CN114694379A - Traffic flow prediction method and system based on self-adaptive dynamic graph convolution - Google Patents
Traffic flow prediction method and system based on self-adaptive dynamic graph convolution Download PDFInfo
- Publication number
- CN114694379A CN114694379A CN202210318281.XA CN202210318281A CN114694379A CN 114694379 A CN114694379 A CN 114694379A CN 202210318281 A CN202210318281 A CN 202210318281A CN 114694379 A CN114694379 A CN 114694379A
- Authority
- CN
- China
- Prior art keywords
- adaptive dynamic
- traffic
- self
- graph
- graph convolution
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000003068 static effect Effects 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000003044 adaptive effect Effects 0.000 claims description 43
- 239000011159 matrix material Substances 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000012512 characterization method Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 125000004122 cyclic group Chemical group 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000001308 synthesis method Methods 0.000 claims description 3
- 125000004432 carbon atom Chemical group C* 0.000 claims description 2
- 230000006870 function Effects 0.000 description 8
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- NTSBMKIZRSBFTA-AIDOXSFESA-N Digoxigenin bisdigitoxoside Chemical compound C1[C@H](O)[C@H](O)[C@@H](C)O[C@H]1O[C@@H]1[C@@H](C)O[C@@H](O[C@@H]2C[C@@H]3[C@]([C@@H]4[C@H]([C@]5(CC[C@@H]([C@@]5(C)[C@H](O)C4)C=4COC(=O)C=4)O)CC3)(C)CC2)C[C@@H]1O NTSBMKIZRSBFTA-AIDOXSFESA-N 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
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
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2474—Sequence data queries, e.g. querying versioned data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
- G06F16/288—Entity relationship models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Chemical & Material Sciences (AREA)
- Fuzzy Systems (AREA)
- Analytical Chemistry (AREA)
- Remote Sensing (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a traffic flow prediction method and a system based on self-adaptive dynamic graph convolution, wherein the method comprises the following steps: acquiring historical traffic data and preprocessing the historical traffic data; constructing a static adjacency graph; constructing a self-adaptive dynamic adjacency graph tensor; constructing a self-adaptive dynamic graph convolution prediction model; training a self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data; and inputting the traffic flow data of the data to be tested into the trained prediction model to obtain a prediction result. The system comprises: the device comprises a preprocessing module, a first building module, a second building module, a model building module, a training module and a prediction module. By using the method and the device, the accuracy of traffic flow prediction can be improved. The traffic flow prediction method and the traffic flow prediction system based on the self-adaptive dynamic graph convolution can be widely applied to the field of traffic prediction.
Description
Technical Field
The invention relates to the field of traffic prediction, in particular to a traffic flow prediction method and a traffic flow prediction system based on self-adaptive dynamic graph convolution.
Background
Traffic flow prediction aims to predict future traffic flow conditions (such as traffic speed, traffic volume, etc.) in a road network based on historical traffic observations. Accurate traffic prediction is an important basis for constructing an intelligent traffic system, and has important significance for various downstream applications such as traffic time estimation, route planning, traffic light control and the like. Since urban traffic networks are highly dynamic and have complex space-time dependencies, accurate traffic prediction remains a challenge.
Traditional statistical signal processing methods such as ARIMA models, Support Vector Regression (SVR) models model traffic predictions by single variable time signal regression. They rely on the assumption of signal stationarity and neglect the interrelationship between traffic nodes, making it difficult to capture real-world complex traffic patterns. With the development of deep learning technology, models such as convolutional neural networks emerge, but the models can only process spatial information in a European spatial rasterization mode and cannot process irregular traffic network topological relations.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a traffic flow prediction method and system based on adaptive dynamic graph convolution, which can improve the accuracy of traffic flow prediction.
The first technical scheme adopted by the invention is as follows: a traffic flow prediction method based on self-adaptive dynamic graph convolution comprises the following steps:
acquiring historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data;
acquiring the geographic space distance of the traffic node and constructing a static adjacency graph;
characterizing the traffic nodes and constructing a self-adaptive dynamic adjacency relation graph tensor;
constructing an adaptive dynamic graph convolution prediction model according to the static adjacency graph and the adaptive dynamic adjacency graph tensor;
training the self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model;
and inputting the traffic flow data of the data to be tested into the trained prediction model to obtain a prediction result.
Further, the step of obtaining historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data specifically includes:
setting a time step interval, the maximum historical observation time step number and the maximum prediction time step number;
dividing equal-length time periods in one day according to the time step intervals to obtain time period index sequences;
performing sliding window slicing on historical traffic data according to the time step interval, the maximum historical observation time step number and the maximum prediction time step number to obtain a traffic flow characteristic sequence;
and constructing a characteristic index pair according to the traffic flow characteristics and the time interval index sequence to obtain the preprocessed historical traffic data.
Further, the expression of the static adjacency graph is as follows:
in the above equation, the static adjacency graph,
representing a traffic node viAnd vjσ represents the standard deviation of the distance between nodes.
Further, the step of characterizing the traffic nodes and constructing the adaptive dynamic adjacency graph tensor specifically includes:
setting traffic nodes and representation dimensions of each time period, and constructing a traffic node representation matrix and a time period representation matrix;
calculating a tensor according to the traffic node characterization matrix and the time period characterization matrix based on a tensor synthesis method;
and carrying out nonlinear mapping and normalization processing on the tensor to obtain the self-adaptive dynamic adjacency graph tensor.
Further, the step of constructing the adaptive dynamic graph convolution prediction model according to the static adjacency graph and the adaptive dynamic adjacency graph tensor specifically includes:
obtaining an adaptive dynamic adjacency graph according to the tensor of the adaptive dynamic adjacency graph;
constructing an adaptive dynamic graph convolution module and performing graph convolution operation by adopting a static adjacency graph and a dynamic adjacency graph;
embedding the self-adaptive dynamic graph convolution module into a gated cyclic unit and replacing full-connection calculation to obtain the gated cyclic unit containing the self-adaptive dynamic graph convolution;
and constructing a model forming an encoder-decoder structure based on a gated cyclic unit containing the self-adaptive dynamic graph convolution to obtain a self-adaptive dynamic graph convolution prediction model.
Further, the step of training the adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model specifically includes:
based on a plan sampling mode, training a decoder in the adaptive dynamic graph convolution prediction model by taking the probability epsilon as input and the historical traffic flow characteristic true value as well as the probability 1-epsilon as input and the output estimation value of the previous time step to obtain the trained prediction model.
The second technical scheme adopted by the invention is as follows: a traffic flow prediction system based on adaptive dynamic graph convolution comprises:
the preprocessing module is used for acquiring historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data;
the first construction module is used for acquiring the geographic space distance of the traffic node and constructing a static adjacency graph;
the second construction module is used for representing the traffic nodes and constructing the self-adaptive dynamic adjacency relation graph tensor;
the model building module is used for building a self-adaptive dynamic graph convolution prediction model according to the static adjacency graph and the self-adaptive dynamic adjacency graph tensor;
the training module is used for training the self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model;
and the prediction module is used for inputting the traffic flow data of the data to be detected into the trained prediction model to obtain a prediction result.
The method and the system have the beneficial effects that: the invention adopts different self-adaptive adjacency graphs to carry out dynamic graph convolution on the traffic node representation at different time points, excavates a complex dynamic mode of a traffic network, improves the accuracy of traffic flow prediction, in addition, respectively sets a trainable node representation matrix and a time period representation matrix, and generates the node dynamic adjacency relation graphs at different time periods in a tensor synthesis mode, thereby avoiding respectively defining a node representation at each time period, and effectively reducing the parameter quantity of a prediction model when the number of the traffic nodes is huge.
Drawings
FIG. 1 is a flow chart illustrating the steps of a traffic flow prediction method based on adaptive dynamic graph convolution according to the present invention;
FIG. 2 is a block diagram of a traffic flow prediction system based on adaptive dynamic graph convolution according to the present invention;
FIG. 3 is a schematic diagram of an adaptive dynamic graph convolution prediction model in accordance with an embodiment of the present invention;
FIG. 4 is a graph comparing the predicted performance of the present invention method and a prior art typical graph convolution based traffic prediction method.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments. For the step numbers in the following embodiments, they are set for convenience of illustration only, the order between the steps is not limited at all, and the execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.
As shown in fig. 1, the present invention provides a traffic flow prediction method based on adaptive dynamic graph convolution, which includes the following steps:
s1, acquiring historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data;
s1.1, setting a time step interval, the maximum historical observation time step number and the maximum prediction time step number;
specifically, the time step interval Δ T is set to 5 minutes, the maximum historical observed time step number P is set to 12 (i.e., 1-hour duration), and the maximum predicted time step number Q is set to 12 (i.e., 1-hour duration);
s1.2, dividing one day into equal-length time intervals according to time step intervals to obtain a time interval index sequence;
specifically, the time of day is divided into 288 equal-length periods according to Δ T of 5 minutes, and the index L of each period in the day is 0,1, …,287 respectively.
S1.3, performing sliding window slicing on historical traffic data according to the time step interval, the maximum historical observation time step number and the maximum prediction time step number to obtain a traffic flow characteristic sequence;
specifically, the historical traffic flow feature data is subjected to sliding window slicing by setting the time step interval Δ T to be 5 minutes, setting the maximum historical observation time step number P to be 12 and setting the maximum predicted time step number Q to be 12, wherein each window slice is a traffic flow feature sequence X with the length of 24t-11,…,Xt,Xt+1,…,Xt+12Each moment of time
207 is the number of traffic nodes, 1 is the number of characteristics of each node (i.e. only one characteristic of traffic speed is adopted), and the indexes of the corresponding time periods in one day are respectively lt-11,…,lt,lt+1,…,lt+12；
S1.4, constructing a characteristic index pair according to the traffic flow characteristics and the time interval index sequence to obtain the preprocessed historical traffic data.
Specifically, the traffic flow characteristic sequence and the corresponding time period index sequence are combined to be a traffic characteristic-time period index pair sequence, and a preprocessed historical traffic flow data sample is obtained.
Is singleThe form of the sample is [ (X)t-11,lt-11),…,(Xt,lt),(Xt+1,lt+1),…,(Xt+12,lt+12)]Is 24.
S2, acquiring the geographic space distance of the traffic node and constructing a static adjacency graph;
specifically, the method adopts a Gaussian kernel function form to calculate the proximity of the traffic nodes to obtain a static adjacency relation graph, and the expression is as follows:
in the above formula, the first and second carbon atoms are,
representing a traffic node viAnd vjσ represents the standard deviation of the distance between nodes.
S3, characterizing the traffic nodes and constructing a self-adaptive dynamic adjacency graph tensor;
s3.1, setting traffic nodes and representation dimensions of each time period, and constructing a traffic node representation matrix and a time period representation matrix;
specifically, the dimension d of the traffic node representation and the dimension d of each time period representation in one day are set to be 30, and a source end traffic node representation matrix is constructed according to the dimension d
Terminal traffic node characterization matrix
Characterization matrix for each time period in a day
Nuclear tensor
Es,Et,EoC, initializing randomly;
s3.2, calculating tensors according to the traffic node representation matrix and the time period representation matrix based on a tensor synthesis method;
in particular, according to Es,Et,EoC, calculating the tensor by adopting a tensor synthesis mode
The calculation expression is as follows:
Ad＝C×1Et×2Es×3Ee
and S3.3, carrying out nonlinear mapping and normalization processing on the tensor to obtain the self-adaptive dynamic adjacency relation graph tensor.
Specifically, for AdCarrying out nonlinear mapping and normalization processing to obtain the final traffic node dynamic adjacency relation graph tensor
The calculation expression is as follows:
in the above formula, the nonlinear mapping adopts a LEAKYRELU function, and the softmax function normalizes the last dimension of the tensor;
s4, constructing an adaptive dynamic graph convolution prediction model according to the static adjacency graph and the adaptive dynamic adjacency graph tensor;
s4.1, obtaining a self-adaptive dynamic adjacency relation graph according to the self-adaptive dynamic adjacency relation graph tensor;
in particular, the node dynamic adjacency graph tensor
First slice in its first dimension
Means a traffic node dynamic adjacency graph in the ith time of the day.
S4.2, constructing an adaptive dynamic graph convolution module dgconv (·), and geometrically performing graph convolution operation by adopting a static adjacency graph and a dynamic adjacency graph, wherein the graph convolution order K is 2;
specifically, the calculation formula is:
in the above formula, Hin,HoutRespectively the input representation of the traffic node and the output representation of the self-adaptive graph convolution module, K is the graph convolution order, DfIs a static adjacent relation graph AsDegree matrix of DbIs AsThe degree matrix of the transposed matrix of (a),
and W is a dynamic adjacency graph corresponding to the first time interval in one day, and is a trainable weight matrix.
S4.3, embedding the self-adaptive dynamic graph convolution module into a gated circulation unit and replacing full-connection calculation to obtain the gated circulation unit containing the self-adaptive dynamic graph convolution;
specifically, dgconv (-) is embedded into a gated cyclic unit GRU to replace the full-connection calculation therein, so as to obtain a GRU unit containing the convolution of the adaptive dynamic graph; that is, for a GRU unit at time step t, the calculation is performed according to the following expression:
Ht＝ut⊙Ht-1+(1-ut)⊙ct,
in the above formula, Xt、HtInput traffic flow characteristics, output hidden states, H, respectively, for the current time step tt-1Is the hidden state of the previous time step; l. thetFor inputting the characteristics X of the in-sample and traffic flowtThe corresponding time of day period index,
is composed of
In the first dimension of the firsttSlicing; σ (-) represents a sigmoid function, | | represents a matrix splicing operation, and | | | represents a matrix Hadamard product operation.
And S4.4, constructing a model forming an encoder-decoder structure based on the gate control cycle unit containing the self-adaptive dynamic graph convolution to obtain a self-adaptive dynamic graph convolution prediction model.
Specifically, the GRU encoder length equal to P is 12, the GRU decoder length equal to Q is 12, the number of encoder-decoder structural layers is 2, the GRU unit including the adaptive dynamic graph convolution is used to form a GRU encoder-decoder prediction model, and a model structure diagram refers to fig. 3.
S5, training the self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model;
based on a plan sampling mode, a decoder in the adaptive dynamic graph convolution prediction model is trained by taking a true value of a historical traffic flow characteristic used by the probability epsilon as an input and an output estimation value of a previous time step used by the probability 1-epsilon as an input, so that a trained prediction model is obtained.
Specifically, when training the model, the first 12 "traffic characteristics-time interval index pairs" of the sequence in each sample are input into the encoder of the model, and the last 12 "traffic characteristics-time interval indexesRefer to "the decoder input into the model; adopting a criterion of minimum Mean Absolute Error (MAE); adopting an Adam optimizer; the learning rate starting value is 0.01, and decays at a rate of 0.1 during the 20 th, 30 th, 40 th and 50 th training sessions; planned sampling probability epsilon of GRU decoder in model at ith iteration in training processiCalculated from the following function:
wherein tau is 2000;
training is performed based on the model output and the true error. The model uses the GRU encoder-decoder structure: (1) during training, the input sequence is generated to correspond to two segments of the encoder and decoder. (2) The encoder inputs the input of the first segment for fixing, and the encoder inputs the hidden vector of the last step into the encoder as the initial hidden vector of the first step of the encoder, but the decoder has output without the output of the model. (3) During training, the decoder is selected by probability for each step input, either (1) the value corresponding to the current time step in the second segment of the input sequence, or the output prediction/estimation value of the previous time step of the decoder.
And S6, inputting traffic flow data of the data to be tested into the trained prediction model to obtain a prediction result.
The traffic flow prediction performance of the present invention is further described below with reference to fig. 4:
comparing the prediction performance of the method of the invention with that of the existing typical traffic prediction method based on graph convolution, the compared method comprises the following steps: DCRNN (diffusion convolutional recurrent neural network) model, STGCN (space-time Graph convolutional network) model, Graph-WaveNet model. The DCRNN model adopts a fixed static distance adjacency graph to carry out diffusion graph convolution and combines an encoder-decoder structure to carry out traffic flow prediction; the STGCN model adopts a fixed static distance adjacency graph to carry out graph convolution in a Chebyshev polynomial form and is combined with 1D-CNN time domain convolution to carry out traffic flow prediction; the Graph-WaveNet model adds a static self-adaptive adjacency Graph shared in a full time period to perform Graph convolution on the basis of an STGCN model, and performs traffic flow prediction by combining hole time domain convolution of various scales. FIG. 4 is a comparison graph of predicted performance showing the error of each method for traffic flow predictions of 15 minutes (3 steps), 30 minutes (6 steps), and 60 minutes (12 steps) into the future; where MAE represents mean absolute error, RMSE represents root mean square error, and MAPE represents mean absolute percent error. It can be seen that the invention obtains better overall traffic flow prediction accuracy by adopting different adaptive adjacency graphs at different time points for dynamic graph convolution, compared with a method based on a fixed static adjacency graph and an adaptive static adjacency graph.
As shown in fig. 2, a traffic flow prediction system based on adaptive dynamic graph convolution includes:
the preprocessing module is used for acquiring historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data;
the first construction module is used for acquiring the geographic space distance of the traffic node and constructing a static adjacency graph;
the second construction module is used for representing the traffic nodes and constructing the self-adaptive dynamic adjacency relation graph tensor;
the model building module is used for building a self-adaptive dynamic graph convolution prediction model according to the static adjacency graph and the self-adaptive dynamic adjacency graph tensor;
the training module is used for training the self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model;
and the prediction module is used for inputting the traffic flow data of the data to be detected into the trained prediction model to obtain a prediction result.
The contents in the above method embodiments are all applicable to the present system embodiment, the functions specifically implemented by the present system embodiment are the same as those in the above method embodiment, and the beneficial effects achieved by the present system embodiment are also the same as those achieved by the above method embodiment.
A traffic flow prediction device based on self-adaptive dynamic graph convolution comprises:
at least one processor;
at least one memory for storing at least one program;
when executed by the at least one processor, cause the at least one processor to implement an adaptive dynamic graph convolution-based traffic flow prediction method as described above.
The contents in the above method embodiments are all applicable to the present apparatus embodiment, the functions specifically implemented by the present apparatus embodiment are the same as those in the above method embodiments, and the advantageous effects achieved by the present apparatus embodiment are also the same as those achieved by the above method embodiments.
A storage medium having stored therein instructions executable by a processor, the storage medium comprising: the processor-executable instructions, when executed by the processor, are for implementing an adaptive dynamic graph convolution based traffic flow prediction method as described above.
The contents in the foregoing method embodiments are all applicable to this storage medium embodiment, the functions specifically implemented by this storage medium embodiment are the same as those in the foregoing method embodiments, and the beneficial effects achieved by this storage medium embodiment are also the same as those achieved by the foregoing method embodiments.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (7)
1. A traffic flow prediction method based on self-adaptive dynamic graph convolution is characterized by comprising the following steps:
acquiring historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data;
acquiring the geographic space distance of the traffic node and constructing a static adjacency graph;
characterizing the traffic nodes and constructing a self-adaptive dynamic adjacency relation graph tensor;
constructing an adaptive dynamic graph convolution prediction model according to the static adjacency graph and the adaptive dynamic adjacency graph tensor;
training the self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model;
and inputting the traffic flow data of the data to be tested into the trained prediction model to obtain a prediction result.
2. The traffic flow prediction method based on the adaptive dynamic graph convolution according to claim 1, wherein the step of obtaining historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data specifically comprises:
setting a time step interval, the maximum historical observation time step number and the maximum prediction time step number;
dividing equal-length time periods in one day according to the time step intervals to obtain time period index sequences;
performing sliding window slicing on historical traffic data according to the time step interval, the maximum historical observation time step number and the maximum prediction time step number to obtain a traffic flow characteristic sequence;
and constructing a characteristic index pair according to the traffic flow characteristics and the time interval index sequence to obtain the preprocessed historical traffic data.
3. The traffic flow prediction method based on the adaptive dynamic graph convolution is characterized in that the expression of the static adjacency graph is as follows:
in the above formula, the first and second carbon atoms are,
representing traffic nodes viAnd vjGround ofThe physical space distance, σ, represents the standard deviation of the distance between nodes.
4. The traffic flow prediction method based on the adaptive dynamic graph convolution according to claim 3, wherein the step of characterizing the traffic nodes and constructing the adaptive dynamic adjacency graph tensor specifically comprises:
setting traffic nodes and representation dimensions of each time period, and constructing a traffic node representation matrix and a time period representation matrix;
calculating a tensor according to the traffic node characterization matrix and the time period characterization matrix based on a tensor synthesis method;
and carrying out nonlinear mapping and normalization processing on the tensor to obtain the self-adaptive dynamic adjacency diagram tensor.
5. The traffic flow prediction method based on adaptive dynamic graph convolution according to claim 4, wherein the step of constructing the adaptive dynamic graph convolution prediction model according to the static adjacency graph and the adaptive dynamic adjacency graph tensor specifically includes:
obtaining an adaptive dynamic adjacency graph according to the tensor of the adaptive dynamic adjacency graph;
constructing an adaptive dynamic graph convolution module and performing graph convolution operation by adopting a static adjacency graph and a dynamic adjacency graph;
embedding the self-adaptive dynamic graph convolution module into a gate control circulation unit and replacing full-connection calculation to obtain a gate control circulation unit containing the self-adaptive dynamic graph convolution;
and constructing a model forming an encoder-decoder structure based on a gated cyclic unit containing the self-adaptive dynamic graph convolution to obtain a self-adaptive dynamic graph convolution prediction model.
6. The traffic flow prediction method based on adaptive dynamic graph convolution according to claim 5, wherein the step of training the adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model specifically comprises:
based on a plan sampling mode, training a decoder in the adaptive dynamic graph convolution prediction model by taking the probability epsilon as input and the historical traffic flow characteristic true value as well as the probability 1-epsilon as input and the output estimation value of the previous time step to obtain the trained prediction model.
7. A traffic flow prediction system based on adaptive dynamic graph convolution is characterized by comprising:
the preprocessing module is used for acquiring historical traffic data and preprocessing the historical traffic data to obtain preprocessed historical traffic data;
the first construction module is used for acquiring the geographic space distance of the traffic node and constructing a static adjacency graph;
the second construction module is used for representing the traffic nodes and constructing the self-adaptive dynamic adjacency relation graph tensor;
the model building module is used for building a self-adaptive dynamic graph convolution prediction model according to the static adjacency graph and the self-adaptive dynamic adjacency graph tensor;
the training module is used for training the self-adaptive dynamic graph convolution prediction model based on the preprocessed historical traffic data to obtain a trained prediction model;
and the prediction module is used for inputting the traffic flow data of the data to be detected into the trained prediction model to obtain a prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210318281.XA CN114694379B (en) | 2022-03-29 | 2022-03-29 | Traffic flow prediction method and system based on self-adaptive dynamic graph convolution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210318281.XA CN114694379B (en) | 2022-03-29 | 2022-03-29 | Traffic flow prediction method and system based on self-adaptive dynamic graph convolution |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114694379A true CN114694379A (en) | 2022-07-01 |
| CN114694379B CN114694379B (en) | 2024-05-03 |
Family
ID=82140419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210318281.XA Active CN114694379B (en) | 2022-03-29 | 2022-03-29 | Traffic flow prediction method and system based on self-adaptive dynamic graph convolution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114694379B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116245255A (en) * | 2023-03-30 | 2023-06-09 | 湖南大学 | Online space-time traffic flow prediction method |
| CN117058886A (en) * | 2023-10-12 | 2023-11-14 | 安徽宇疆科技有限公司 | Beidou space-time data model based on third-order tensor and traffic flow analysis method |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111710154A (en) * | 2020-05-15 | 2020-09-25 | 湖州师范学院 | Expressway traffic flow prediction method |
| CN111931905A (en) * | 2020-07-13 | 2020-11-13 | 江苏大学 | Graph convolution neural network model and vehicle track prediction method using same |
| CN112766551A (en) * | 2021-01-08 | 2021-05-07 | 鹏城实验室 | Traffic prediction method, intelligent terminal and computer readable storage medium |
| CN112801404A (en) * | 2021-02-14 | 2021-05-14 | 北京工业大学 | Traffic prediction method based on self-adaptive spatial self-attention-seeking convolution |
| CN112863180A (en) * | 2021-01-11 | 2021-05-28 | 腾讯大地通途（北京）科技有限公司 | Traffic speed prediction method, device, electronic equipment and computer readable medium |
| CN113362491A (en) * | 2021-05-31 | 2021-09-07 | 湖南大学 | Vehicle track prediction and driving behavior analysis method |
| 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 |
| CN114220271A (en) * | 2021-12-21 | 2022-03-22 | 南京理工大学 | Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network |
-
2022
- 2022-03-29 CN CN202210318281.XA patent/CN114694379B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111710154A (en) * | 2020-05-15 | 2020-09-25 | 湖州师范学院 | Expressway traffic flow prediction method |
| CN111931905A (en) * | 2020-07-13 | 2020-11-13 | 江苏大学 | Graph convolution neural network model and vehicle track prediction method using same |
| CN112766551A (en) * | 2021-01-08 | 2021-05-07 | 鹏城实验室 | Traffic prediction method, intelligent terminal and computer readable storage medium |
| CN112863180A (en) * | 2021-01-11 | 2021-05-28 | 腾讯大地通途（北京）科技有限公司 | Traffic speed prediction method, device, electronic equipment and computer readable medium |
| CN112801404A (en) * | 2021-02-14 | 2021-05-14 | 北京工业大学 | Traffic prediction method based on self-adaptive spatial self-attention-seeking convolution |
| CN113362491A (en) * | 2021-05-31 | 2021-09-07 | 湖南大学 | Vehicle track prediction and driving behavior analysis method |
| 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 |
| CN114220271A (en) * | 2021-12-21 | 2022-03-22 | 南京理工大学 | Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116245255A (en) * | 2023-03-30 | 2023-06-09 | 湖南大学 | Online space-time traffic flow prediction method |
| CN117058886A (en) * | 2023-10-12 | 2023-11-14 | 安徽宇疆科技有限公司 | Beidou space-time data model based on third-order tensor and traffic flow analysis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114694379B (en) | 2024-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Labach et al. | Survey of dropout methods for deep neural networks | |
| CN111191841B (en) | Power load prediction method and device, computer equipment and storage medium | |
| US20200272905A1 (en) | Artificial neural network compression via iterative hybrid reinforcement learning approach | |
| Dubois et al. | Data-driven predictions of the Lorenz system | |
| CN110619430A (en) | Space-time attention mechanism method for traffic prediction | |
| CN114694379B (en) | Traffic flow prediction method and system based on self-adaptive dynamic graph convolution | |
| CN114220271A (en) | Traffic flow prediction method, equipment and storage medium based on dynamic space-time graph convolution cycle network | |
| CN110138595A (en) | Time link prediction technique, device, equipment and the medium of dynamic weighting network | |
| CN112381673B (en) | Park electricity utilization information analysis method and device based on digital twin | |
| CN115511177A (en) | Ultra-short-term wind speed prediction method based on INGO-SWGMN hybrid model | |
| CN117175588B (en) | Space-time correlation-based electricity load prediction method and device | |
| CN113411216A (en) | Network flow prediction method based on discrete wavelet transform and FA-ELM | |
| CN116346639A (en) | Network traffic prediction method, system, medium, equipment and terminal | |
| CN112434888A (en) | PM2.5 prediction method of bidirectional long and short term memory network based on deep learning | |
| CN116166642A (en) | Spatio-temporal data filling method, system, equipment and medium based on guide information | |
| Lei et al. | A novel time-delay neural grey model and its applications | |
| CN115389743A (en) | Method, medium and system for predicting content interval of dissolved gas in transformer oil | |
| Xing et al. | Research of a novel combined deterministic and probabilistic forecasting system for air pollutant concentration | |
| CN117495071A (en) | Flow discovery method and system based on predictive log enhancement | |
| Shen et al. | Variational sequential optimal experimental design using reinforcement learning | |
| Flores et al. | Model for time series imputation based on average of historical vectors, fitting and smoothing | |
| CN116737943A (en) | News field-oriented time sequence knowledge graph link prediction method | |
| Cheikhi et al. | On the statistical benefits of temporal difference learning | |
| Caldana et al. | Neural ordinary differential equations for model order reduction of stiff systems | |
| CN118779582B (en) | Marine environment short-term prediction method and system based on condition countermeasure network |
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 |