Cross-regional traffic prediction method and device based on artificial intelligence

Abstract

The invention relates to the technical field of traffic prediction, in particular to a cross-regional traffic prediction method and device based on artificial intelligence, and the method comprises the steps: collecting target traffic data meeting processing conditions of a plurality of cities, and constructing initial traffic models of the plurality of cities through the target traffic data; and migrating the spatial-temporal characteristics in the target traffic data to an original traffic data set of the target city by using the initial traffic model to obtain a target traffic data set of the target city, thereby optimizing the initial traffic prediction model of the target city, and obtaining a target MDTLGCN model based on the optimized initial traffic prediction model, and performing cross-regional traffic flow prediction on the target city by using the target MDTLGCN model. Therefore, the problems that in the prior art, when cross-city traffic flow prediction is processed, it is difficult to effectively fuse multi-source heterogeneous data, the cross-city space-time dependency relationship cannot be accurately captured, and the accuracy of traffic flow prediction is reduced are solved.

Description

Technical Field

[0001] The present application relates to the technical field of traffic prediction, and in particular to a method and device for cross-regional traffic prediction based on artificial intelligence. Background Art

[0002] With the acceleration of global urbanization, the complexity of urban traffic management has greatly increased. Traffic flow forecasting is one of the core technologies in traffic management, and its accuracy directly impacts traffic scheduling, flow control, and public transportation optimization. Typically, traffic data comes from sources such as traffic sensors, GPS signals, video surveillance, and traffic control systems. This data provides a crucial basis for traffic flow forecasting, but sensor failures, network transmission errors, or system issues often result in missing data.

[0003] Traditional traffic flow prediction methods rely heavily on complete datasets. If data is missing or incomplete, the accuracy of existing methods will drop significantly. Deep learning-based prediction methods, such as LSTM (Long Short-Term Memory Network) and GCN (Graph Convolutional Network), are capable of modeling spatiotemporal data, but their performance degrades significantly when data is severely missing. Therefore, related technologies utilize transfer learning techniques to transfer knowledge from the source domain to the target domain, thereby compensating for data loss in the target domain.

[0004] However, when dealing with cross-city traffic flow prediction, the transfer learning technology in related technologies has difficulty in effectively integrating multi-source heterogeneous data and cannot accurately capture the spatiotemporal dependencies across cities, which reduces the reliability and accuracy of traffic flow prediction and needs to be urgently addressed. Summary of the Invention

[0005] The present application provides an artificial intelligence-based cross-regional traffic prediction method and device to solve the problems in related technologies such as the difficulty of effectively integrating multi-source heterogeneous data and the inability to accurately capture the spatiotemporal dependencies across cities when processing cross-city traffic flow predictions using transfer learning technology, thereby reducing the applicability and accuracy of traffic flow predictions.

[0006] The first aspect of the present application provides an artificial intelligence-based cross-regional traffic prediction method, comprising the following steps: collecting target traffic data that meets preset processing conditions for multiple cities, and using the target traffic data to construct initial traffic models for the multiple cities; using the initial traffic model to migrate the spatiotemporal features in the target traffic data to the original traffic dataset of the target city to obtain the target traffic dataset of the target city; using the target traffic dataset to optimize the initial traffic prediction model of the target city, and obtaining a target MDTLGCN (Missing Data Transfer Learning Graph Convolutional Network) model of the target city based on the optimized initial traffic prediction model, so as to use the target MDTLGCN model to perform cross-regional traffic flow prediction for the target city.

[0007] Optionally, in one embodiment of the present application, the collecting of target traffic data of multiple cities that meet preset processing conditions includes: obtaining traffic flow characteristics and road network structure of the target city; determining the multiple cities that meet preset similarity conditions based on the traffic flow characteristics and the road network structure, and collecting target traffic data of the multiple cities that meet the preset processing conditions.

[0008] Optionally, in one embodiment of the present application, the use of the initial traffic model to migrate the spatiotemporal features in the target traffic data to the original traffic dataset of the target city to obtain the target traffic dataset of the target city includes: extracting the spatiotemporal features in the target traffic data of the multiple cities based on the initial traffic model; and using knowledge distillation technology to transfer the spatiotemporal features to the original traffic dataset of the target city to update the original traffic dataset of the target city and obtain the target traffic dataset of the target city.

[0009] Optionally, in one embodiment of the present application, the use of the target traffic dataset to optimize the initial traffic prediction model of the target city includes: determining a target multi-source dataset based on the target traffic data of the multiple cities and the original traffic dataset of the target city; and optimizing the initial traffic prediction model of the target city based on the target multi-source dataset and the target traffic dataset.

[0010] Optionally, in one embodiment of the present application, the target missing data transfer learning graph convolutional network MDTLGCN model of the target city is obtained based on the optimized initial traffic prediction model, including: based on the optimized initial traffic prediction model, converting the traffic flow data of the target city into a graph structure, and determining the spatial dependency relationship of each traffic node in the graph structure; based on the spatial dependency relationship of each traffic node and the time series data and spatial topological structure of the target traffic flow of the target city, determining the target missing data transfer learning graph convolutional network MDTLGCN model.

[0011] The second aspect of the present application provides an artificial intelligence-based cross-regional traffic prediction device, including: an acquisition module for collecting target traffic data that meets preset processing conditions from multiple cities, and using the target traffic data to construct initial traffic models for the multiple cities; a migration module for using the initial traffic model to migrate the spatiotemporal features in the target traffic data to the original traffic data set of the target city to obtain the target traffic data set of the target city; a prediction module for using the target traffic data set to optimize the initial traffic prediction model of the target city, and based on the optimized initial traffic prediction model, obtain a target missing data transfer learning graph convolutional network MDTLGCN model of the target city, so as to use the target MDTLGCN model to perform cross-regional traffic flow prediction for the target city.

[0012] Optionally, in one embodiment of the present application, the acquisition module includes: an acquisition unit, used to acquire the traffic flow characteristics and road network structure of the target city; a first determination unit, used to determine the multiple cities that meet the preset similarity conditions based on the traffic flow characteristics and the road network structure, and collect target traffic data of the multiple cities that meet the preset processing conditions.

[0013] Optionally, in one embodiment of the present application, the migration module includes: an extraction unit for extracting spatiotemporal features from the target traffic data of the multiple cities based on the initial traffic model; a transfer unit for transferring the spatiotemporal features to the original traffic dataset of the target city using knowledge distillation technology to update the original traffic dataset of the target city and obtain the target traffic dataset of the target city.

[0014] Optionally, in one embodiment of the present application, the prediction module includes: a second determination unit, used to determine a target multi-source data set based on the target traffic data of the multiple cities and the original traffic data set of the target city; an optimization unit, used to optimize the initial traffic prediction model of the target city based on the target multi-source data set and the target traffic data set.

[0015] Optionally, in one embodiment of the present application, the prediction module includes: a conversion unit for converting the traffic flow data of the target city into a graph structure based on the optimized initial traffic prediction model, and determining the spatial dependency relationship of each traffic node in the graph structure; a third determination unit for determining the target missing data transfer learning graph convolutional network MDTLGCN model based on the spatial dependency relationship of each traffic node and the time series data and spatial topological structure of the target traffic flow of the target city.

[0016] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the cross-regional traffic prediction method based on artificial intelligence as described in the above embodiment.

[0017] The fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program. When the program is executed by a processor, it implements the above-mentioned artificial intelligence-based cross-regional traffic prediction method.

[0018] The fifth aspect of the present application provides a computer program product, including a computer program, which, when executed, is used to implement the above-mentioned artificial intelligence-based cross-regional traffic prediction method.

[0019] The embodiment of the present application can construct initial traffic models for multiple cities based on target traffic data that meets processing conditions in multiple cities, thereby migrating the spatiotemporal features in the target traffic data to the original traffic data set of the target city to obtain the target traffic data set of the target city, and then optimizing the initial traffic prediction model of the target city to obtain the target MDTLGCN model, so as to use the target MDTLGCN model to predict cross-regional traffic flow for the target city, effectively improving the accuracy of traffic flow prediction. This solves the problem in the related art that when processing cross-city traffic flow prediction, the transfer learning technology is difficult to effectively integrate multi-source heterogeneous data and cannot accurately capture the spatiotemporal dependencies across cities, thereby reducing the accuracy of traffic flow prediction.

[0020] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0022] Figure 1A flowchart of an artificial intelligence-based cross-regional traffic prediction method provided according to an embodiment of the present application;

[0023] Figure 2 This is a schematic diagram of data interpolation prediction according to a specific embodiment of the present application;

[0024] Figure 3 This is a schematic diagram of a transfer learning framework based on knowledge distillation according to a specific embodiment of the present application;

[0025] Figure 4 This is a schematic diagram of the overall MDTLGCN model framework of a specific embodiment of the present application;

[0026] Figure 5 This is a traffic prediction model framework diagram based on a graph convolutional neural network according to a specific embodiment of the present application;

[0027] Figure 6 A schematic diagram of the loss curve of the MDTLGCN model training process according to a specific embodiment of the present application;

[0028] Figure 7 This is a schematic diagram comparing the RMSE (Root Mean Square Error) of the MDTLGCN model prediction method and the comparative method under different missing rates in a specific embodiment of the present application;

[0029] Figure 8 A schematic diagram of the structure of an artificial intelligence-based cross-regional traffic prediction device provided according to an embodiment of the present application;

[0030] Figure 9 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0032] The following describes the cross-regional traffic prediction method and device based on artificial intelligence in the embodiment of the present application with reference to the accompanying drawings. In view of the problem that the transfer learning technology in the related technology mentioned in the background technology center is difficult to effectively integrate multi-source heterogeneous data when processing cross-city traffic flow prediction, it is impossible to accurately capture the spatiotemporal dependency across cities, and the accuracy of traffic flow prediction is reduced, the present application provides a cross-regional traffic prediction method based on artificial intelligence, in which the initial traffic models of multiple cities can be constructed based on the target traffic data that meet the processing conditions of multiple cities, so that the spatiotemporal features in the target traffic data are migrated to the original traffic data set of the target city to obtain the target traffic data set of the target city, and then the initial traffic prediction model of the target city is optimized to obtain the target MDTLGCN model, so as to use the target MDTLGCN model to perform cross-regional traffic flow prediction on the target city, effectively improving the accuracy of traffic flow prediction. Thus, it solves the problem that the transfer learning technology in the related technology is difficult to effectively integrate multi-source heterogeneous data when processing cross-city traffic flow prediction, it is impossible to accurately capture the spatiotemporal dependency across cities, and the accuracy of traffic flow prediction is reduced.

[0033] Specifically, Figure 1 A flow chart of an artificial intelligence-based cross-regional traffic prediction method provided in an embodiment of the present application.

[0034] like Figure 1 As shown, the cross-regional traffic prediction method based on artificial intelligence includes the following steps:

[0035] In step S101 , target traffic data of multiple cities that meet preset processing conditions are collected, and initial traffic models of the multiple cities are constructed using the target traffic data.

[0036] In the embodiments of the present application, the preset processing conditions are conditions for performing standardized processing, which can be specifically set by those skilled in the art.

[0037] It can be understood that the embodiments of the present application can collect traffic data from multiple cities. For example, data from multiple cities with traffic flow characteristics similar to those of the target city are selected as source fields. Since traffic data may be missing to varying degrees, such as traffic flow data is usually affected by factors such as sensor failure, data loss or transmission errors, resulting in missing or incomplete traffic data, the collected traffic data can be standardized to obtain target traffic data, and then the target traffic data can be used to construct initial traffic models for multiple cities to capture the spatiotemporal characteristics of the source field, effectively improving the feasibility of cross-regional traffic prediction.

[0038] In one embodiment of the present application, target traffic data that meets preset processing conditions is collected from multiple cities, including: obtaining traffic flow characteristics and road network structure of the target city; determining multiple cities that meet preset similarity conditions based on the traffic flow characteristics and road network structure, and collecting target traffic data that meets the preset processing conditions from multiple cities.

[0039] In the embodiment of the present application, the target city is the city for which cross-regional traffic prediction is performed.

[0040] In the actual implementation process, the embodiment of the present application can obtain the traffic flow characteristics and road network structure of the target city, so as to determine multiple cities that meet similar conditions based on the traffic flow characteristics and road network structure of the target city, and then collect target traffic data that meet the processing conditions for multiple cities, for example, Figure 2 As shown, the interpolation method can be used to fill in some missing data in multiple cities to obtain the target traffic data of multiple cities, which effectively improves the integrity and reliability of the traffic data of multiple cities.

[0041] In step S102, the initial traffic model is used to migrate the spatiotemporal features in the target traffic data to the original traffic dataset of the target city to obtain the target traffic dataset of the target city.

[0042] It is understandable that the embodiments of the present application can use the initial traffic model to migrate the spatiotemporal features in the target traffic data to the original traffic data set of the target city. For example, through transfer learning technology, valid traffic data from multiple other cities can be migrated to the target city to update the original traffic data set of the target city, thereby obtaining the target traffic data set of the target city and further optimizing the traffic data prediction of the target city.

[0043] In one embodiment of the present application, the spatiotemporal features in the target traffic data are migrated to the original traffic dataset of the target city using the initial traffic model to obtain the target traffic dataset of the target city, including: extracting the spatiotemporal features in the target traffic data of multiple cities based on the initial traffic model; and transferring the spatiotemporal features to the original traffic dataset of the target city using knowledge distillation technology to update the original traffic dataset of the target city and obtain the target traffic dataset of the target city.

[0044] In the embodiment of the present application, transfer learning compensates for the lack of target city traffic data by transferring knowledge from the source domain to the target domain. Knowledge distillation is a method of transfer learning that extracts important knowledge from the source domain and transfers it to the target domain, thereby enhancing the performance of the target domain model.

[0045] In some embodiments, Figure 3 As shown, the embodiment of the present application can extract the spatiotemporal features in the target traffic data of multiple cities based on the initial traffic model, and use the knowledge distillation technology to transfer the spatiotemporal features to the original traffic data set of the target city. For example, in the transfer learning process, the embodiment of the present application can use the knowledge distillation technology to refine information, and obtain effective traffic flow data from the source domain (i.e., multiple other cities). By using knowledge distillation, efficient features and patterns can be extracted from the source domain and transferred to the target domain (i.e., the target city), so that the target domain can obtain knowledge support from the source domain even in the absence of data.

[0046] Among them, the knowledge distillation process transforms the model knowledge of the source domain into a learning signal that is effective for the target domain. Specifically, during the training process of the source domain, the intermediate layer features of the source domain model, that is, the initial traffic model, are extracted, and these features are transferred to the target domain through knowledge distillation technology to update the original traffic dataset of the target city and obtain the target traffic dataset of the target city, thereby accelerating the convergence of the target domain model and improving its prediction performance.

[0047] In addition, the embodiment of the present application adopts an adversarial training method, so that the source domain model and the target domain model, that is, the initial traffic model and the initial traffic prediction model, can adapt to each other during the training process, narrowing the distribution difference between the two. Through domain adaptation, the prediction ability of the target domain is not seriously affected by missing data.

[0048] In step S103, the target traffic data set is used to optimize the initial traffic prediction model of the target city, and the target missing data transfer learning graph convolutional network MDTLGCN model of the target city is obtained based on the optimized initial traffic prediction model, so as to use the target MDTLGCN model to perform cross-regional traffic flow prediction for the target city.

[0049] It is understandable that the embodiment of the present application can use the target traffic data set to optimize the initial traffic prediction model of the target city, for example, Figure 4 As shown, the embodiment of the present application can train the MDTLGCN model determined by optimizing the initial traffic prediction model of the target city. During the training process, by combining multi-source data and transfer learning strategies, it is ensured that the trained MDTLGCN model can quickly adapt to the data characteristics of the target domain with the support of the source domain data. That is to say, the present application can optimize the initial traffic prediction model of the target city to obtain the target MDTLGCN model, so that the target MDTLGCN model can maintain a high cross-regional traffic flow prediction accuracy for the target city under high missing rate conditions.

[0050] In one embodiment of the present application, the target traffic dataset is used to optimize the initial traffic prediction model of the target city, including: determining the target multi-source dataset based on the target traffic data of multiple cities and the original traffic dataset of the target city; and optimizing the initial traffic prediction model of the target city based on the target multi-source dataset and the target traffic dataset.

[0051] As a possible implementation method, the embodiments of the present application can integrate multi-source data sets. For example, the target traffic data of multiple cities and the original traffic data set of the target city are integrated to determine the multi-source data set. Then, the initial traffic prediction model of the target city is optimized through the multi-source data set and the target traffic data set, thereby effectively improving the accuracy and reliability of the optimization of the initial traffic prediction model.

[0052] Optionally, in one embodiment of the present application, a target missing data transfer learning graph convolutional network MDTLGCN model of the target city is obtained based on the optimized initial traffic prediction model, including: based on the optimized initial traffic prediction model, converting the traffic flow data of the target city into a graph structure, and determining the spatial dependency relationship of each traffic node in the graph structure; based on the spatial dependency relationship of each traffic node and the time series data and spatial topological structure of the target traffic flow of the target city, determining the target missing data transfer learning graph convolutional network MDTLGCN model.

[0053] In an embodiment of the present application, since traffic flow data has significant spatiotemporal dependencies, traditional deep learning methods, such as LSTM (Long Short-Term Memory networks) and CNN (Convolutional Neural Network), have limited modeling effects on spatiotemporal data. Therefore, an embodiment of the present application can construct the MDTLGCN model in the following steps, using GCN (Graph Convolutional Network) to capture the spatial dependencies between nodes through a graph structure, which can better process the spatiotemporal characteristics in traffic data.

[0054] Specifically, if Figure 5As shown, the embodiment of the present application first constructs a traffic network graph, converts traffic flow data into a graph structure, and each traffic node (such as an intersection, road section, etc.) is a node in the graph, and the connection relationship between nodes (such as road section connection) is used as an edge. Through the graph convolution operation, the spatial dependency relationship between traffic nodes is captured; then, spatiotemporal feature extraction is performed, and the target missing data transfer learning graph convolution network MDTLGCN model is determined by combining the time series data and spatial topological structure of traffic flow, so that spatiotemporal features can be extracted through GCN, and the future value of traffic flow can be predicted. Among them, GCN can capture the flow relationship between nodes in the spatial dimension, and at the same time process the dynamic changes of traffic flow in the time dimension.

[0055] During the actual implementation process, the embodiment of the present application can train the constructed MDTLGCN model. The training process uses traffic data from the source domain and the target domain, and is optimized in combination with the transfer learning strategy.

[0056] First, the training goal is determined. By minimizing the error of traffic flow prediction in the target city (target domain), the initial traffic prediction model of the target domain is optimized. During the training process, the optimized initial traffic prediction model needs to be able to effectively handle the missing data in the target city and make up for the missing part with the data from the source domain.

[0057] Secondly, multi-source data is integrated. During the training process, this application uses multi-source data, including traffic data of the target city and traffic data of multiple other cities. Through the integration of multi-source data, the MDTLGCN model can more robustly cope with changes in data from different cities.

[0058] Again, the MDTLGCN model is trained using a gradient descent-based optimization algorithm (e.g., Adam optimizer), and the network parameters are iteratively updated until the model converges. Figure 6 As shown in Figure 3, we use the target missing data transfer learning graph convolutional network (MDTLGCN) model to obtain the target city.

[0059] For example, the working principle of the embodiment of the present application is described in detail below using a specific embodiment, and the experimental steps are as follows:

[0060] 1. Dataset and Experimental Setup

[0061] Dataset: We use traffic datasets from two cities. These datasets include traffic speed information at different time periods (such as peak and off-peak hours). The dataset spans two months and covers traffic conditions under different weather conditions.

[0062] Missing rate setting: In order to simulate the situation of missing data, some data points are randomly marked as missing, and different missing rates (such as 20%, 40%, 60%, 80%) are set.

[0063] Experimental comparison model: The MDTLGCN model of this application is compared with a variety of advanced baseline methods, including the traditional KNN (K-Nearest Neighbors) method, STGCN (Spatio-Temporal Graph Convolutional Network) method, DastNet method and the integrated model TransGTR.

[0064] 2. Experimental Results

[0065] Missing rate comparison: Figure 7 As shown in the figure, the experimental results show that as the missing rate increases, the prediction error (RMSE as the evaluation indicator) of all methods increases. However, the MDTLGCN model still performs well under high missing rate conditions and maintains a low RMSE. In particular, when the missing rate reaches 80%, the RMSE of MDTLGCN is only 5.674, which is about 10% lower than the RMSE of the TransGTR model.

[0066] Model performance comparison: Compared with other baseline models (such as KNN, STGCN, and DastNet), MDTLGCN demonstrates the most stable performance under different missing rate conditions. In particular, under high missing rate conditions, MDTLGCN has the smallest prediction error, demonstrating its strong robustness in dealing with high missing data.

[0067] The effectiveness of knowledge distillation and transfer learning: This application also conducted comparative experiments to verify the role of knowledge distillation and transfer learning in MDTLGCN. The experiments showed that the performance of the MDTLGCN model decreased significantly after removing knowledge distillation and transfer learning, indicating that these two techniques played a crucial role in improving model accuracy.

[0068] 3. Results Analysis

[0069] Experimental results demonstrate that our proposed MDTLGCN model can maintain high prediction accuracy even under high data loss conditions. Specifically, compared to traditional methods, MDTLGCN can reduce the RMSE by 7%-13%, significantly improving the model's robustness and prediction accuracy, especially when data loss is severe.

[0070] 4. Application Scenarios and Expansion

[0071] In addition to urban traffic flow prediction, the model in this application can also be used for public transportation scheduling, traffic accident prediction, energy consumption prediction, etc. In these fields, data missing and incompleteness are common problems. The MDTLGCN model in this application, through the combination of cross-domain transfer learning and graph convolutional neural networks, can effectively solve these problems and provide accurate solutions for prediction tasks in related fields.

[0072] According to the cross-regional traffic prediction method based on artificial intelligence proposed in the embodiment of the present application, the initial traffic models of multiple cities can be constructed based on the target traffic data of multiple cities that meet the processing conditions, thereby migrating the spatiotemporal features in the target traffic data to the original traffic data set of the target city to obtain the target traffic data set of the target city, and then optimizing the initial traffic prediction model of the target city to obtain the target MDTLGCN model, so as to use the target MDTLGCN model to predict cross-regional traffic flow for the target city, thereby effectively improving the accuracy of traffic flow prediction. Thus, the problem that the transfer learning technology in the related art is difficult to effectively integrate multi-source heterogeneous data when processing cross-city traffic flow prediction, and cannot accurately capture the spatiotemporal dependency relationship between cities, thereby reducing the accuracy of traffic flow prediction, is solved.

[0073] Next, an artificial intelligence-based cross-regional traffic prediction device proposed according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0074] Figure 8 It is a block diagram of an artificial intelligence-based cross-regional traffic prediction device according to an embodiment of the present application.

[0075] like Figure 8 As shown, the artificial intelligence-based cross-regional traffic prediction device 10 includes: a collection module 100, a migration module 200 and a prediction module 300.

[0076] Specifically, the collection module 100 is used to collect target traffic data of multiple cities that meet preset processing conditions, and use the target traffic data to build initial traffic models for the multiple cities.

[0077] The migration module 200 is used to migrate the spatiotemporal features in the target traffic data to the original traffic dataset of the target city using the initial traffic model to obtain the target traffic dataset of the target city.

[0078] The prediction module 300 is used to optimize the initial traffic prediction model of the target city using the target traffic data set, and obtain the target missing data transfer learning graph convolutional network MDTLGCN model of the target city based on the optimized initial traffic prediction model, so as to use the target MDTLGCN model to predict cross-regional traffic flow for the target city.

[0079] Optionally, in one embodiment of the present application, the acquisition module 100 includes: an acquisition unit and a first determination unit.

[0080] The acquisition unit is used to obtain the traffic flow characteristics and road network structure of the target city.

[0081] The first determining unit is used to determine a plurality of cities that meet preset similarity conditions according to traffic flow characteristics and road network structure, and collect target traffic data that meet preset processing conditions from the plurality of cities.

[0082] Optionally, in one embodiment of the present application, the migration module 200 includes: an extraction unit and a transmission unit.

[0083] The extraction unit is used to extract the spatiotemporal features from target traffic data of multiple cities based on the initial traffic model.

[0084] The transfer unit is used to transfer the spatiotemporal features to the original traffic dataset of the target city by using the knowledge distillation technology, so as to update the original traffic dataset of the target city and obtain the target traffic dataset of the target city.

[0085] Optionally, in one embodiment of the present application, the prediction module 300 includes: a second determination unit and an optimization unit.

[0086] The second determining unit is configured to determine a target multi-source data set based on target traffic data of multiple cities and an original traffic data set of a target city.

[0087] The optimization unit is used to optimize the initial traffic prediction model of the target city based on the target multi-source dataset and the target traffic dataset.

[0088] Optionally, in one embodiment of the present application, the prediction module 300 includes: a conversion unit and a third determination unit.

[0089] Among them, the conversion unit is used to convert the traffic flow data of the target city into a graph structure based on the optimized initial traffic prediction model, and determine the spatial dependency relationship of each traffic node in the graph structure.

[0090] The third determination unit is used to determine the target missing data transfer learning graph convolutional network MDTLGCN model based on the spatial dependency relationship of each traffic node and the time series data and spatial topological structure of the target traffic flow of the target city.

[0091] It should be noted that the above explanation of the embodiment of the cross-regional traffic prediction method based on artificial intelligence is also applicable to the cross-regional traffic prediction device based on artificial intelligence in this embodiment, and will not be repeated here.

[0092] According to the cross-regional traffic prediction device based on artificial intelligence proposed in the embodiment of the present application, the initial traffic models of multiple cities can be constructed based on the target traffic data of multiple cities that meet the processing conditions, thereby migrating the spatiotemporal features in the target traffic data to the original traffic data set of the target city to obtain the target traffic data set of the target city, and then optimizing the initial traffic prediction model of the target city to obtain the target MDTLGCN model, so as to use the target MDTLGCN model to predict cross-regional traffic flow for the target city, effectively improving the accuracy of traffic flow prediction. Thus, the problem that the transfer learning technology in the related art is difficult to effectively integrate multi-source heterogeneous data when processing cross-city traffic flow prediction, and cannot accurately capture the spatiotemporal dependency relationship between cities, thereby reducing the accuracy of traffic flow prediction, is solved.

[0093] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0094] A memory 901 , a processor 902 , and a computer program stored in the memory 901 and executable on the processor 902 .

[0095] When the processor 902 executes the program, the cross-regional traffic prediction method based on artificial intelligence provided in the above embodiment is implemented.

[0096] Furthermore, the electronic device further includes:

[0097] The communication interface 903 is used for communication between the memory 901 and the processor 902 .

[0098] The memory 901 is used to store computer programs that can be run on the processor 902 .

[0099] The memory 901 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0100] If the memory 901, processor 902, and communication interface 903 are implemented independently, the communication interface 903, memory 901, and processor 902 can be interconnected via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0101] Optionally, in a specific implementation, if the memory 901, the processor 902 and the communication interface 903 are integrated on a chip, the memory 901, the processor 902 and the communication interface 903 can communicate with each other through an internal interface.

[0102] The processor 902 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0103] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned artificial intelligence-based cross-regional traffic prediction method.

[0104] This embodiment also provides a computer program product, including a computer program. When the computer program is executed, it is used to implement the above-mentioned cross-regional traffic prediction method based on artificial intelligence.

[0105] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0106] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0107] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing a custom logical function or process step, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

[0108] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or N wires (electronic devices), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program can be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing it in other suitable ways as necessary, and then storing it in a computer memory.

[0109] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, it can be implemented using any one or a combination of the following technologies known in the art: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0110] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0111] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0112] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A cross-regional traffic prediction method based on artificial intelligence, characterized in that: The following steps are involved: Collecting target traffic data that meets preset processing conditions for a plurality of cities, and constructing initial traffic models for the plurality of cities using the target traffic data; Using the initial traffic model, the spatiotemporal features in the target traffic data are transferred to the original traffic dataset of the target city to obtain the target traffic dataset of the target city; The target traffic dataset is used to optimize the initial traffic prediction model of the target city, and based on the optimized initial traffic prediction model, a target missing data transfer learning graph convolutional network MDTLGCN model of the target city is obtained, so as to use the target MDTLGCN model to perform cross-regional traffic flow prediction for the target city.

2. The method according to claim 1, characterized in that The target traffic data of multiple cities meeting the preset processing conditions are collected, including: Obtaining traffic flow characteristics and road network structure of the target city; The plurality of cities that meet the preset similarity conditions are determined according to the traffic flow characteristics and the road network structure, and target traffic data that meet the preset processing conditions are collected from the plurality of cities.

3. The method according to claim 1, characterized in that The step of migrating the spatiotemporal features of the target traffic data to the original traffic dataset of the target city using the initial traffic model to obtain the target traffic dataset of the target city includes: extracting spatiotemporal features from target traffic data of the plurality of cities based on the initial traffic model; The spatiotemporal features are transferred to the original traffic dataset of the target city by using knowledge distillation technology to update the original traffic dataset of the target city and obtain the target traffic dataset of the target city.

4. The method according to claim 1, wherein Optimizing the initial traffic prediction model of the target city by using the target traffic dataset includes: determining a target multi-source dataset based on the target traffic data of the plurality of cities and an original traffic dataset of the target city; An initial traffic prediction model for the target city is optimized based on the target multi-source dataset and the target traffic dataset.

5. The method according to claim 1, wherein The target missing data transfer learning graph convolutional network MDTLGCN model for the target city is obtained based on the optimized initial traffic prediction model, including: Based on the optimized initial traffic prediction model, the traffic flow data of the target city is converted into a graph structure, and the spatial dependency relationship of each traffic node in the graph structure is determined; Based on the spatial dependency relationship of each traffic node and the time series data and spatial topological structure of the target traffic flow of the target city, the target missing data transfer learning graph convolutional network MDTLGCN model is determined.

6. An artificial intelligence-based cross-regional traffic prediction device, characterized in that: include: A collection module, configured to collect target traffic data of multiple cities that meet preset processing conditions, and construct initial traffic models of the multiple cities using the target traffic data; a migration module, configured to migrate the spatiotemporal features in the target traffic data to an original traffic dataset of a target city using the initial traffic model, so as to obtain a target traffic dataset of the target city; A prediction module is used to optimize the initial traffic prediction model of the target city using the target traffic data set, and obtain the target missing data transfer learning graph convolutional network MDTLGCN model of the target city based on the optimized initial traffic prediction model, so as to use the target MDTLGCN model to predict cross-regional traffic flow for the target city.

7. The device according to claim 6, characterized in that The acquisition module includes: an acquisition unit, configured to acquire traffic flow characteristics and road network structure of the target city; A determination unit is used to determine the multiple cities that meet the preset similarity conditions based on the traffic flow characteristics and the road network structure, and collect target traffic data of the multiple cities that meet the preset processing conditions.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the cross-regional traffic prediction method based on artificial intelligence as described in any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the cross-regional traffic prediction method based on artificial intelligence as described in any one of claims 1 to 5.

10. A computer program product comprising a computer program, characterized in that The computer program is executed by a processor to implement the cross-regional traffic prediction method based on artificial intelligence as described in any one of claims 1 to 5.

Images