic flow of the target area in the first time period based on the spatial characteristics of the target area.

[0039] In another possible implementation, when the processing module predicts the flow of the target area based on the spatial characteristics of the target area, it specifically inputs the spatial characteristics of the target area into the flow prediction model and outputs the flow of the target area.

[0040] Fourthly, a data processing apparatus is provided, comprising modules for performing operational steps of the method in the second aspect or any possible implementation thereof. For example, the data processing apparatus includes a communication module and a processing module.

[0041] The communication module acquires map and traffic data for the training area. The map data indicates the objects and their characteristics within the training area. The processing module acquires spatial features of the training area from the map data and traffic and temporal features from the traffic data. Spatial features indicate the distribution of objects in the training area, while temporal features indicate the temporal distribution of traffic. The processing module also trains an artificial intelligence model based on the spatial, temporal, and traffic features of the training area to obtain a traffic prediction model. This model is used to predict traffic in the target area.

[0042] In another possible implementation, when the processing module obtains the spatial features of the training area based on the map data, it is specifically used to: obtain the spatial features of multiple grid cells in the training area based on the map data, and the spatial features of the grid cells are used to indicate the distribution of objects in the grid cells.

[0043] In another possible implementation, when the processing module obtains the traffic and temporal characteristics of the training area based on the traffic data, it specifically acquires the temporal and traffic characteristics of multiple grid cells in the training area based on the traffic data. The temporal characteristics of the grid cells are used to indicate the temporal distribution of traffic within the grid cells, and the traffic characteristics of the grid cells are used to indicate the traffic distribution within the grid cells.

[0044] In another possible implementation, the traffic data includes at least one of location data and signal quality-related data.

[0045] In another possible implementation, the spatial features include at least one of the density or coverage of objects contained in the training region.

[0046] In another possible implementation, when the processing module trains the model to obtain the traffic prediction model based on spatial features, temporal features, and traffic features, it is specifically used to: input the spatial features and temporal features into the artificial intelligence model to obtain the predicted traffic of the training area; adjust the artificial intelligence model according to the error between the traffic features and the predicted traffic until the artificial intelligence model converges to obtain the traffic prediction model.

[0047] Fifthly, a computer device is provided, the computer device including a memory and a plurality of processors, the memory being used to store a set of computer instructions; when the processors execute the set of computer instructions, the plurality of processors jointly execute the operation steps of the method in the first aspect or any possible implementation of the first aspect, or execute the operation steps of the method in the second aspect or any possible implementation of the second aspect.

[0048] In a sixth aspect, a traffic prediction system is provided, which is used to perform the operational steps of the method in the first aspect or any possible implementation of the first aspect to achieve traffic prediction for a target area.

[0049] In a seventh aspect, a model training system is provided, which is used to execute the operational steps of the method in the second aspect or any possible implementation of the second aspect to train an artificial intelligence model to obtain a traffic prediction model.

[0050] Eighthly, a computer-readable storage medium is provided, comprising: computer software instructions; when the computer software instructions are executed in a processor, causing the processor to perform operation steps of the method as described in the first aspect or any possible implementation thereof, or to perform operation steps of the method as described in the second aspect or any possible implementation thereof.

[0051] Ninthly, a computer program product is provided that, when run on a computer, causes the computer to perform operational steps of the method as described in the first aspect or any possible implementation thereof, or to perform operational steps of the method as described in the second aspect or any possible implementation thereof.

[0052] The technical effects of any of the design methods in aspects two through nine can be found in aspect one or in different design methods in aspect one, and will not be repeated here.

[0053] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods. Attached Figure Description

[0054] Figure 1 A schematic diagram of a traffic prediction system provided for the prior art;

[0055] Figure 2 A schematic diagram illustrating an alternative traffic prediction method provided by existing technology;

[0056] Figure 3 A schematic diagram of a traffic prediction system architecture is provided for this application;

[0057] Figure 4 A schematic diagram of the logical architecture of a traffic prediction system provided in this application;

[0058] Figure 5 A flowchart illustrating a traffic prediction method provided in this application;

[0059] Figure 6 A flowchart illustrating a model training method provided in this application;

[0060] Figure 7 A schematic diagram of the structure of a data processing device provided in this application;

[0061] Figure 8 This is a schematic diagram of the structure of a computer device provided in this application. Detailed Implementation

[0062] To better understand the embodiments of this application, some terms or technologies involved in the embodiments of this application will be explained below.

[0063] Traffic forecasting refers to predicting traffic flow in a region over a specific time period by analyzing historical traffic data, and then using the predicted traffic to optimize resource allocation or provide decision support. Traffic flow includes, but is not limited to, network traffic, traffic flow, or user traffic. Historical traffic data refers to traffic data that has already been generated within the region.

[0064] For example, in the telecommunications sector, by predicting network traffic within a region, base station coverage and bandwidth allocation can be dynamically adjusted. During peak traffic periods, network capacity can be optimized in advance to reduce dropped calls and buffering, thus improving the user communication experience. During off-peak traffic periods, energy conservation can be achieved by reducing base station power or shutting down some base station functions, and resource waste can be reduced by dynamically allocating bandwidth.

[0065] For example, in the business sector, by predicting user traffic within a region, advertising can be placed in areas and time periods with high traffic, thereby reducing decision-making risks and increasing revenue.

[0066] In some embodiments, a traffic prediction model is trained based on historical traffic data, and the traffic prediction result is output through this model. For example, Figure 1 A schematic diagram of a traffic prediction system provided by existing technology. (Example) Figure 1 As shown, the traffic prediction system 110 includes a preprocessing module 111, an encoding network 112, an attention mechanism 113, and a prediction network 114. The components of the encoding network 112 and the prediction network 114 include, but are not limited to, convolutional long short-term memory networks.

[0067] The preprocessing module 111 is used to divide the target area into grids, preprocess the historical flow data of the target area, and obtain the inflow and outflow of each grid at different times.

[0068] The encoding network 112 is used to take the inflow and outflow obtained by the preprocessing module 111 as input to obtain the output state of the encoding network at each time.

[0069] The attention mechanism 113 is used to obtain attention weights based on the output state of the encoding network obtained by the encoding network 112 and the state of the prediction network 114 at the previous time step. The attention weights are then weighted and summed with the output state of the encoding network at each time step to obtain the vector at each time step.

[0070] The prediction network 114 is used to take the vector obtained by the attention mechanism 113 as input to obtain the traffic prediction result of the target area.

[0071] Therefore, it is evident that the model requires a large amount of historical traffic data for the target area to predict traffic flow. Without sufficient historical data, the model cannot predict traffic flow in the target area, or the accuracy of the prediction will be low, leading to the cold start problem. The cold start problem refers to the low prediction accuracy of traffic prediction models when historical data is lacking.

[0072] In other embodiments, a traffic prediction model is trained based on historical route data and urban road network data, and the traffic prediction result is obtained through this model. For example, Figure 2 A schematic diagram of a traffic prediction method provided for existing technologies, such as... Figure 2 As shown, the method may include the following steps.

[0073] Step 210: Divide the irregular area.

[0074] The region is divided into multiple unconnected, irregular areas using urban road network data.

[0075] Step 220: Calculate the inflow and outflow of the region.

[0076] Spatially simplify the historical route data and calculate the inflow and outflow of all irregular areas obtained in step 210 at each time point.

[0077] Step 230: Create a relationship graph.

[0078] For all irregular regions obtained in step 210, establish an inter-regional association graph and construct the corresponding adjacency matrix. The adjacency matrix is ​​used to represent the spatial association between irregular regions.

[0079] Step 240: Spatial association modeling.

[0080] Based on the inter-regional correlation graph obtained in step 230, a multi-graph convolutional neural network is designed to fuse diverse spatial correlation features between regions, resulting in a multi-graph convolutional fusion result.

[0081] Step 250: Time-related modeling.

[0082] Based on the multi-graph convolution fusion result obtained in step 240, a gated recurrent unit (GRU) is used to capture temporal correlations.

[0083] Step 260: Traffic prediction results for irregular areas.

[0084] Based on the spatial correlation obtained in step 230 and the temporal correlation obtained in step 250, a flow prediction model is trained, and the inflow and outflow of all irregular areas are obtained through the flow prediction model.

[0085] The traffic prediction model can only predict traffic in the region used when training the model, and cannot predict traffic in other regions, which leads to poor cross-regional migration capability.

[0086] Cross-regional migration capability refers to the ability of a traffic prediction model to predict traffic flow when applied to different regions. For example, a traffic prediction model trained on historical traffic data from region A, when applied to region B for traffic prediction, will result in low prediction accuracy, indicating poor cross-regional migration capability of the traffic prediction system. To address the problem of low accuracy in traffic prediction, this application provides a traffic prediction method. The method includes acquiring map data of a target region, obtaining spatial features of the target region based on the map data, and predicting traffic flow in the target region based on the spatial features. The map data is used to indicate the objects contained in the target region and their characteristics. The spatial features are used to indicate the distribution of objects in the target region.

[0087] Compared to using historical traffic data to predict traffic in a region, the current method suffers from low accuracy in traffic prediction. The method provided in this application uses map data to predict traffic. Since map data is readily available and fully represents the objects and relationships within the target area, analyzing the map data reveals the distribution of objects in the target area—that is, spatial information in multiple dimensions such as points, lines, and polygons. Based on this information, the accuracy of traffic prediction for the target area can be effectively improved.

[0088] Furthermore, since the traffic prediction method provided in this application does not require the use of dynamic data (e.g., historical traffic data) to predict traffic in the target area, but instead uses readily available map data, the cold start problem is solved. In addition, the map data fully represents the objects and relationships within the target area, enabling the traffic prediction method provided in this application to adapt to the geographical differences in different regions and improving cross-regional migration capabilities.

[0089] The following is a detailed description of a traffic prediction method provided in this application, with reference to the accompanying drawings.

[0090] Figure 3 This is a schematic diagram of the architecture of a traffic prediction system provided in this application. Figure 3 As shown, the traffic prediction system 300 includes a client 310, a computing cluster 320, and a storage cluster 330.

[0091] Storage cluster 330 comprises multiple storage nodes 331. Each storage node 331 includes one or more controllers, a network interface card (NIC), and multiple hard drives. Hard drives are used to store data. Hard drives are disks or other types of storage media, such as solid-state drives (SSDs) or shingled magnetic recording (SMR) hard drives. The NICs are used to communicate with the compute nodes 321 included in compute cluster 320. The controllers are used to write data to or read data from the hard drives based on read / write data requests sent by the compute nodes 321. During the read / write process, the controller needs to translate the address carried in the read / write data request into an address that the hard drive can recognize.

[0092] The computing cluster 320 contains multiple computing nodes 321. For example, a computing node 321 is a computing device, such as a server.

[0093] In some embodiments, the computing cluster 320 is a heterogeneous computing architecture to provide high-performance computing. For example, computing nodes 321 include computing units with computing capabilities such as central processing units (CPUs), graphics processing units (GPUs), data processing units (DPUs), neural processing units (NPUs), and neural-network processing units (NPUs) to provide high-performance computing.

[0094] In other embodiments, multiple computing nodes 321 are connected via network devices (such as switches, network interface cards, etc.) based on high-speed interconnect technology, enabling communication between the multiple computing nodes 321.

[0095] In this application, computing node 321 is used to obtain the spatial characteristics of the target area based on map data, and predict the traffic flow of the target area based on the spatial characteristics of the target area.

[0096] Optionally, computing node 321 is used to train a traffic prediction model, which is used to predict traffic in a target area.

[0097] In some embodiments, client 310 communicates with computing cluster 320 and storage cluster 330 via network 340. For example, client 310 sends a request to computing cluster 320 via network 340, requesting computing cluster 320 to perform traffic prediction. Network 340 refers to an enterprise intranet (e.g., a local area network, LAN) or the Internet. Client 310 refers to a computer connected to network 340, also known as a workstation. Different clients share network resources (e.g., computing resources, storage resources).

[0098] Optionally, the computing cluster 320 may also include a control node 322. For example, the control node and the computing nodes may be independent physical devices. Alternatively, the control node and multiple computing nodes may reside on the same physical device. The control node is the CPU. The multiple computing nodes include computing units such as GPUs, NPUs, and DPUs. The control node 322 is used to manage and allocate traffic prediction tasks, with multiple computing nodes executing multiple tasks in parallel to improve the rate of traffic prediction.

[0099] Optionally, the control node 322 is used to instruct multiple computing nodes to perform distributed traffic prediction based on the above embodiments upon request.

[0100] In this embodiment, storage cluster 330 stores map data, traffic data, and parameters. Parameters include model parameters and optimizer parameters, etc.

[0101] In other embodiments, client 310 has client program 311 installed. Client 310 runs client program 311 and displays a user interface (UI). User 350 manipulates the user interface to submit a request. For example, user 350 manipulates the user interface to submit a request. After receiving the request, control node 322 loads map data from storage cluster 330 and distributes the map data to multiple computing nodes, enabling the multiple computing nodes to perform traffic prediction tasks.

[0102] Optionally, the system administrator 360 may configure system information, such as the map data provided in this application, by calling the application platform interface (API) 312 or the command-line interface (CLI) 313 through the client 310.

[0103] Figure 3 This is merely an illustrative diagram; the embodiments of this application do not limit the device connection method or the number of devices in the traffic prediction system. For example, the traffic prediction system includes multiple clients. One client connects to multiple computing nodes. Different clients establish connections with different computing nodes.

[0104] Figure 4 This is a schematic diagram of the logical architecture of a traffic prediction system provided in this application. Figure 4 As shown, the traffic prediction system 400 includes a data layer 410, a representation layer 420, a model layer 430, and an optimization layer 440.

[0105] Data layer 410 includes a map database 411 and a traffic database 412. Map database 411 provides map data. Traffic database 412 provides traffic data. The map data and traffic data are used by representation layer 420 to construct features.

[0106] Map data refers to a collection of data used to describe the location and attribute information of map elements within a specific area. Map elements include at least one of point elements, line elements, or polygon elements. Point elements include, but are not limited to, points of interest (POIs). Linear elements include at least one of road networks and waterways. Polygon elements include at least one of areas of interest (AOIs), building information, land use types, or community divisions. Location information includes, but is not limited to, spatial coordinates composed of latitude and longitude. Attribute information includes, but is not limited to, the name, type, length, or area of ​​the map element.

[0107] For example, map elements in a certain area may include water systems and areas of interest (AOIs). The attribute information of a water system includes its name, type, and length, while the attribute information of an AOI includes its name, type, and area.

[0108] Traffic data includes at least one of location data and signal quality-related data. For example, location data includes, but is not limited to, GPS data, and signal quality-related data includes, but is not limited to, MR data.

[0109] The representation layer 420 is used to acquire map data and traffic data from the data layer 410, and to construct features based on the map data and traffic data. For example, map data is acquired from the map database 411, and traffic data is acquired from the traffic database 412.

[0110] In some embodiments, the characterization layer 420 includes a spatial feature construction module 421, a temporal feature construction module 422, and a flow feature construction module 423.

[0111] The spatial feature construction module 421 is used to construct spatial features based on map data. Spatial features refer to the distribution of objects in a target area. Objects are, for example, map elements, which include point elements, line elements, and polygon elements. For example, spatial features include, but are not limited to, the density of point elements, the density of line elements, or the coverage of polygon elements in the target area.

[0112] The time feature construction module 422 is used to construct time features based on traffic data. Time features are used to indicate the temporal distribution of traffic in the training area. For example, the time period in which traffic occurs in the training area can be any of the following: hour, day, week, or month. Another example is the morning peak period (e.g., 7:00 AM to 9:00 AM), which represents the network traffic generated in the training area during the morning peak period.

[0113] The traffic feature construction module 423 is used to construct traffic features based on traffic data. Traffic features are used to indicate the traffic distribution in the training area. For example, traffic includes any one of traffic flow, user flow, or network traffic. Traffic distribution includes at least one of uplink traffic or downlink traffic. For instance, uplink traffic in network traffic refers to the traffic consumed by local devices within the training area sending data to a remote server, while downlink traffic refers to the traffic consumed by the remote server sending data to local devices within the training area.

[0114] In some embodiments, the model layer 430 includes a training module 431, a transfer module 432, and a processing module 433.

[0115] Training module 431 is used to train a traffic prediction model based on the spatial, temporal, and traffic features constructed by representation layer 420. For example, the model includes, but is not limited to, an artificial intelligence model.

[0116] In some scenarios, the training module 431 inputs the spatial and temporal features constructed by the representation layer 420 into the artificial intelligence model to obtain the predicted flow of the training area. The artificial intelligence model is then adjusted based on the error between the predicted flow and the flow features constructed by the representation layer 420 until the artificial intelligence model converges, thus obtaining the flow prediction model.

[0117] The migration module 432 is used to predict traffic flow in a target area using a traffic prediction model and to fine-tune the traffic prediction model. For example, it can predict traffic flow in area B based on a traffic prediction model trained using map data and traffic data contained in area A.

[0118] Optionally, a traffic prediction model is trained based on region A. When using this model to predict traffic in region B, the model is fine-tuned. For example, the model can be incrementally trained using traffic data from region B, and its parameters can be fine-tuned to achieve better prediction results in region B. The parameters of the traffic prediction model include, but are not limited to, weights, bias terms, or learning rates.

[0119] The processing module 433 takes the spatial features obtained by the spatial feature construction module 421 as input and uses the traffic prediction model to obtain the traffic prediction value of the target area in the first time period.

[0120] Optionally, the processing module 433 is further configured to take the temporal features obtained by the temporal feature construction module 422 and the spatial features obtained by the spatial feature construction module 421 as inputs, and obtain the traffic prediction value of the target area in the first time period through the traffic prediction model. The temporal features indicate the first time period.

[0121] The optimization layer 440 receives the traffic prediction results from the processing module 433 in the model layer 430 and adjusts the resource configuration based on the traffic prediction results. For example, it adjusts the coverage, direction, or power of the base station antennas based on the obtained network traffic prediction values.

[0122] The optimization layer 440 includes a short-cycle optimization strategy module 441 and a long-cycle optimization strategy module 442.

[0123] The short-cycle optimization strategy module 441 is used to adjust resource allocation in real time based on traffic forecast values ​​to cope with short-term traffic fluctuations. For example, it dynamically adjusts the coverage, orientation, and power of base station antennas based on traffic forecasts for the next 15 minutes.

[0124] The long-term optimization strategy module 442 is used for structured resource planning based on long-term traffic trends. For example, it plans the deployment of new base stations or the upgrading of existing hardware based on annual traffic forecasts.

[0125] Next, the method for traffic flow prediction will be explained in detail with reference to the attached diagram.

[0126] Figure 5 This is a flowchart illustrating a traffic prediction method provided in this application. Here, we mainly focus on... Figure 3 or Figure 4 The diagram illustrates the architecture of a traffic prediction system. For example, the traffic prediction method is executed by a server. Figure 5 As shown, the method includes the following steps.

[0127] Step 510: Divide the target area into multiple units.

[0128] This application does not limit the size, shape, or method of division of the units. For example, the target area can be divided into multiple units of equal size. For instance, the size of a unit could be 50 meters × 50 meters. Optionally, the shape of the unit can be rectangular, square, or irregular.

[0129] For example, at least two of the multiple cells may have different sizes. For instance, cell sizes can be set based on the distribution of map elements in the target area, dividing areas with densely distributed map elements into smaller cells and areas with sparsely distributed map elements into larger cells.

[0130] Optionally, the cell is also called a grid cell.

[0131] Optionally, each unit can be assigned a unique number. This number identifies the corresponding unit. For example, combining the number of each unit in the target area with the traffic prediction result for that unit yields the traffic prediction result for the target area.

[0132] In other embodiments, different configuration methods are provided for receiving user input of the prediction region and cell size. These different configuration methods include, but are not limited to, page-based or background configuration options. Pages include, but are not limited to, web pages or application pages, while background configuration options include, but are not limited to, configuration files, environment variables, or databases. For example, a page can provide selection functionality for regions, cell sizes, or cell shapes. The client receives the region, cell size, or cell shape selected by the user through the page.

[0133] Optionally, step 510 is an optional step.

[0134] Step 520: Obtain map data for the target area.

[0135] In some embodiments, map data is used to indicate the objects and object characteristics contained within a target area. For example, an object refers to a map element, which includes point elements, line elements, and polygon elements.

[0136] Point elements are used to represent a specific geographical location or object on a map, including but not limited to Points of Interest (POIs). A POI refers to a point element on a map that has a specific function or meaning, such as a school, hospital, or restaurant.

[0137] Line elements are used to represent connections or routes between geographical locations or objects, including but not limited to road networks or waterways.

[0138] Polygonal elements are used to represent areas with boundaries and areas on a map, including but not limited to areas of interest (AOI), building information, land use types, or community divisions. AOI refers to polygonal elements on a map that have specific functions or meanings, such as commercial areas, residential areas, or industrial areas.

[0139] Object characteristics include at least one of object relationships, object shape, or object size. For example, object relationships refer to the location information of map elements. Object shape includes the shape of line elements and polygon elements. For example, the shape of a line element includes a straight line or a curve. For instance, a straight line represents the shape of a road in a line element. A curve represents the shape of a water system in a line element. The shape of a polygon element includes rectangles or irregular polygons. Object size refers to information such as the length of a line element or the area of ​​a polygon element.

[0140] Optionally, map data can be retrieved from a map database. This map database is deployed on a server other than the one executing the traffic prediction method.

[0141] Step 530: Obtain the spatial characteristics of the target area based on map data.

[0142] Spatial features are used to indicate the distribution of objects within a target area. In some embodiments, spatial features include at least one of the density of point elements, the density of line elements, or the coverage of area elements within the target area.

[0143] For example, the density of point elements refers to the POI density. POI density is obtained by the ratio of the number of POIs within a target area to the area of ​​the target area. In some embodiments, POIs are categorized based on different objects, including but not limited to hospitals, schools, or restaurants. For each type of POI, the POI density is calculated. For example, POI density includes hospital density, school density, or restaurant density. For instance, assuming the target area is 5 square kilometers, and there are 5 hospitals, 15 schools, and 80 restaurants within the target area, the hospital density is 1 / square kilometer, the school density is 3 / square kilometer, and the restaurant density is 16 / square kilometer.

[0144] The density of line elements is obtained by the ratio of the length of line elements within the target area to the area of ​​the target area. For example, the density of line elements includes, but is not limited to, road network density or water system density.

[0145] For example, road network density is obtained by the ratio of the length of the road network within the target area to the area of ​​the target area. In some embodiments, the types of road networks include, but are not limited to, highways, arterial roads, and secondary arterial roads. For each type of road network, the road network density is calculated. For example, road network density includes highway density, arterial road density, or secondary arterial road density. For instance, assuming the target area (e.g., an urban area) has an area of ​​1000 square kilometers, the length of highways within the target area is 100 kilometers, the length of arterial roads is 400 kilometers, and the length of secondary arterial roads is 800 kilometers, the highway density of the target area is 0.1 kilometers / square kilometer, the arterial road density is 0.4 kilometers / square kilometer, and the secondary arterial road density is 0.8 kilometers / square kilometer.

[0146] Water system types include, but are not limited to, main streams, tributaries, or streams. The calculation method for water system density can be referenced from the calculation method for road network density, and will not be repeated here.

[0147] The coverage rate of a surface element is obtained by the ratio of the area of ​​the surface element within the target area to the area of ​​the target area. For example, the coverage rate of a surface element includes, but is not limited to, at least one of AOI coverage rate, building coverage rate, land use type coverage rate, or community zoning coverage rate.

[0148] For example, AOI coverage is obtained by the ratio of the area of ​​AOIs within the target area to the area of ​​the target area. In some embodiments, AOIs are categorized according to different regions, including but not limited to commercial areas or residential areas. For each type of AOI, an AOI coverage rate is calculated. For example, AOI coverage rates include commercial area coverage rate or residential area coverage rate. For instance, assuming the target area is 100 square kilometers, the commercial area within the target area is 10 square kilometers, and the residential area is 70 square kilometers, then the commercial area coverage rate of the target area is 10%, and the residential area coverage rate is 70%.

[0149] Building types include, but are not limited to, industrial or public buildings. Land use types include, but are not limited to, residential or public service facility land. Community division includes, but is not limited to, the administrative divisions of the target area. The calculation methods for building coverage, land use type coverage, and community division coverage can be found in the calculation method for AOI coverage, and will not be repeated here.

[0150] Optionally, in some embodiments, the temporal characteristics of the target area are obtained. Temporal characteristics refer to the temporal distribution of traffic flow in the target area. For example, the time period to which the traffic flow in the target area belongs is predicted, including but not limited to hours, days, weeks, months, whether it is a holiday, or whether it is a weekend. Another example is predicting the traffic flow in the target area during the morning rush hour, and adjusting traffic signal resources in advance based on the predicted traffic flow values ​​to alleviate road congestion in the target area during the morning rush hour.

[0151] In some scenarios, the temporal characteristics are derived from the inherent temporal attributes of the traffic data. For example, traffic data includes location data and signal quality-related data, such as GPS and MR data. The time period in which the traffic occurred can be determined based on the timestamps in the GPS data.

[0152] In other scenarios, different configuration methods are provided for receiving user input time characteristics. These different configuration methods include, but are not limited to, page-based and backend configuration items.

[0153] Step 540: Predict the flow rate of the target area based on the spatial characteristics of the target area.

[0154] In some embodiments, traffic flow in a target area is predicted for at least one time period based on the spatial characteristics of the target area. The spatial characteristics of the target area are input into a traffic prediction model to obtain traffic prediction results for the target area for at least one time period. Traffic flow includes, but is not limited to, network traffic, traffic flow, or user traffic. The traffic prediction model includes, but is not limited to, artificial intelligence models.

[0155] For example, by inputting the spatial characteristics of the target area into a traffic prediction model, we can obtain traffic prediction results for the target area in the first time period and the second time period. For instance, the first time period represents weekends, and the second time period represents weekdays. Based on the network traffic prediction results for the target area on weekends and weekdays, we can formulate optimization strategies, adjust base station coverage in the target area, reduce network resource waste, and improve the user communication experience.

[0156] In some scenarios, time periods are derived from traffic data used to train artificial intelligence models. For example, time periods can be obtained through timestamps in GPS data.

[0157] In other scenarios, at least one time period is set by the user. For example, the user sets at least one time period for the target area through configuration methods such as page or backend configuration items.

[0158] Optionally, the flow rate in the target area is predicted for at least one time period based on the spatial and temporal characteristics of the target area. The spatial and temporal characteristics of the target area are input into the flow prediction model to obtain the flow prediction results for the target area for at least one time period.

[0159] Optionally, traffic data from the target region can be used to incrementally train the traffic prediction model, fine-tuning its parameters. The spatial and temporal features of the target region are then input into the fine-tuned traffic prediction model to obtain traffic prediction results for at least one time period. The parameters of the traffic prediction model include, but are not limited to, weights, bias terms, or learning rates.

[0160] Optionally, after the prediction is completed, the traffic prediction results for the target area under at least one time period are output, and the method further includes step 550.

[0161] Step 550: Output optimization strategy.

[0162] In some embodiments, an optimization strategy generated based on traffic prediction results is output. The optimization strategy is used to adjust resource allocation in the target area or to provide decision support.

[0163] This application does not limit the application area of ​​the optimization strategy. For example, the optimization strategy can be applied to the field of digital antennas, including but not limited to adjusting the coverage, direction, or power of the antenna.

[0164] For example, if there are two units, A and B, in the prediction area, and the predicted traffic value of unit A is greater than that of unit B, the antenna direction of the prediction area can be adjusted according to the prediction results to concentrate the beam towards unit A, thereby increasing the antenna coverage and transmission power of unit A, and achieving the purpose of optimizing network resource allocation and improving user experience.

[0165] The traffic prediction method provided in this application uses map data to predict traffic. Since map data is readily available and fully represents the objects and relationships contained in the target area, the distribution of objects in the target area can be obtained by analyzing the map data. For example, spatial information of points, lines, and polygons in multiple dimensions within the target area can be obtained. Based on this information, the traffic in the target area can be predicted, which can effectively improve the accuracy of traffic prediction.

[0166] Furthermore, since the traffic prediction method provided in this application uses readily available map data to predict traffic in the target area, it solves the cold start problem. Also, the map data fully represents the objects and relationships within the target area, enabling the traffic prediction method provided in this application to adapt to the geographical differences in different regions and improving its cross-regional migration capability.

[0167] The above embodiments illustrate traffic prediction at the target region level. In some embodiments, step 510 is also performed, which involves dividing the target region into multiple units and performing traffic prediction on the target region at the level of the multiple units contained within the target region. For example, steps 530 to 550 include the following method.

[0168] Step 531: Obtain the spatial characteristics of each unit in the target area based on the map data.

[0169] In some embodiments, map elements contained in each cell of the target area are obtained based on the location information of map data in the target area, and spatial features of each cell in the target area are obtained based on the map elements contained in each cell. For example, map elements contained within the range of each cell of the target area are calculated based on the location information of the map elements. The location information includes, but is not limited to, coordinates composed of latitude and longitude.

[0170] The spatial characteristics of a cell refer to the distribution of map elements within each cell. In some embodiments, spatial characteristics include at least one of the density of point elements, the density of line elements, or the coverage of area elements in each cell of the target area.

[0171] Point element density includes, but is not limited to, POI density. Line element density includes, but is not limited to, road network density or water system density. Area element coverage includes, but is not limited to, AOI coverage, building coverage, land use type coverage, or community zoning coverage.

[0172] The density of point elements is obtained by the ratio of the number of POIs within a cell to the cell area. The density of line elements is obtained by the ratio of the length of the line elements within a cell to the cell area. The coverage of polygon elements is obtained by the ratio of the area of ​​the polygon elements within a cell to the cell area. The calculation methods for the density of point elements, the density of line elements, and the coverage of polygon elements can be found in step 530, and will not be repeated here.

[0173] Optionally, in some embodiments, the temporal characteristics of each unit in the target area are obtained. Temporal characteristics refer to the temporal distribution of traffic in each unit of the target area. For example, predicting the time period to which the traffic of each unit in the target area belongs. Time periods include, but are not limited to, hours, days, weeks, months, whether it is a holiday, or whether it is a weekend, etc.

[0174] Step 541: Predict the flow rate of the target area based on the spatial characteristics of each unit in the target area.

[0175] In some embodiments, traffic flow in the target area is predicted for at least one time period based on the spatial characteristics of each unit in the target area. The spatial characteristics of each unit in the target area are input into the traffic prediction model to obtain the traffic prediction result for each unit in the target area for at least one time period. The traffic prediction table is composed of the unit number in the target area and the traffic prediction result for that unit, and the traffic prediction table includes the traffic prediction result for the target area.

[0176] In some scenarios, the time period is determined based on traffic data used to train the artificial intelligence model. In other scenarios, at least one time period is set by the user.

[0177] Optionally, traffic data from the target region can be used to incrementally train the traffic prediction model, fine-tuning its parameters. The spatial and temporal features of each unit in the target region are input into the traffic prediction model to obtain the traffic volume of each unit over at least one time period. Based on the traffic prediction results for each unit over at least one time period, the traffic prediction result for the target region over at least one time period is obtained. The parameters of the traffic prediction model include, but are not limited to, weights, bias terms, or learning rates.

[0178] Optionally, after the prediction is completed, the traffic prediction results for the target area in the first time period are output, and the method also includes step 551.

[0179] Step 551: Output optimization strategy.

[0180] In some embodiments, an optimization strategy generated based on traffic prediction results is output. The optimization strategy is used to adjust the resource configuration of each unit in the target area or to provide decision support. The method for adjusting resource configuration can be found in step 550, and will not be repeated here.

[0181] Next, the model training method will be explained in detail with reference to the attached figures.

[0182] Figure 6 This is a flowchart illustrating a model training method provided in this application, which is mainly used here. Figure 3 or Figure 4 The architecture diagram of the traffic prediction system shown is used for illustration. Figure 6As shown, the method includes the following steps.

[0183] Step 610: Divide the training area into multiple units.

[0184] The method for dividing the training region can be found in step 510, and will not be repeated here.

[0185] Optionally, step 610 is an optional step.

[0186] Step 620: Obtain map data for the training area.

[0187] The method for obtaining map data for the training area can be found in step 520, and will not be repeated here.

[0188] Step 630: Obtain traffic data for the training area.

[0189] In some embodiments, traffic data refers to the amount of data in the training area. Traffic data includes at least one of network traffic data, traffic flow data, or user traffic data. For example, network traffic data includes location data and signal quality-related data, such as GPS data and MR data.

[0190] Alternatively, traffic data can be obtained from a traffic database deployed on a server other than the one executing the traffic prediction method.

[0191] Step 640: Obtain the spatial features of the training area based on the map data.

[0192] Spatial features are used to indicate the distribution of objects in the training region. Spatial features include the density of point elements, the density of line elements, and the coverage of area elements in the training region.

[0193] The calculation method for its spatial characteristics can be found in step 530, and will not be repeated here.

[0194] Step 650: Obtain the traffic characteristics and time characteristics of the training area based on the traffic data.

[0195] In some embodiments, traffic characteristics of the training area are obtained based on location information contained in the traffic data of the training area. Traffic characteristics refer to the traffic distribution in the training area.

[0196] For example, upstream network traffic refers to the data consumed when a local device within the training area sends data to a remote server. Another example is the amount of data consumed when a user uploads a photo to a cloud storage server using their mobile phone. Downstream network traffic refers to the data consumed when a remote server sends data to a local device within the training area. Another example is the amount of data consumed when a user downloads an image from a cloud storage server to their mobile phone.

[0197] In some embodiments, the temporal characteristics of the training region are obtained based on traffic data. Temporal characteristics refer to the temporal distribution of traffic in the training region. For example, the time periods during which traffic is generated in the training region.

[0198] In some scenarios, temporal characteristics are derived from the inherent temporal attributes of the traffic data. For example, based on the temporal attributes in GPS data, information such as the month, week, day, or hour in which traffic was generated in the training area can be obtained.

[0199] In other scenarios, different configuration methods are provided for receiving user input time characteristics. These different configuration methods include, but are not limited to, page-based and backend configuration items.

[0200] Step 660: Model training.

[0201] In some embodiments, an artificial intelligence model is trained based on the spatial, temporal, and traffic characteristics of the training area to obtain a traffic prediction model. The traffic prediction model is used to predict traffic in the target area.

[0202] For example, the spatial and temporal features of the training region are input into the artificial intelligence model to obtain the predicted flow rate for the training region. A loss function is set for the artificial intelligence model to measure the error between the predicted flow rate and the flow rate features obtained in step 650. The parameters of the artificial intelligence model are updated using the backpropagation algorithm and gradient descent optimization method to minimize the loss function. For example, when the error is large, the parameters of the artificial intelligence model are adjusted to reduce the error. This process continues until the artificial intelligence model converges, resulting in the flow rate prediction model.

[0203] This application does not impose any restrictions on the selection of artificial intelligence models and loss functions.

[0204] The above embodiments illustrate model training at the training region level. In some embodiments, step 610 is also performed, which involves dividing the training region into multiple units and training the traffic prediction model at the granularity of the multiple units contained in the training region. For example, steps 640 to 660 include the following methods.

[0205] Step 641: Obtain the spatial features of each unit in the training area based on the map data.

[0206] In some embodiments, map elements contained in each cell of the training area are obtained based on the location information of the map data in the training area, and spatial features of each cell in the training area are obtained based on the map elements contained in each cell. Spatial features are used to indicate the distribution of objects in the training area. Spatial features include the density of point elements, the density of line elements, and the coverage of area elements in the training area.

[0207] The calculation method for spatial features can be found in step 531, and will not be repeated here.

[0208] Step 651: Obtain the traffic characteristics and time characteristics of each unit in the training area based on the traffic data.

[0209] In some embodiments, the traffic characteristics of each unit in the training region are obtained based on the location information contained in the traffic data in the training region. The traffic characteristics refer to the traffic distribution of each unit in the training region.

[0210] In some embodiments, the temporal characteristics of each unit in the training region are obtained based on the location information contained in the traffic data in the training region. The temporal characteristics refer to the temporal distribution of traffic in each unit of the training region.

[0211] Optionally, the temporal features of each unit in the training region can be the same time period or different time periods.

[0212] Step 661: Model training.

[0213] In some embodiments, an artificial intelligence model is trained based on the spatial, temporal, and traffic characteristics of each unit in the training region to obtain a traffic prediction model. The traffic prediction model is used to predict the traffic of each unit in the target region. The training method of the model can be referred to step 660, and will not be repeated here.

[0214] It is understood that, in order to achieve the functions in the above embodiments, the computer device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and method steps described in conjunction with the embodiments disclosed in this application, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application scenario and design constraints of the technical solution.

[0215] The above text combines Figures 1 to 6 The method for traffic prediction provided in this application is described in detail below, and will be combined with... Figure 7 This application describes the apparatus provided according to the present application. These apparatuses are used to implement the functions of the flow prediction system in the above-described method embodiments, and therefore also achieve the beneficial effects of the above-described method embodiments. In this embodiment, the apparatus is as follows: Figure 3 or Figure 4 The traffic prediction system shown is also a module (such as a chip) applied to computer equipment.

[0216] like Figure 7 As shown, the data processing device 700 includes a communication module 710, a processing module 720, and a storage module 730.

[0217] The data processing device 700 is used to achieve the above. Figure 4 The method embodiment shown computes the functionality of the node.

[0218] Communication module 710 is used to acquire map data of the target area. For example, communication module 710 is used to perform... Figure 5 Step 520 in the process. Processing module 720 is used to construct spatial features based on map data and predict traffic flow in the target area based on the spatial features of the target area. For example, processing module 720 is used to perform... Figure 5 Steps 510, 530, and 540 in the process. For example, processing module 720 is used to execute... Figure 6 Steps 610, 640 to 660 are included. Optionally, the communication module 710 is also used to acquire traffic data from the training area. For example, the communication module 710 is used to perform... Figure 6 Step 630 in the process.

[0219] Storage module 730 is used to store map data, traffic data, and traffic prediction results, etc.

[0220] It should be understood that the data processing apparatus 700 in this application embodiment is implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). The PLD can be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. It can also be implemented using software. Figure 5 or Figure 6 The method shown, and its various modules, are also software modules, as are the data processing device 700 and its various modules.

[0221] The data processing apparatus 700 according to the embodiments of this application can correspondingly execute the methods described in the embodiments of this application, and the above and other operations and / or functions of each unit in the data processing apparatus 700 are respectively for implementing Figure 5 or Figure 6 For the sake of brevity, the corresponding processes of each method in the code will not be elaborated here.

[0222] Figure 8 This is a structural schematic diagram of a computer device 800 provided in this application. Figure 8As shown, the computer device 800 includes a processor 810, a bus 820, a memory 830, a communication interface 840, a main memory unit 850 (also called a main memory unit), and a processor 860. The processor 810, processor 860, memory 830, main memory unit 850, and communication interface 840 are connected via the bus 820.

[0223] It should be understood that in this embodiment, processor 810 is a CPU, but it can also be other general-purpose processors, digital signal processors (DSPs), ASICs, FPGAs, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor is a microprocessor or any conventional processor.

[0224] The computer device 800 also includes a graphics processing unit (GPU), a neural network processing unit (NPU), a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of programs according to the present application. For example, processor 860 is a GPU or an NPU.

[0225] The communication interface 840 is used to enable communication between the computer device 800 and external devices or components.

[0226] In this application, computer device 800 is used to implement Figure 3 or Figure 4 The traffic prediction system shown in the diagram uses a communication interface 840 to acquire map data, which enables the processor 810 to obtain the spatial characteristics of the target area based on the map data. The processor 860 is used to predict the traffic flow in the target area based on the spatial characteristics, and to store model parameters and optimizer parameters, etc.

[0227] Bus 820 includes a pathway for transferring information between the aforementioned components (such as processor 810, memory 850, and storage 830). In addition to the data bus, bus 820 also includes a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled as bus 820 in the diagram. Bus 820 can be a Peripheral Component Interconnect Express (PCIe) bus, or an Extended Industry Standard Architecture (EISA) bus, a unified bus (Ubus or UB), a compute express link (CXL), a cachecoherent interconnect for accelerators (CCIX), etc. Bus 820 is divided into address bus, data bus, and control bus.

[0228] As an example, computer device 800 includes multiple processors. A processor is a multi-core (multi-CPU) processor. Here, a processor refers to one or more devices, circuits, and / or computing units used to process data (e.g., computer program instructions).

[0229] It is worth noting that, Figure 8 Taking a computer device 800 comprising one processor 810 and one memory 830 as an example, the processor 810 and memory 830 are used to indicate a type of device or equipment. In a specific embodiment, the number of each type of device or equipment is determined according to business requirements. For example, the computer device 800 includes multiple GPUs or NPUs.

[0230] Memory 850 can be non-transitory memory, which can be a pool of volatile memory or a pool of non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory is random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). Memory 850 is used to store map data, traffic data, and traffic prediction results, etc.

[0231] The memory 830 corresponds to the storage medium used in the above method embodiments for storing information such as system map data, traffic data and traffic prediction results, for example, a disk, such as a mechanical hard disk or a solid-state hard disk.

[0232] The aforementioned computer device 800 may be a general-purpose device or a special-purpose device. For example, computer device 800 may also be a server or other device with computing capabilities.

[0233] It should be understood that the computer device 800 according to this embodiment may correspond to the data processing device 700 in this embodiment, and to the device executing the data processing device according to this embodiment. Figure 5 or Figure 6 The corresponding subject in any of the methods, and the above and other operations and / or functions of each module in the data processing apparatus 700 are respectively for implementing Figure 5 or Figure 6 For the sake of brevity, the corresponding processes of each method in the code will not be elaborated here.

[0234] The method steps in this embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Alternatively, the ASIC can reside in a computing device. Of course, the processor and storage medium can also exist as discrete components in the computing device.

[0235] Some embodiments of this application provide a computer-readable storage medium (e.g., a non-transitory computer-readable storage medium) storing computer program instructions that, when executed on a computer, cause the computer to perform one or more steps in the traffic prediction method as described in any of the above embodiments.

[0236] For example, the aforementioned computer-readable storage media include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks, or magnetic tapes), optical discs (e.g., CDs (Compact Disks), DVDs (Digital Versatile Disks), etc.), smart cards, and flash memory devices (e.g., EPROMs (Erasable Programmable Read-Only Memory), cards, sticks, or key drives, etc.). The various computer-readable storage media described in the embodiments of this application may represent one or more devices and / or other machine-readable storage media for storing information. The term "machine-readable storage medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing, and / or carrying instructions and / or data.

[0237] Some embodiments of this application also provide a computer program product. This computer program product includes computer program instructions carried on a non-transitory computer-readable storage medium, which, when executed on a computer, cause the computer to perform one or more steps of the data storage method as described in the above embodiments.

[0238] The beneficial effects of the computer-readable storage medium and computer program product described above are the same as those of the data storage method described in some of the above embodiments, and will not be repeated here.

[0239] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are performed entirely or partially. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, or other programmable device. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video disc (DVD); or it can be a semiconductor medium, such as a solid-state drive (SSD). The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A flow prediction method, characterized in that, include: Acquire map data of the target area, wherein the map data is used to indicate the objects and object features contained in the target area; The spatial features of the target area are obtained based on the map data, and the spatial features are used to indicate the distribution of objects in the target area. Predict the flow rate of the target area based on the spatial characteristics of the target area.

2. The method according to claim 1, characterized in that, Obtaining the spatial features of the target area based on the map data includes: Based on the map data, spatial features of multiple grid cells in the target area are obtained, and the spatial features of the grid cells are used to indicate the distribution of objects in the grid cells; Predicting the flow rate of the target area based on its spatial characteristics includes: The flow rate of the multiple grid cells is predicted based on their spatial characteristics, and the flow rate of the target area is obtained based on the flow rate of the multiple grid cells.

3. The method according to claim 1 or 2, characterized in that, The object characteristics include at least one of the following: object relationships, object shape, and object size.

4. The method according to any one of claims 1-3, characterized in that, The spatial features include at least one of the density or coverage of objects contained in the target area.

5. The method according to any one of claims 1-4, characterized in that, The multiple grid cells are all the same size.

6. The method according to any one of claims 1-5, characterized in that, Predicting the flow rate of the target area based on its spatial characteristics includes: Predict the flow rate of the target area during at least one time period based on the spatial characteristics of the target area.

7. The method according to any one of claims 1-6, characterized in that, Predicting the flow rate of the target area based on its spatial characteristics includes: The spatial characteristics of the target area are input into the traffic prediction model, and the traffic flow of the target area is output.

8. A model training method, characterized in that, include: Acquire map data and traffic data for the training area, wherein the map data is used to indicate the objects and object features contained in the training area; The spatial features of the training area are obtained based on the map data, and the spatial features are used to indicate the distribution of objects in the training area. Based on the traffic data, the traffic characteristics and time characteristics of the training area are obtained. The time characteristics are used to indicate the time distribution of traffic in the training area, and the traffic characteristics are used to indicate the traffic distribution of the training area. An artificial intelligence model is trained based on the spatial, temporal, and flow characteristics of the training area to obtain a flow prediction model, which is used to predict the flow in the target area.

9. The method according to claim 8, characterized in that, The spatial features of the training area are obtained based on the map data, including: The spatial features of multiple grid cells in the training area are obtained based on the map data. The spatial features of the grid cells are used to indicate the distribution of objects in the grid cells. The spatial features of the training area include the spatial features of the multiple grid cells.

10. The method according to claim 8 or 9, characterized in that, Based on the traffic data, the traffic characteristics and time characteristics of the training area are obtained, including: The time and flow characteristics of multiple grid cells in the training area are obtained based on the flow data. The time characteristics of the grid cells are used to indicate the time distribution of flow in the grid cells, and the flow characteristics of the grid cells are used to indicate the flow distribution in the grid cells. The flow characteristics of the training area include the flow characteristics of the multiple grid cells, and the time characteristics of the training area include the time characteristics of the multiple grid cells.

11. The method according to any one of claims 8-10, characterized in that, The traffic data includes at least one of location data and signal quality-related data.

12. The method according to any one of claims 8-11, characterized in that, The spatial features include at least one of the density or coverage of objects contained in the training region.

13. The method according to any one of claims 8-12, characterized in that, A traffic prediction model is obtained by training a model based on the spatial features, temporal features, and traffic flow features, including: The spatial and temporal features are input into the artificial intelligence model to obtain the predicted traffic flow of the training area. The artificial intelligence model is adjusted according to the error between the traffic flow features and the predicted traffic flow until the artificial intelligence model converges to obtain the traffic prediction model.

14. A data processing apparatus, characterized in that, The data processing apparatus includes modules for performing the operational steps of the method according to any one of claims 1-13.

15. A computer device, characterized in that, The computer device includes a memory and multiple processors, the memory being used to store a set of computer instructions; when the processors execute the set of computer instructions, the multiple processors perform the operational steps of the method according to any one of claims 1-13.

16. A flow prediction system, characterized in that, The traffic prediction system is used to perform the method as described in any one of claims 1-7 to predict traffic flow in a target area.

17. A model training system, characterized in that, The model training system is used to perform the method as described in any one of claims 8-13 to train an artificial intelligence model to obtain a traffic prediction model.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions, which, when executed by a computing device, implement the method as described in any one of claims 1-7 or the method as described in any one of claims 8-13.

19. A computer program product, characterized in that, The computer program product includes a computer program or instructions that, when executed by a computing device, implement the method as described in any one of claims 1-7 or the method as described in any one of claims 8-13.

Images