CN114220271A - Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network - Google Patents

Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network Download PDF

Info

Publication number
CN114220271A
CN114220271A CN202111575360.0A CN202111575360A CN114220271A CN 114220271 A CN114220271 A CN 114220271A CN 202111575360 A CN202111575360 A CN 202111575360A CN 114220271 A CN114220271 A CN 114220271A
Authority
CN
China
Prior art keywords
time
traffic flow
dynamic space
graph
dynamic
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202111575360.0A
Other languages
Chinese (zh)
Other versions
CN114220271B (en
Inventor
潘卫鹏
郭唐仪
唐坤
陆奕
蒋继扬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Science and Technology
Original Assignee
Nanjing University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN202111575360.0A priority Critical patent/CN114220271B/en
Publication of CN114220271A publication Critical patent/CN114220271A/en
Application granted granted Critical
Publication of CN114220271B publication Critical patent/CN114220271B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/12—Discovery or management of network topologies
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Abstract

The invention discloses a traffic flow prediction method based on a dynamic space-time graph convolution cycle network, which comprises the following steps: constructing a dynamic space-time relation graph of a road network; constructing a dynamic space-time graph convolution cyclic module; establishing a traffic flow prediction model based on a dynamic space-time diagram convolution cycle network; preprocessing traffic flow data and dividing a data set; training a traffic flow prediction model based on a dynamic space-time graph convolution cycle network; and carrying out traffic flow prediction and evaluating the prediction performance of the model based on the trained model. The invention can effectively capture the dynamic space-time dependence of traffic flow data, improve the accuracy of traffic flow prediction and provide scientific traffic guidance basis for traffic management departments. The method has extremely important significance for relieving urban traffic jam, improving running efficiency and safety and reducing traffic environment load. The method is simple and practical, and has strong operability and effectiveness.

Description

Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network
Technical Field
The invention belongs to the field of intelligent traffic, and particularly relates to a traffic flow prediction method, equipment and a storage medium based on a dynamic space-time graph convolution cycle network.
Background
Road traffic is an artery of social and economic activities, and plays an extremely important role in sustainable development of urban economy and improvement of the living standard of people. In recent years, with the continuous promotion of the urbanization process, a large number of people are rushed into urban work and life, the quantity of motor vehicles and the travel demand are increased day by day, and the problem of urban traffic jam is more and more serious. In order to relieve the traffic pressure of cities, various countries have been directing attention to Intelligent Transportation Systems (ITS). Traffic flow prediction is a hotspot of ITS research, and the prediction result can provide scientific traffic guidance basis for traffic management departments. The traffic system is a complex system, and the traffic flow change process is a real-time, nonlinear, high-dimensional and non-stable random process. The traditional traffic flow prediction method relies on data driving, only considers the time sequence characteristics of traffic flow data, ignores the spatial topology of a road network and splits the time correlation and the spatial correlation of the traffic flow. In recent years, a graph-convolution network has been applied to traffic flow prediction, and a complicated topological structure of a road network can be efficiently excavated. However, it only constructs a static graph and gives a fixed weight to represent the association between nodes, ignoring that the association between nodes is dynamically changed with time. Therefore, a traffic flow prediction method based on a dynamic space-time graph convolution cycle network is urgently needed at present, different dynamic space-time relation graphs are constructed along with time variation, dynamic space-time dependency of traffic flow is fully captured, accuracy of traffic flow prediction results is improved, reliable information support is provided for traffic management departments, and the method has positive significance for making corresponding traffic control measures.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides a traffic flow prediction method based on a dynamic space-time graph convolution cycle network.
The technical solution for realizing the purpose of the invention is as follows: a traffic flow prediction method based on a dynamic space-time graph convolution cycle network is characterized by comprising the following steps:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, predicting the traffic flow based on the trained model.
Preferably, step 1 constructs a dynamic spatiotemporal relationship graph of a road network, and the specific steps include:
step 1-1, constructing a static graph based on spatial distance:
according to geographical position information of traffic flow data sensors, constructing a weighted topological relation graph of a road network, wherein the sensors are used as nodes of the graph, and the connection relations of the sensors are used as edges of the graph;
constructing an adjacent matrix to describe the spatial relationship of the sensors, wherein the weight of the adjacent matrix is calculated by the spatial distance of the sensors in the road network, and specifically comprises the following steps:
Figure BDA0003424671310000021
in the formula, Aij DistRepresenting the connection weights of node i and node j in the adjacency matrix of the static graph,
Figure BDA0003424671310000022
spatial distance in the road network, σ, for node i and node j1Is the standard deviation of the distance between nodes, epsilon1Is a control threshold;
step 1-2, constructing a dynamic graph based on flow correlation:
calculating the correlation of the flow between the adjacent nodes by using the Pearson correlation coefficient, which specifically comprises the following steps:
suppose node ViHas a flow rate of
Figure BDA0003424671310000023
Neighbor node VjHas a flow rate of
Figure BDA0003424671310000024
T is the current time, T is the time sliding time window, then node ViAnd VjThe correlation between flows is expressed as:
Figure BDA0003424671310000025
correlation between nodes Sim (V)i,Vj) Comprises the following steps:
Figure BDA0003424671310000031
constructing an adjacency matrix to describe a dynamic space-time topological graph of the correlation between nodes, wherein the weight of the adjacency matrix is calculated by the correlation between the nodes, and specifically comprises the following steps:
Figure BDA0003424671310000032
in the formula, Aij SimThe connection weights of node i and node j in the adjacency matrix representing the dynamic graph,
Figure BDA0003424671310000033
spatial distance in the road network, σ, for node i and node j2For correlation of traffic between nodesStandard deviation of epsilon2Is a control threshold;
step 1-3, stacking a static graph based on space distance and a dynamic graph based on flow correlation, and constructing a dynamic spatiotemporal relationship graph of a road network:
Figure BDA0003424671310000034
in the formula, AijThe connection weights of the node i and the node j in the adjacency matrix representing the dynamic spatio-temporal relationship diagram,
Figure BDA0003424671310000035
spatial distance in the road network, σ, for node i and node j1Is the standard deviation, σ, of the distance between nodes2The standard deviation of the correlation of the flow among the nodes is shown, alpha and beta are weight coefficients, and epsilon is a control threshold value.
Preferably, the dynamic space-time graph convolution cyclic module comprises a dynamic space-time graph convolution layer and a cyclic layer;
the dynamic space-time graph convolutional layer takes the dynamic space-time relationship graph as the graph input of the network, and the dynamic space-time relationship graph is continuously updated along with the movement of the sliding time window; and inputting traffic flow data into a dynamic space-time diagram convolution layer, capturing the dynamic space dependency of the traffic flow, and outputting time series information with space characteristics, as shown in the following formula:
Figure BDA0003424671310000036
wherein X is an input traffic flow time sequence matrix, X' is an output new time sequence matrix with spatial characteristics,
Figure BDA0003424671310000037
representing a processing process, wherein W is a weight matrix, sigma represents a sigmoid activation function, and GCN ((-) represents graph convolution operation;
Figure BDA0003424671310000038
and is
Figure BDA0003424671310000039
INIs an identity matrix, a denotes an adjacency matrix,
Figure BDA00034246713100000310
is composed of
Figure BDA00034246713100000311
The elements of (1);
the time series information with space characteristics output by the dynamic space-time diagram convolution layer is input into the circulation layer, the dynamic time dependency of traffic flow is captured, and hidden layer information with space-time characteristics is output, and the formula is shown as follows:
X″＝LSTM(X′)
the specific binding process is as follows:
ft＝σ(Wf[ht-1，X′t]+bf)
it＝σ(Wi[ht-1，X′t]+bi)
Figure BDA0003424671310000041
Figure BDA0003424671310000042
ot＝σ(Wo[ht-1，X′t]+bo)
ht＝ot⊙tanh(Ct)
in the formula, LSTM (·) represents a loop layer operation, X' represents an input of a loop layer, and X ″ represents an output of the loop layer. Xt' represents input at time t, ht-1Hidden state at the previous moment, htIs an updated hidden state; f. oft、it、otThe display device comprises a forgetting gate, an input gate and an output gate;
Figure BDA0003424671310000043
Ct-1、Ctis the state of the memory cell; wf、Wi、WoAnd bf、bi、boIndicating a weight matrix and an offset entry in the training process, an indicates a corresponding multiplication of matrix elements.
Preferably, the step 3 of establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network specifically includes:
step 3-1, connecting a plurality of dynamic space-time graph convolution circulation modules through residual errors, integrating dynamic space-time graph convolution circulation layers, extracting space-time characteristics from shallow information and aggregating the space-time characteristics to generate deeper hidden layer information:
DSTGCRL(X)＝DSTGCRM(ReLU(…ReLU(DSTGCRM(X))))
in the formula, DSTGCRL (inverse discrete cosine transform) · represents a dynamic space-time diagram convolution loop layer operation, and ReLU (inverse discrete cosine transform) · represents an activation function;
3-2, sequentially connecting 2 dynamic space-time diagram convolution cycle layers, 1 attention mechanism layer and 1 full connection layer, and establishing a traffic flow prediction model based on a dynamic space-time diagram convolution cycle network;
the dynamic space-time graph convolution cycle layer is used for outputting hidden layer information with dynamic space-time characteristics, the attention mechanism layer distributes different weights for nodes of the hidden layer, and the full connection layer outputs results:
f(X,A)＝FC(LeakyReLU(Attention(DSTGCRL(ReLU(DSTGCRL(X))))))
in the formula, f (-) represents the model output, Attention (-) represents the Attention mechanism, FC (-) represents the fully-connected layer, and LeakyReLU (-) and ReLU (-) are activation functions.
Preferably, the traffic flow data preprocessing and data set partitioning in step 4 specifically includes:
step 4-1, selecting a real world traffic flow data set comprising geographical position information of a sensor and traffic flow data acquired by the sensor;
step 4-2, performing data cleaning on the traffic flow data set, and filling missing data by adopting a linear interpolation method;
step 4-3, carrying out standardization processing on the data set by adopting a Z-Score method;
and 4-4, dividing the data set into a training set, a testing set and a verification set by adopting a retention method.
Preferably, the specific method for training the traffic flow prediction model based on the dynamic space-time graph convolution cycle network comprises the following steps:
step 5-1, selecting Mean Square Error (MSE) as a loss function, selecting an Adam algorithm as a parameter optimizer, setting a model parameter learning rate, and setting the size of Batch;
and 5-2, randomly selecting data with the size of Batch from the training set, sending the data into a model for training, continuously updating model parameters until the training stopping condition is met, and obtaining a trained traffic flow prediction model.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, predicting the traffic flow based on the trained model.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, predicting the traffic flow based on the trained model.
Compared with the prior art, the invention has the following remarkable advantages:
1) the invention constructs a dynamic spatiotemporal relationship graph which changes along with time, and gets rid of the limitation of the existing static graph method; calculating the correlation between the nodes by using the Pearson correlation coefficient of the flow between the nodes, and continuously updating the correlation between the nodes along with the movement of a sliding time window; stacking a static graph based on distance and a dynamic graph based on traffic correlation among nodes to construct a spatiotemporal relation graph of a road network; the space-time relation graph of the road network is dynamically changed along with the input value, and the dynamic space-time dependency of the traffic flow can be effectively explained.
2) The invention constructs a dynamic space-time graph convolution cycle module, integrates a dynamic space-time relation graph, a graph convolution network and a long-term and short-term memory network, gets rid of the limitation that the traditional deep learning method only depends on data driving, gives consideration to the space characteristic and the time characteristic of the dynamic change of the traffic flow, and effectively captures the dynamic space-time dependency of the traffic flow.
3) The invention establishes a traffic flow prediction model based on a dynamic space-time graph convolution cycle network, data is firstly input into a dynamic space-time graph convolution cycle network layer, dynamic space-time characteristics of the traffic flow are extracted, attention mechanisms are introduced to distribute different weights, so that differentiated information aggregation is carried out, and finally, results are output by a full connection layer, so that the prediction performance of the model is improved.
The present invention is described in further detail below with reference to the attached drawing figures.
Drawings
Fig. 1 is a flow chart of a traffic flow prediction method based on a dynamic space-time graph convolution cycle network in one embodiment.
FIG. 2 is a flow chart of construction of a dynamic spatiotemporal relationship graph of a road network in one embodiment.
FIG. 3 is a block diagram of a dynamic space-time graph convolution cycle module in one embodiment.
FIG. 4 is a block diagram of a traffic flow prediction model based on a dynamic space-time graph convolutional loop network in one embodiment.
Fig. 5 is a schematic illustration of a PeMS04 data set traffic flow prediction result in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In addition, it should be noted that if there is a description of "first", "second", etc. in the embodiments of the present invention, the description of "first", "second", etc. is only for descriptive purposes and is not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In addition, technical solutions between various embodiments may be combined with each other, but must be realized by a person skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination should not be considered to exist, and is not within the protection scope of the present invention.
In one embodiment, in combination with fig. 1, there is provided a traffic flow prediction method based on a dynamic space-time graph convolutional loop network, the method including the steps of:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, carrying out traffic flow prediction based on the trained model and evaluating the prediction performance of the model.
Further, the step 1 of constructing a Dynamic Spatio-temporal Graph (DST-Graph) of the road network includes:
step 1-1, constructing a static graph based on spatial distance:
and constructing a weighted topological relation graph of the road network according to the geographical position information of the traffic flow data sensor. The sensors serve as nodes of the graph, and the connection relations of the sensors serve as edges of the graph. And constructing an adjacent matrix to describe the spatial relationship of the sensors, wherein the weight of the adjacent matrix is calculated by the spatial distance of the sensors in the road network. The weight of the adjacency matrix can be obtained by a threshold Gaussian kernel function:
Figure BDA0003424671310000071
in the formula, Aij DistRepresenting the connection weights of node i and node j in the adjacency matrix of the static graph,
Figure BDA0003424671310000072
spatial distance in the road network, σ, for node i and node j1Is the standard deviation of the distance between nodes, epsilon1Is a control threshold;
step 1-2, constructing a dynamic graph based on flow correlation:
if the interval of each time slice is short enough, at a certain moment, the traffic of the adjacent nodes can be considered to have linear correlation, that is, the traffic states of two adjacent nodes in the road network have similarity at the same time. The Correlation of the traffic between the adjacent nodes is calculated using the Pearson Correlation Coefficient (Pearson Correlation Coefficient).
Suppose node ViHas a flow rate of
Figure BDA0003424671310000073
Neighbor node VjOfMeasured as
Figure BDA0003424671310000074
T is the current time and T is the sliding time window. Then node ViAnd VjThe correlation between traffic can be expressed as:
Figure BDA0003424671310000081
due to r (X)i,Xj) Has a value of [ -1,1 [)]Namely, the traffic between the nodes has positive correlation and negative correlation. Only considering the correlation of adjacent nodes, not calculating when no edges are connected between the nodes, and obtaining the correlation Sim (V) between the nodes after taking the absolute valuei,Vj) I.e. by
Figure BDA0003424671310000082
And constructing a dynamic space-time topological graph for describing the correlation among the nodes by constructing an adjacency matrix, wherein the weight of the adjacency matrix is calculated by the correlation among the nodes. The weight of the adjacency matrix can be obtained by a threshold Gaussian kernel function:
Figure BDA0003424671310000083
in the formula, Aij SimThe connection weights of node i and node j in the adjacency matrix representing the dynamic graph,
Figure BDA0003424671310000084
spatial distance in the road network, σ, for node i and node j2Is the standard deviation, ε, of the correlation of traffic between nodes2Is a control threshold;
step 1-3, stacking a static graph based on space distance and a dynamic graph based on flow correlation, and constructing a dynamic spatiotemporal relationship graph of a road network:
Figure BDA0003424671310000085
in the formula, AijThe connection weights of the node i and the node j in the adjacency matrix representing the dynamic spatio-temporal relationship diagram,
Figure BDA0003424671310000087
spatial distance in the road network, σ, for node i and node j1Is the standard deviation, σ, of the distance between nodes2The standard deviation of the correlation of the flow among the nodes is shown, alpha and beta are weight coefficients, and epsilon is a control threshold value.
Inputting the flow of the road network nodes in a sliding time window T
Figure BDA0003424671310000086
Expressed as:
Figure BDA0003424671310000091
with the movement of the sliding time window T, the node flow input each time is changed continuously, and the correlation Sim (V) between the nodesi,Vj) And also constantly changing. Thus, A of the adjacency matrixijThe dynamic space-time relation graph of the road network is constructed by dynamically changing along with the input values. Because the graph construction method takes space distance and flow correlation into account, the dynamic space-time dependence of traffic flow can be explained.
Further, in one embodiment, with reference to fig. 3, the constructing a dynamic space-time graph convolution cycle module in step 2 specifically includes:
step 2-1, building a dynamic space-time diagram convolutional layer based on the dynamic space-time relationship diagram and the graph convolutional network:
and (3) sequentially inputting traffic flow data of different sliding time windows according to the sequence of the time steps, and constructing a dynamic spatiotemporal relationship diagram of the current time step according to the step 1. Based on a Graph Convolution Network (GCN) architecture of first-order Chebyshev approximation proposed by Kipf and Welling, a dynamic spatiotemporal relationship graph is used as graph input of the network to build a dynamic spatiotemporal graph convolution layer. The basic algorithm of GCN is as follows:
in atlas theory, a graph can be represented by its corresponding laplacian matrix, defined as L ═ D-a, normalized in the form of
Figure BDA0003424671310000092
Wherein A represents an adjacency matrix, INIs an identity matrix, a degree matrix D is a diagonal matrix, Dii＝∑jAij。
In graph signal processing, graph signals
Figure BDA0003424671310000093
Is the feature vector of the node on the graph. According to the graph convolution theory, the convolution operation of the input signal x and the convolution kernel g in the time domain can be converted into the inner product form of the frequency domain, namely:
Figure BDA0003424671310000094
in the formula (I), the compound is shown in the specification,
Figure BDA0003424671310000095
is Fourier transform,. is Hadamard productGIs a graph convolution operation.
Writing the Fourier transform of the convolution kernel g to a diagonal matrix, i.e. gθ＝diag(UTg) Then the graph convolution can be simplified to
x*Ggθ＝UgθUTx
Due to gθThe computational complexity of (Λ) is high, considering the approximate calculation with first order Chebyshev polynomials, i.e.
Figure BDA0003424671310000101
Then it is determined that,
Figure BDA0003424671310000102
further simplification, let beta0＝-β1θ, then xGgθ＝θ(IN+D-1/2AD-1/2)x
To avoid explosion or disappearance of the gradient, there is
Figure BDA0003424671310000103
Wherein
Figure BDA0003424671310000104
And is
Figure BDA0003424671310000105
Thus, the final formula for the graph convolution network is obtained:
Figure BDA0003424671310000106
with the movement of the sliding time window T, a dynamically changing road network space-time relation graph is generated, and the space-time relation graph in the above formula is continuously updated
Figure BDA0003424671310000109
Thereby building up a dynamic space-time diagram convolutional layer. And inputting traffic flow data into the dynamic space-time diagram convolution layer, capturing the dynamic space dependency of the traffic flow, and outputting time series information with space characteristics, namely:
Figure BDA0003424671310000107
wherein X is an input traffic flow time sequence matrix, X' is an output new time sequence matrix with spatial characteristics,
Figure BDA0003424671310000108
representing a processing process, wherein W is a weight matrix, sigma represents a sigmoid activation function, and GCN ((-) represents graph convolution operation;
step 2-2, building a circulation layer based on the long-term and short-term memory network:
a loop layer is built based on a Long short-term memory network (LSTM) architecture proposed by Hochreiter and Schmidhuber. LSTM replaces the hidden layer of the recurrent neural network with memory cells controlled by input gates, forget gates, and output gates. And (3) inputting the time sequence information with the space characteristics output in the step (2-1) into a circulation layer, capturing the dynamic time dependency of traffic flow, and outputting hidden layer information with space-time characteristics.
X″＝LSTM(X′)
The specific binding process is as follows:
ft＝σ(Wf[ht-1,X′t]+bf)
it＝σ(Wi[ht-1,X′t]+bi)
Figure BDA0003424671310000111
Figure BDA0003424671310000112
ot＝σ(Wo[ht-1,X′t]+bo)
ht＝ot⊙tanh(Ct)
in the formula, LSTM (·) represents a loop layer operation, X' represents an input of a loop layer, and X ″ represents an output of the loop layer. Xt' represents input at time t, ht-1Hidden state at the previous moment, htIs an updated hidden state; f. oft、it、otThe display device comprises a forgetting gate, an input gate and an output gate;
Figure BDA0003424671310000113
Ct-1、Ctis the state of the memory cell; wf、Wi、WoAnd bf、bi、boIndicating a weight matrix and an offset entry in the training process, an indicates a corresponding multiplication of matrix elements.
Therefore, the operation process of the Dynamic space-time Graph convolution loop module (DSTGCRM) is as follows:
DSTGCRM(X)＝X″＝LSTM(X′)＝LSTM(GCN(X))
in the formula, DSTGCRM (-) represents the operation of the dynamic space-time graph convolution cycle module.
Further, in one embodiment, with reference to fig. 4, the establishing a traffic flow prediction model based on a dynamic space-time graph convolution cycle network in step 3 specifically includes:
and 3-1, integrating Dynamic space-time Graph convolution loop Layers (Dynamic spatial-temporal Graph constraint Layers) by connecting a plurality of Dynamic space-time Graph convolution loop modules through residual errors (Residualbock). Network depth is increased by stacking a dynamic space-time graph convolution circulation module, space-time characteristics are extracted from shallow information to be aggregated, and deeper hidden layer information is generated.
DSTGCRL(X)＝DSTGCRM(ReLU(…ReLU(DSTGCRM(X))))
In the formula, DSTGCRL (-) represents the dynamic space-time diagram convolution loop layer operation, and ReLU (-) represents the activation function.
And 3-2, sequentially connecting 2 Dynamic space-time Graph convolution cycle Layers, 1 Attention Mechanism layer (Attention Mechanism layer) and 1 full-connected layer (full-connected Layers), and establishing a traffic flow prediction model based on a Dynamic space-time Graph convolution Network (DSTGCRN). The activation function of the convolution cycle layer of the dynamic space-time diagram is a ReLU function, and the activation function of the attention mechanism layer is a LeakyReLU function. Hidden layer information with dynamic space-time characteristics is output through a dynamic space-time graph convolution cycle layer, different weights are distributed to nodes of the hidden layer by introducing an attention mechanism, so that differentiated information aggregation is carried out, and finally a result is output by a full-connection layer.
f(X,A)＝FC(LeakyReLU(Attention(DSTGCRL(ReLU(DSTGCRL(X))))))
In the formula, f (-) represents the model output, Attention (-) represents the Attention mechanism, FC (-) represents the fully-connected layer, and LeakyReLU (-) and ReLU (-) are activation functions.
Further, in one embodiment, the traffic flow data preprocessing and data set partitioning in step 4 specifically includes:
step 4-1, selecting a real world traffic flow data set comprising geographical position information of a sensor and traffic flow data acquired by the sensor;
and 4-2, performing data cleaning on the traffic flow data set, and filling missing data by adopting a linear interpolation method.
Step 4-3, carrying out standardization processing on the data set by adopting a Z-Score method, wherein the mean value of the processed data is 0, and the standard deviation is 1, namely
Figure BDA0003424671310000121
In the formula, x is traffic flow data acquired by a sensor, mu is an average value of original data, sigma is a standard deviation of the original data, and x' is standardized data;
and 4-4, dividing the data set into a training set, a testing set and a verification set by using a retention method, wherein the principle of the retention method is similar to that of hierarchical sampling, dividing the data set according to the proportion for multiple times in order to ensure the randomness, and then averaging the results of the multiple divisions.
Further, in one embodiment, the training of the traffic flow prediction model based on the dynamic space-time graph convolutional loop network in step 5 specifically includes:
step 5-1, selecting Mean Square Error (MSE) as a loss function, selecting an Adam algorithm as a parameter optimizer, setting a model parameter learning rate, and setting the size of Batch;
and 5-2, randomly selecting data with the size of Batch from the training set, sending the data into a model for training, continuously updating model parameters until the training stopping condition is met, and obtaining a trained traffic flow prediction model.
6, carrying out traffic flow prediction based on the trained model, outputting a traffic flow prediction result, and selecting a baseline model to compare with a reference;
for the performance evaluation of the prediction model, it is essential to compare the actual values with the predicted values. And selecting three error functions of average absolute error, average absolute percentage error and root mean square error as indexes for evaluating the prediction effect of the model.
The modules in the traffic flow prediction method based on the dynamic space-time graph convolutional loop network can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, carrying out traffic flow prediction based on the trained model and evaluating the prediction performance of the model.
For specific definition of each step, reference may be made to the above definition of the traffic flow prediction method based on the dynamic space-time graph convolutional loop network, and details are not described here.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, carrying out traffic flow prediction based on the trained model and evaluating the prediction performance of the model.
For specific definition of each step, reference may be made to the above definition of the traffic flow prediction method based on the dynamic space-time graph convolutional loop network, and details are not described here.
As a specific example, in one embodiment, the traffic flow prediction method based on the dynamic space-time graph convolutional loop network is further verified and explained.
In this embodiment, a PeMS04 data set collected by the traffic bureau Performance Measurement System (PeMS) in california is taken as an example. PeMS04 was collected in the gulf of san francisco and contained 307 detectors at intervals ranging from 1/2018 to 28/2/5 min, for a total of 169992 samples. The PeMS04 includes the distance between sensors and traffic flow data, wherein the traffic flow data is data of flow, speed and density collected by the sensors, and the flow is selected for predictive analysis in this embodiment.
And constructing a dynamic space-time relation graph of the road network according to the distance between the sensors and the traffic flow data and by combining the Pearson correlation coefficient. And fusing the dynamic space-time relationship graph, the graph convolution network and the long-term and short-term memory network to construct a dynamic space-time graph convolution cycle module, thereby establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network.
The samples were divided into a training set, a test set, and a validation set at a ratio of about 7:2: 1. The size of the sliding time window is defined to be 7, the first 6 time sequence data are input values, and the last 1 time sequence data are labels, so that a training sample is formed, namely the traffic state of the next 5 minutes is predicted by the traffic data of the first 30 minutes. In trainingBefore, the data set is standardized by Z-Score, and the final result is obtained by inverse standardization during prediction. VAR, SVR, LSTM and STGCN are selected as baseline models and are compared and analyzed with the DSTGCRN model. A network model is built by utilizing a Pythrch deep neural network framework, mean square error MSE is set as a loss function, Adam algorithm is selected as a parameter optimizer, and the learning rate of model parameters is set to be 10-3And Batch size is 64. The experiments were performed on an Intel (R) Xeon (R) CPU E5-2660 v2 processor and NVIDIA GeForce GTX Titan X graphics card.
With reference to fig. 5, the performance evaluation indexes of the method of the present invention are shown in table 1:
TABLE 1 Performance indicators on datasets for different models
Figure BDA0003424671310000141
The prediction result obtained by the prediction process of the method can show that the method can realize traffic flow prediction of the PeMS04 data set. As shown in table 1, compared with the baseline model, the DSTGCRN model has significantly improved effects under the 3 performance indexes of MAE, MAPE and RMSE, which are 19.15%, 12.89% and 28.88, respectively. Referring to fig. 5, the predicted result of the DSTGCRN model in one day is closest to the real data group route, which shows that the DSTGCRN model obtains the best predicted performance through training.
In conclusion, the traffic flow prediction method based on the dynamic space-time graph convolution cycle network can effectively capture the dynamic space-time dependency of the traffic flow, and the model has high prediction precision and strong interpretability. The predicted result can provide reliable information support for traffic management departments, and has positive significance for making corresponding traffic control measures.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only illustrative of the present invention and are not intended to limit the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (8)

1. A traffic flow prediction method based on a dynamic space-time graph convolution cycle network is characterized by comprising the following steps:
step 1, constructing a dynamic spatio-temporal relationship graph of a road network;
step 2, constructing a dynamic space-time diagram convolution circulation module;
step 3, establishing a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
step 4, preprocessing traffic flow data and dividing a data set;
step 5, training a traffic flow prediction model based on the dynamic space-time graph convolution cycle network;
and 6, predicting the traffic flow based on the trained model.
2. The traffic flow prediction method based on the dynamic space-time graph convolution cycle network according to claim 1, wherein the step 1 of constructing the dynamic space-time relationship graph of the road network comprises the following specific steps:
step 1-1, constructing a static graph based on spatial distance:
according to geographical position information of traffic flow data sensors, constructing a weighted topological relation graph of a road network, wherein the sensors are used as nodes of the graph, and the connection relations of the sensors are used as edges of the graph;
constructing an adjacent matrix to describe the spatial relationship of the sensors, wherein the weight of the adjacent matrix is calculated by the spatial distance of the sensors in the road network, and specifically comprises the following steps:
Figure FDA0003424671300000011
in the formula, Aij DistRepresenting the connection weights of node i and node j in the adjacency matrix of the static graph,
Figure FDA0003424671300000012
spatial distance in the road network, σ, for node i and node j1Is the standard deviation of the distance between nodes, epsilon1Is a control threshold;
step 1-2, constructing a dynamic graph based on flow correlation:
calculating the correlation of the flow between the adjacent nodes by using the Pearson correlation coefficient, which specifically comprises the following steps:
suppose node ViHas a flow rate of
Figure FDA0003424671300000013
Neighbor node VjHas a flow rate of
Figure FDA0003424671300000014
T is the current time, T is the time sliding time window, then node ViAnd VjThe correlation between flows is expressed as:
Figure FDA0003424671300000021
correlation between nodes Sim (V)i,Vj) Comprises the following steps:
Figure FDA0003424671300000022
constructing an adjacency matrix to describe a dynamic space-time topological graph of the correlation between nodes, wherein the weight of the adjacency matrix is calculated by the correlation between the nodes, and specifically comprises the following steps:
Figure FDA0003424671300000023
in the formula, Aij SimThe connection weights of node i and node j in the adjacency matrix representing the dynamic graph,
Figure FDA0003424671300000024
spatial distance in the road network, σ, for node i and node j2Is the standard deviation, ε, of the correlation of traffic between nodes2Is a control threshold;
step 1-3, stacking a static graph based on space distance and a dynamic graph based on flow correlation, and constructing a dynamic spatiotemporal relationship graph of a road network:
Figure FDA0003424671300000025
in the formula, AijThe connection weights of the node i and the node j in the adjacency matrix representing the dynamic spatio-temporal relationship diagram,
Figure FDA0003424671300000026
spatial distance in the road network, σ, for node i and node j1Is the standard deviation, σ, of the distance between nodes2The standard deviation of the correlation of the flow among the nodes is shown, alpha and beta are weight coefficients, and epsilon is a control threshold value.
3. The traffic flow prediction method based on the dynamic space-time graph convolution cycle network according to claim 1, characterized in that the dynamic space-time graph convolution cycle module comprises a dynamic space-time graph convolution layer and a cycle layer;
the dynamic space-time graph convolutional layer takes the dynamic space-time relationship graph as the graph input of the network, and the dynamic space-time relationship graph is continuously updated along with the movement of the sliding time window; and inputting traffic flow data into a dynamic space-time diagram convolution layer, capturing the dynamic space dependency of the traffic flow, and outputting time series information with space characteristics, as shown in the following formula:
Figure FDA0003424671300000031
wherein X is an input traffic flow time sequence matrix, X' is an output new time sequence matrix with spatial characteristics,
Figure FDA0003424671300000032
representing a processing process, wherein W is a weight matrix, sigma represents a sigmoid activation function, and GCN ((-) represents graph convolution operation;
Figure FDA0003424671300000033
and is
Figure FDA0003424671300000034
INIs an identity matrix, a denotes an adjacency matrix,
Figure FDA0003424671300000035
is composed of
Figure FDA0003424671300000036
The elements of (1);
the time series information with space characteristics output by the dynamic space-time diagram convolution layer is input into the circulation layer, the dynamic time dependency of traffic flow is captured, and hidden layer information with space-time characteristics is output, and the formula is shown as follows:
X"＝LSTM(X′)
the specific binding process is as follows:
ft＝σ(Wf[ht-1,X′t]+bf)
it＝σ(Wi[ht-1,X′t]+bi)
Figure FDA0003424671300000037
Figure FDA0003424671300000038
ot＝σ(Wo[ht-1，X′t]+bo)
ht＝ot⊙tanh(Ct)
in the formula, LSTM (·) represents a loop layer operation, X' represents an input of a loop layer, and X ″ represents an output of the loop layer. Xt' represents input at time t, ht-1Hidden state at the previous moment, htIs an updated hidden state; f. oft、it、otThe display device comprises a forgetting gate, an input gate and an output gate;
Figure FDA0003424671300000039
Ct-1、Ctis the state of the memory cell; wf、Wi、WoAnd bf、bi、boIndicating a weight matrix and an offset entry in the training process, an indicates a corresponding multiplication of matrix elements.
4. The traffic flow prediction method based on the dynamic space-time graph convolution cycle network according to claim 1, wherein the step 3 of establishing the traffic flow prediction model based on the dynamic space-time graph convolution cycle network specifically comprises the following steps:
step 3-1, connecting a plurality of dynamic space-time graph convolution circulation modules through residual errors, integrating dynamic space-time graph convolution circulation layers, extracting space-time characteristics from shallow information and aggregating the space-time characteristics to generate deeper hidden layer information:
DSTGCRL(X)＝DSTGCRM(ReLU(…ReLU(DSTGCRM(X))))
in the formula, DSTGCRL (inverse discrete cosine transform) · represents a dynamic space-time diagram convolution loop layer operation, and ReLU (inverse discrete cosine transform) · represents an activation function;
3-2, sequentially connecting 2 dynamic space-time diagram convolution cycle layers, 1 attention mechanism layer and 1 full connection layer, and establishing a traffic flow prediction model based on a dynamic space-time diagram convolution cycle network;
the dynamic space-time graph convolution cycle layer is used for outputting hidden layer information with dynamic space-time characteristics, the attention mechanism layer distributes different weights for nodes of the hidden layer, and the full connection layer outputs results:
f(X,A)＝FC(LeakyReLU(Attention(DSTGCRL(ReLU(DSTGCRL(X))))))
in the formula, f (-) represents the model output, Attention (-) represents the Attention mechanism, FC (-) represents the fully-connected layer, and LeakyReLU (-) and ReLU (-) are activation functions.
5. The traffic flow prediction method based on the dynamic space-time graph convolution cycle network according to claim 1, wherein the traffic flow data preprocessing and the data set partitioning in step 4 specifically include:
step 4-1, selecting a real world traffic flow data set comprising geographical position information of a sensor and traffic flow data acquired by the sensor;
step 4-2, performing data cleaning on the traffic flow data set, and filling missing data by adopting a linear interpolation method;
step 4-3, carrying out standardization processing on the data set by adopting a Z-Score method;
and 4-4, dividing the data set into a training set, a testing set and a verification set by adopting a retention method.
6. The traffic flow prediction method based on the dynamic space-time graph convolution cycle network according to claim 1, characterized in that a specific method for training the traffic flow prediction model based on the dynamic space-time graph convolution cycle network is as follows:
step 5-1, selecting Mean Square Error (MSE) as a loss function, selecting an Adam algorithm as a parameter optimizer, setting a model parameter learning rate, and setting the size of Batch;
and 5-2, randomly selecting data with the size of Batch from the training set, sending the data into a model for training, continuously updating model parameters until the training stopping condition is met, and obtaining a trained traffic flow prediction model.
7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 6 are implemented when the computer program is executed by the processor.
8. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
CN202111575360.0A 2021-12-21 2021-12-21 Traffic flow prediction method, equipment and storage medium based on dynamic space-time diagram convolution circulation network Active CN114220271B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111575360.0A CN114220271B (en) 2021-12-21 2021-12-21 Traffic flow prediction method, equipment and storage medium based on dynamic space-time diagram convolution circulation network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111575360.0A CN114220271B (en) 2021-12-21 2021-12-21 Traffic flow prediction method, equipment and storage medium based on dynamic space-time diagram convolution circulation network

Publications (2)

Publication Number Publication Date
CN114220271A true CN114220271A (en) 2022-03-22
CN114220271B CN114220271B (en) 2023-06-30

Family

ID=80704891

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111575360.0A Active CN114220271B (en) 2021-12-21 2021-12-21 Traffic flow prediction method, equipment and storage medium based on dynamic space-time diagram convolution circulation network

Country Status (1)

Country Link
CN (1) CN114220271B (en)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114004152A (en) * 2021-10-29 2022-02-01 河海大学 Multi-wind-field wind speed space-time prediction method based on graph convolution and recurrent neural network
CN114664090A (en) * 2022-04-14 2022-06-24 山东大学 Traffic data filling method and system based on recurrent neural network
CN114694379A (en) * 2022-03-29 2022-07-01 中山大学 Traffic flow prediction method and system based on self-adaptive dynamic graph convolution
CN114724386A (en) * 2022-03-31 2022-07-08 温州大学 Short-time traffic flow prediction method and system under intelligent traffic and electronic equipment
CN114781609A (en) * 2022-04-19 2022-07-22 华东交通大学 Traffic flow prediction method based on multi-mode dynamic residual image convolution network
CN114973653A (en) * 2022-04-27 2022-08-30 中国计量大学 Traffic flow prediction method based on space-time graph convolution network
CN115205306A (en) * 2022-08-02 2022-10-18 吉林建筑大学 Medical image segmentation method based on graph convolution
CN115376317A (en) * 2022-08-22 2022-11-22 重庆邮电大学 Traffic flow prediction method based on dynamic graph convolution and time sequence convolution network
CN115482656A (en) * 2022-05-23 2022-12-16 汕头大学 Method for predicting traffic flow by using space dynamic graph convolution network
CN115512545A (en) * 2022-09-30 2022-12-23 浙江财经大学 Traffic speed prediction method based on spatio-temporal dynamic graph convolution network
CN115755219A (en) * 2022-10-18 2023-03-07 长江水利委员会水文局 Flood forecast error real-time correction method and system based on STGCN
CN116363878A (en) * 2023-05-26 2023-06-30 云南大学 Traffic flow prediction system and method based on continuous dynamic ordinary differential equation
CN116363874A (en) * 2023-03-20 2023-06-30 南京理工大学 Urban traffic hypergraph convolution prediction method integrating multimode high-order semantic correlation
CN114694379B (en) * 2022-03-29 2024-05-03 中山大学 Traffic flow prediction method and system based on self-adaptive dynamic graph convolution

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160314686A1 (en) * 2013-12-30 2016-10-27 Fudan University Method for traffic flow prediction based on spatio-temporal correlation mining
CN109461311A (en) * 2018-12-19 2019-03-12 哈尔滨工业大学（深圳） A kind of road network traffic flow spatio-temporal prediction method towards intelligent transportation and intelligent driving
CN110491129A (en) * 2019-09-24 2019-11-22 重庆城市管理职业学院 The traffic flow forecasting method of divergent convolution Recognition with Recurrent Neural Network based on space-time diagram
CN110827544A (en) * 2019-11-11 2020-02-21 重庆邮电大学 Short-term traffic flow control method based on graph convolution recurrent neural network
US20200135018A1 (en) * 2018-10-24 2020-04-30 Bluesignal Corporation Method of predicting traffic congestion and controlling traffic signals based on deep learning and server for performing the same
CN112216108A (en) * 2020-10-12 2021-01-12 中南大学 Traffic prediction method based on attribute-enhanced space-time graph convolution model
CN113487088A (en) * 2021-07-06 2021-10-08 哈尔滨工业大学（深圳） Traffic prediction method and device based on dynamic space-time diagram convolution attention model

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160314686A1 (en) * 2013-12-30 2016-10-27 Fudan University Method for traffic flow prediction based on spatio-temporal correlation mining
US20200135018A1 (en) * 2018-10-24 2020-04-30 Bluesignal Corporation Method of predicting traffic congestion and controlling traffic signals based on deep learning and server for performing the same
CN109461311A (en) * 2018-12-19 2019-03-12 哈尔滨工业大学（深圳） A kind of road network traffic flow spatio-temporal prediction method towards intelligent transportation and intelligent driving
CN110491129A (en) * 2019-09-24 2019-11-22 重庆城市管理职业学院 The traffic flow forecasting method of divergent convolution Recognition with Recurrent Neural Network based on space-time diagram
CN110827544A (en) * 2019-11-11 2020-02-21 重庆邮电大学 Short-term traffic flow control method based on graph convolution recurrent neural network
CN112216108A (en) * 2020-10-12 2021-01-12 中南大学 Traffic prediction method based on attribute-enhanced space-time graph convolution model
CN113487088A (en) * 2021-07-06 2021-10-08 哈尔滨工业大学（深圳） Traffic prediction method and device based on dynamic space-time diagram convolution attention model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
闫旭 等: "基于图卷积神经网络的城市交通态势预测算法", 《浙江大学学报（工学版）》 *

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114004152A (en) * 2021-10-29 2022-02-01 河海大学 Multi-wind-field wind speed space-time prediction method based on graph convolution and recurrent neural network
CN114694379A (en) * 2022-03-29 2022-07-01 中山大学 Traffic flow prediction method and system based on self-adaptive dynamic graph convolution
CN114694379B (en) * 2022-03-29 2024-05-03 中山大学 Traffic flow prediction method and system based on self-adaptive dynamic graph convolution
CN114724386A (en) * 2022-03-31 2022-07-08 温州大学 Short-time traffic flow prediction method and system under intelligent traffic and electronic equipment
CN114724386B (en) * 2022-03-31 2023-10-27 温州大学 Short-time traffic flow prediction method and system under intelligent traffic and electronic equipment
CN114664090A (en) * 2022-04-14 2022-06-24 山东大学 Traffic data filling method and system based on recurrent neural network
CN114781609A (en) * 2022-04-19 2022-07-22 华东交通大学 Traffic flow prediction method based on multi-mode dynamic residual image convolution network
CN114973653A (en) * 2022-04-27 2022-08-30 中国计量大学 Traffic flow prediction method based on space-time graph convolution network
CN114973653B (en) * 2022-04-27 2023-12-19 中国计量大学 Traffic flow prediction method based on space-time diagram convolutional network
CN115482656A (en) * 2022-05-23 2022-12-16 汕头大学 Method for predicting traffic flow by using space dynamic graph convolution network
CN115482656B (en) * 2022-05-23 2023-09-26 汕头大学 Traffic flow prediction method by using space dynamic graph convolutional network
CN115205306A (en) * 2022-08-02 2022-10-18 吉林建筑大学 Medical image segmentation method based on graph convolution
CN115376317B (en) * 2022-08-22 2023-08-11 重庆邮电大学 Traffic flow prediction method based on dynamic graph convolution and time sequence convolution network
CN115376317A (en) * 2022-08-22 2022-11-22 重庆邮电大学 Traffic flow prediction method based on dynamic graph convolution and time sequence convolution network
CN115512545B (en) * 2022-09-30 2023-06-27 浙江财经大学 Traffic speed prediction method based on space-time dynamic graph convolution network
CN115512545A (en) * 2022-09-30 2022-12-23 浙江财经大学 Traffic speed prediction method based on spatio-temporal dynamic graph convolution network
CN115755219A (en) * 2022-10-18 2023-03-07 长江水利委员会水文局 Flood forecast error real-time correction method and system based on STGCN
CN115755219B (en) * 2022-10-18 2024-04-02 长江水利委员会水文局 STGCN-based flood forecast error real-time correction method and system
CN116363874A (en) * 2023-03-20 2023-06-30 南京理工大学 Urban traffic hypergraph convolution prediction method integrating multimode high-order semantic correlation
CN116363874B (en) * 2023-03-20 2024-04-23 南京理工大学 Urban traffic hypergraph convolution prediction method integrating multimode high-order semantic correlation
CN116363878B (en) * 2023-05-26 2023-08-11 云南大学 Traffic flow prediction system and method based on continuous dynamic ordinary differential equation
CN116363878A (en) * 2023-05-26 2023-06-30 云南大学 Traffic flow prediction system and method based on continuous dynamic ordinary differential equation

Also Published As

Publication number Publication date
CN114220271B (en) 2023-06-30

Similar Documents

Publication Publication Date Title
CN114220271B (en) Traffic flow prediction method, equipment and storage medium based on dynamic space-time diagram convolution circulation network
Ma et al. Improving air quality prediction accuracy at larger temporal resolutions using deep learning and transfer learning techniques
Wang et al. Multiple convolutional neural networks for multivariate time series prediction
CN110675623B (en) Short-term traffic flow prediction method, system and device based on hybrid deep learning
CN110827544B (en) Short-term traffic flow control method based on graph convolution recurrent neural network
Liang et al. A novel wind speed prediction strategy based on Bi-LSTM, MOOFADA and transfer learning for centralized control centers
CN109492822B (en) Air pollutant concentration time-space domain correlation prediction method
Seo et al. Daily water level forecasting using wavelet decomposition and artificial intelligence techniques
CN111899510A (en) Intelligent traffic system flow short-term prediction method and system based on divergent convolution and GAT
CN110570035B (en) People flow prediction system for simultaneously modeling space-time dependency and daily flow dependency
CN113673769B (en) Traffic flow prediction method of graph neural network based on multivariate time sequence interpolation
CN113570859B (en) Traffic flow prediction method based on asynchronous space-time expansion graph convolution network
Lu et al. Short-term demand forecasting for online car-hailing using ConvLSTM networks
CN111047078B (en) Traffic characteristic prediction method, system and storage medium
Bi et al. Multi-indicator water quality prediction with attention-assisted bidirectional LSTM and encoder-decoder
CN112862177B (en) Urban area aggregation degree prediction method, device and medium based on deep neural network
CN113516304B (en) Regional pollutant space-time joint prediction method and device based on space-time diagram network
CN116307152A (en) Traffic prediction method for space-time interactive dynamic graph attention network
Yan et al. Small watershed stream-flow forecasting based on LSTM
CN112598165A (en) Private car data-based urban functional area transfer flow prediction method and device
CN116052254A (en) Visual continuous emotion recognition method based on extended Kalman filtering neural network
Li et al. Multi-task spatio-temporal augmented net for industry equipment remaining useful life prediction
CN111141879A (en) Deep learning air quality monitoring method, device and equipment
Utku et al. Deep learning based prediction model for the next purchase
Xu et al. Residual autoencoder-LSTM for city region vehicle emission pollution prediction

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant