CN-121998163-A - Traffic flow prediction method based on floating car data and mutual information topology reconstruction

CN121998163ACN 121998163 ACN121998163 ACN 121998163ACN-121998163-A

Abstract

The invention provides a traffic flow prediction method based on floating car data and mutual information topology reconstruction, which comprises the steps of obtaining non-full-sample floating car track data, generating traffic flow time sequences of road nodes, calculating mutual information values between any two road nodes based on the traffic flow time sequences of the road nodes, constructing an adjacent matrix according to the mutual information values, wherein elements of the adjacent matrix represent dependency relations among corresponding nodes, constructing a traffic network diagram according to the adjacent matrix, wherein the traffic network diagram comprises node sets, edge sets and adjacent matrixes, the node sets represent road nodes, the edge sets represent connectivity among the nodes, training a traffic flow prediction model by using a graph neural network model by taking the traffic network diagram and the traffic flow time sequences as inputs, predicting traffic flow in a future time period by using the trained traffic flow prediction model, and outputting a prediction result, so that the traffic flow prediction accuracy is effectively improved.

Inventors

LIU CHANG
LIU HONGYAN
WANG HONGMEI
LI JIANHUA
Wei Jitao

Assignees

北京易智时代数字科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251218

Claims (10)

1. The traffic flow prediction method based on the topology reconstruction of floating car data and mutual information is characterized by comprising the following steps: Acquiring non-full-sample floating vehicle track data, preprocessing the non-full-sample floating vehicle track data, and generating a traffic flow time sequence of each road node; calculating mutual information values between any two road nodes based on the traffic flow time sequence of each road node, and constructing an adjacency matrix according to the mutual information values, wherein elements of the adjacency matrix represent the dependency relationship between the corresponding nodes; Constructing a traffic network graph according to the adjacency matrix, wherein the traffic network graph comprises a node set, an edge set and the adjacency matrix, the node set represents road nodes, and the edge set represents connectivity among the nodes; Using the traffic network graph and the traffic flow time sequence as input, and training a traffic flow prediction model by using a graph neural network model; and predicting the traffic flow in the future time period by using the trained traffic flow prediction model and outputting a prediction result.
2. The traffic flow prediction method based on floating car data and mutual information topology reconstruction according to claim 1, wherein the calculating a mutual information value between any two road nodes comprises: for any two road nodes, respectively acquiring corresponding traffic flow time sequences; Estimating joint probability distribution and respective edge probability distribution based on the numerical distribution of the traffic flow time series; And inputting the joint probability distribution and the edge probability distribution into a mutual information calculation formula, and outputting a mutual information value between the two road nodes.
3. The traffic flow prediction method based on floating car data and mutual information topology reconstruction according to claim 2, wherein the mutual information calculation formula is: ; in the formula, And Which represents any two road nodes and, And Respectively represent And A corresponding time series of traffic flows is provided, The joint probability distribution is represented by a graph, And Respectively representing the respective corresponding edge probability distributions.
4. The traffic flow prediction method based on floating car data and mutual information topology reconstruction of claim 3, wherein said constructing an adjacency matrix from said mutual information values comprises: and constructing an adjacency matrix by taking the mutual information value as a matrix element.
5. The traffic flow prediction method based on floating car data and mutual information topology reconstruction of claim 4, further comprising, after said constructing the adjacency matrix: performing row normalization processing on the adjacent matrix, and processing elements of each row by adopting a softMax function so that the sum of the connection weights of each node and all other nodes is 1; And setting elements smaller than a preset threshold value in the adjacent matrix after the line normalization processing to zero so as to remove weak connection and obtain a sparse adjacent matrix.
6. The traffic flow prediction method based on floating car data and mutual information topology reconstruction of claim 5, wherein the adjacency matrix is dynamically updated, and the updating process comprises: Determining a sliding time window; sliding the time window every time new non-full-sample floating car track data is acquired; And (3) based on the data in the latest time window, recalculating to obtain a new mutual information value and a new adjacency matrix, and finishing updating the adjacency matrix.
7. The traffic flow prediction method based on floating car data and mutual information topology reconstruction according to any one of claims 1 to 6, wherein said obtaining non-full sample floating car trajectory data comprises: receiving an original floating car data stream from a navigation software service provider, a network taxi platform or a vehicle-mounted Telematics system, wherein the original floating car data stream comprises at least one of a car ID, longitude and latitude coordinates, an instantaneous speed and a time stamp; And analyzing the original floating car data stream, and carrying out standardized packaging according to a preset data format to form structured non-full-sample floating car track data.
8. The traffic flow prediction method based on floating car data and mutual information topology reconstruction of claim 7, wherein said preprocessing the non-full sample floating car trajectory data comprises: Identifying and removing space drift points and speed abnormal points in the non-whole sample floating car track data based on a geographical surrounding frame and a speed threshold value, and finishing data cleaning; matching the non-whole sample floating vehicle track data after the data cleaning to a target node or road section of a road network by using a map matching algorithm; And aggregating the track points matched with each node at fixed time intervals, and statistically generating a traffic flow time sequence of each node in a corresponding time slice.
9. The traffic flow prediction method based on floating car data and mutual information topology reconstruction according to claim 1, wherein the graph neural network model is a space-time graph convolutional network, and the training process comprises: Inputting the historical traffic flow time sequence and an adjacency matrix constructed by mutual information into the space-time diagram convolution network in a forward propagation mode, and calculating to obtain a predicted value of a future time period; and updating the network weight by using the average absolute error between the predicted value and the true value as a loss function through a back propagation algorithm until the model converges to obtain a traffic flow prediction model.
10. A traffic flow prediction system based on floating car data and mutual information topology reconstruction, comprising: The acquisition module is used for acquiring non-full-sample floating vehicle track data, preprocessing the non-full-sample floating vehicle track data and generating a traffic flow time sequence of each road node; The calculation module is used for calculating a mutual information value between any two road nodes based on the traffic flow time sequence of each road node and constructing an adjacent matrix according to the mutual information value, wherein elements of the adjacent matrix represent the dependency relationship between the corresponding nodes; the construction module is used for constructing a traffic network graph according to the adjacency matrix, wherein the traffic network graph comprises node sets, edge sets and the adjacency matrix, the node sets represent road nodes, and the edge sets represent connectivity among the nodes; the prediction module is used for taking the traffic network graph and the traffic flow time sequence as input, training a traffic flow prediction model by using the graph neural network model, and predicting the traffic flow of a future time period by using the trained traffic flow prediction model and outputting a prediction result.

Description

Traffic flow prediction method based on floating car data and mutual information topology reconstruction Technical Field The invention relates to the technical field of traffic flow prediction, in particular to a traffic flow prediction method based on floating car data and mutual information topology reconstruction. Background Traffic flow prediction methods based on Graph Neural Networks (GNNs) are current research hotspots, and the core of the methods is to construct an adjacency matrix which can accurately reflect the spatial dependency relationship between roads and serve as input of the GNNs. In the prior art, the adjacency matrix is typically constructed based on the linear correlation (e.g., pearson coefficients) of road physical connections, inter-node distances, or traffic parameter time series. However, the method has the obvious defects that dynamic and nonlinear traffic flow transmission characteristics cannot be captured by a method based on physical connection and distance when dealing with floating car data of non-full samples, and the calculation result is seriously distorted under the condition of sparse and missing non-full samples of data by a method based on linear correlation and severely depends on complete and continuous time sequence data. Therefore, how to accurately predict the traffic flow based on the floating car data of the non-whole samples becomes a technical problem to be solved by the skilled person. Disclosure of Invention The invention provides a traffic flow prediction method based on floating car data and mutual information topology reconstruction, which is used for solving the defect of poor accuracy of traffic flow prediction in the prior art when floating car data facing non-full samples are used. In a first aspect, the present invention provides a traffic flow prediction method based on topology reconstruction of floating car data and mutual information, comprising: Acquiring non-full-sample floating vehicle track data, preprocessing the non-full-sample floating vehicle track data, and generating a traffic flow time sequence of each road node; calculating mutual information values between any two road nodes based on the traffic flow time sequence of each road node, and constructing an adjacency matrix according to the mutual information values, wherein elements of the adjacency matrix represent the dependency relationship between the corresponding nodes; Constructing a traffic network graph according to the adjacency matrix, wherein the traffic network graph comprises a node set, an edge set and the adjacency matrix, the node set represents road nodes, and the edge set represents connectivity among the nodes; Using the traffic network graph and the traffic flow time sequence as input, and training a traffic flow prediction model by using a graph neural network model; and predicting the traffic flow in the future time period by using the trained traffic flow prediction model and outputting a prediction result. According to the traffic flow prediction method based on floating car data and mutual information topology reconstruction, the calculating of the mutual information value between any two road nodes comprises the following steps: for any two road nodes, respectively acquiring corresponding traffic flow time sequences; Estimating joint probability distribution and respective edge probability distribution based on the numerical distribution of the traffic flow time series; And inputting the joint probability distribution and the edge probability distribution into a mutual information calculation formula, and outputting a mutual information value between the two road nodes. According to the traffic flow prediction method based on floating car data and mutual information topology reconstruction, the mutual information calculation formula is as follows: ; in the formula, AndWhich represents any two road nodes and,AndRespectively representAndA corresponding time series of traffic flows is provided,The joint probability distribution is represented by a graph,AndRespectively representing the respective corresponding edge probability distributions. According to the traffic flow prediction method based on floating car data and mutual information topology reconstruction, the construction of the adjacency matrix according to the mutual information value comprises the following steps: and constructing an adjacency matrix by taking the mutual information value as a matrix element. According to the traffic flow prediction method based on the topology reconstruction of floating car data and mutual information, after the adjacency matrix is constructed, the traffic flow prediction method further comprises the following steps: performing row normalization processing on the adjacent matrix, and processing elements of each row by adopting a softMax function so that the sum of the connection weights of each node and all other nodes is 1; And setting elements smaller than a preset t