CN112668797B - Long-short-period traffic prediction method - Google Patents
Long-short-period traffic prediction method Download PDFInfo
- Publication number
- CN112668797B CN112668797B CN202011641479.9A CN202011641479A CN112668797B CN 112668797 B CN112668797 B CN 112668797B CN 202011641479 A CN202011641479 A CN 202011641479A CN 112668797 B CN112668797 B CN 112668797B
- Authority
- CN
- China
- Prior art keywords
- traffic
- traffic prediction
- feature
- convolution
- prediction
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 29
- 238000010586 diagram Methods 0.000 claims abstract description 12
- 230000007774 longterm Effects 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 41
- 238000009792 diffusion process Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 12
- 230000003044 adaptive effect Effects 0.000 claims description 11
- 239000000284 extract Substances 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 9
- 238000012546 transfer Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 2
- 238000009825 accumulation Methods 0.000 abstract description 8
- 230000006870 function Effects 0.000 description 10
- 230000001364 causal effect Effects 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- 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
Landscapes
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a long-term and short-term traffic prediction method, which comprises the steps of obtaining first historical traffic data of nodes in a constructed traffic road network diagram, and carrying out convolution processing on the first historical traffic data through a first convolution layer in a preset traffic prediction model; traffic prediction is carried out on the convolved first historical traffic data through a first iteration RNN operator in the model, a traffic prediction result of a first time step is output, and the traffic prediction result of the first time step is input to a next iteration RNN operator for traffic prediction until the T-th traffic prediction is carried out p T of the iterative RNN operator output p Traffic prediction results of the time steps are obtained through a splicing module in the model p And the traffic prediction results of the time steps are spliced and then input into a second convolution layer for convolution processing, and the final traffic prediction result is output. The traffic prediction method solves the technical problems that the existing traffic prediction method has high prediction error accumulation and cannot simultaneously consider the long-term prediction precision and the short-term prediction precision.
Description
Technical Field
The application relates to the technical field of traffic prediction, in particular to a long-short-period traffic prediction method.
Background
Traffic prediction is a classical space-time prediction problem and is widely applied in actual life, such as smart city road network planning, intelligent travel path planning, urban public transportation systems and the like. Because of the high nonlinearity and complexity of traffic data, the prior art has mainly solved the traffic prediction problem through deep learning. The existing traffic prediction method has the problems that prediction errors are accumulated more highly, and long-term prediction precision cannot be considered at the same time.
Disclosure of Invention
The application provides a long-short-term traffic prediction method, which is used for solving the technical problems that the existing traffic prediction method has higher prediction error accumulation and can not simultaneously consider the long-short-term prediction precision.
In view of this, a first aspect of the present application provides a long-short-term traffic prediction method, including:
constructing a traffic road network into a graph structure to obtain a traffic road network graph;
acquiring first historical traffic data of nodes in the traffic road network graph;
inputting the first historical traffic data into a computer system comprising a first convolution layer, T p The first convolution layer carries out convolution processing on the first historical traffic data, the first iteration RNN operator carries out traffic prediction on the convolved first historical traffic data, a traffic prediction result of a first time step is output, and the traffic prediction result of the first time step is input to the next iteration RNN operator to carry out traffic prediction until the T-th p T of the iterative RNN operator output p Traffic prediction results of the time steps, the splicing module pairs T p And splicing the traffic prediction results of the time steps, and carrying out convolution processing on the spliced traffic prediction results by the second convolution layer to output final traffic prediction results.
Optionally, the iterative RNN operator includes a gated linear unit, a diffusion convolution layer, and a full join layer;
the first iterative RNN operator performs traffic prediction on the convolved first historical traffic data, and outputs a traffic prediction result of a first time step, including:
the gating linear unit extracts the time dependence of the convolved first historical traffic data and outputs a first feature;
the diffusion convolution layer extracts the space dependence relation of the first feature and outputs a second feature;
and the full-connection layer predicts the traffic of the second feature and outputs a traffic prediction result of the first time step.
Optionally, the gating linear unit extracts a time dependency relationship of the convolved first historical traffic data, and outputs a first feature, including:
the gating linear unit carries out one-dimensional convolution processing on the first historical traffic data after the convolution processing to obtain a first convolution characteristic and a second convolution characteristic,
the gating linear unit performs activation processing on the second convolution feature through an activation function, and calculates a Hadamard product of the first convolution feature and the activated second convolution feature;
and the gating linear unit carries out residual connection on the convolved first historical traffic data and the Hadamard product and outputs a first characteristic.
Optionally, the diffusion convolution layer extracts a spatial dependency relationship of the first feature, and outputs a second feature, including:
and the diffusion convolution layer performs diffusion convolution feature extraction with an adaptive matrix on the first feature and outputs a second feature.
Optionally, the second feature is:
wherein Z is a second feature, P f =a/rowsum (a) is the forward transfer matrix, a is the weight adjacency matrix; p (P) b ＝A T /rowsum(A T ) Is a backward transfer matrix; x is a first characteristic, W is a parameter matrix of the diffusion convolution layer, and K is a constant;
is an adaptive matrix, E 1 、E 2 Respectively embedding the source node and the target node.
Optionally, the method further comprises:
calculating the weight adjacency matrix based on the traffic road network graph, wherein the calculation formula of the weight in the weight adjacency matrix is as follows:
wherein ,aij Is the weight of the edge between the neighbor nodes i and j in the traffic road network diagram, d ij Is a neighbor node i,j, delta is a threshold parameter, epsilon is a super parameter.
Optionally, the inputting the first historical traffic data into a computer system comprising a first convolution layer, T p The iterative RNN operator, the splicing module and the preset traffic prediction model of the second convolution layer further include:
and preprocessing the first historical traffic data.
Optionally, the configuration process of the preset traffic prediction model is as follows:
acquiring second historical traffic data of nodes in the traffic road network graph;
and training a traffic prediction network through the second historical traffic data to obtain the preset traffic prediction model.
Optionally, the loss function of the traffic prediction network is:
wherein ,
for predicted T p Traffic data of individual time steps +.>
For future T p Real traffic data of individual time steps, W θ Training parameters for the traffic prediction network.
From the above technical scheme, the application has the following advantages:
the application provides a long-short-period traffic prediction method, which comprises the following steps: constructing a traffic road network into a graph structure to obtain a traffic road network graph; acquiring first historical traffic data of nodes in a traffic network diagram; inputting the first historical traffic data into a computer system comprising a first convolution layer, T p The iterative RNN operator, the splicing module and the preset traffic prediction model of the second convolution layer enable the first convolution layer to carry out convolution processing on the first historical traffic data, and the first iterationThe generation RNN operator carries out traffic prediction on the convolved first historical traffic data, outputs the traffic prediction result of the first time step, and inputs the traffic prediction result of the first time step to the next iteration RNN operator for traffic prediction until the T-th p T of the iterative RNN operator output p Traffic prediction results of each time step, and the splicing module pair T p And splicing the traffic prediction results of the time steps, carrying out convolution processing on the spliced traffic prediction results by the second convolution layer, and outputting the final traffic prediction results.
According to the traffic prediction method, the obtained first historical traffic data is input into a preset traffic prediction model, one time step of prediction is carried out through the iteration RNN operator, the traffic prediction result of each time is input into the next iteration RNN operator to carry out the prediction of the next time step, meanwhile, as a part of a final output result, one time step of prediction is carried out through each iteration RNN operator, effective long-period traffic space-time information can be extracted, the phenomenon of error accumulation is reduced, and the short-period traffic prediction precision is improved; the traffic prediction results of all time steps are spliced through the splicing module, and the final traffic prediction result is output through convolution processing of the spliced traffic prediction results, so that the long-short-term traffic prediction result is obtained, the long-short-term traffic prediction precision is considered, and the technical problems that the existing traffic prediction method is high in prediction error accumulation and cannot simultaneously consider the long-short-term prediction precision are solved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive faculty for a person skilled in the art.
Fig. 1 is a schematic flow chart of a long-short-period traffic prediction method according to an embodiment of the present application;
fig. 2 is a schematic diagram of a traffic road network sensor distribution according to an embodiment of the present application;
FIG. 3 is a schematic diagram of an RNN structure according to an embodiment of the present disclosure;
fig. 4 is a schematic structural diagram of a gating linear cell according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a preset traffic prediction network model according to an embodiment of the present application;
fig. 6 is a schematic flow chart of traffic prediction by the first iterative RNN operator according to the embodiment of the present application.
Detailed Description
The application provides a long-short-term traffic prediction method, which is used for solving the technical problems that the existing traffic prediction method has higher prediction error accumulation and can not simultaneously consider the long-short-term prediction precision.
In order to make the present application solution better understood by those skilled in the art, the following description will clearly and completely describe the technical solution in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
For ease of understanding, referring to fig. 1, an embodiment of a long-short-term traffic prediction method provided in the present application includes:
and 101, constructing a traffic road network into a graph structure to obtain a traffic road network graph.
In this embodiment of the present application, a traffic road network may be constructed as a directed graph structure, to obtain a traffic road network graph g= (V, E, a), where V is a node set, that is, a set of sensors on the traffic road network, and reference may be made to fig. 2, where each sensor may record historical traffic data, such as speed, traffic flow, and so on. E is the edge set among nodes, A E R N×N Is a weight adjacency matrix of the traffic road network graph G.
Further, the calculation formula of the weights in the weight adjacency matrix is:
wherein ,aij Is the weight of the edge between the neighbor nodes i and j in the traffic road network diagram, d ij For the distance between the neighboring nodes i and j, delta is a threshold parameter, which can be set to 0.1, epsilon is a super parameter, and a specific value can be set according to actual conditions. The weight adjacency matrix A is composed of the weights of the edges between all the neighbor nodes i and j in the traffic road network diagram.
After the traffic network graph G is obtained, the first historical traffic data may be obtained through the nodes therein.
X is traffic data of the ith node at the t-th time step t ∈R N Traffic data for all nodes at time step t. In the embodiment of the application, the T of the future is predicted by acquiring the historical tau traffic data p Traffic data for each time step. Thus, the first historical traffic data obtained in the embodiments of the present application may be expressed as:
X＝(x 1 ，x 2 ，…，x τ )∈R N×τ ；
traffic data that needs to be predicted in the future can be expressed as:
RNNs are a type of neural network that processes sequence data, and differ from other underlying neural networks that establish connections only from layer to layer in that RNNs also establish connections between neurons from layer to layer, since the latter data of the sequence data is also closely related to the former data, as shown in fig. 3.
Based on the RNN idea, the embodiment of the application takes a one-time iterative prediction process as an iterative RNN operator, the output prediction result of the one-time iterative prediction process is taken as a part of the final output result, and the single-step prediction result is input to the next iterative RNN operator at the same time, and the preset traffic prediction model in the embodiment of the application comprises T p Iterative RNN operators for predicting future T p Traffic data for each time step. The iterative RNN operator in the embodiment of the present application is different from a conventional RNN structure, each RNN operator in the conventional RNN structure shares parameters, and the iterative RNN operator in the embodiment of the present application includes a time and space module (i.e., a linear gating unit and a diffusion convolution layer) for one-time prediction, which are respectively used for extracting a time dependency relationship and a space dependency relationship, so as to complete single-step prediction, and the parameters of different iterative RNN operators are different.
The preset traffic prediction model in the embodiment of the application comprises a first convolution layer and T p The iterative RNN operator, splice module and second convolution layer can be referenced in fig. 5. Predicting future T by inputting the acquired first historical traffic data into a preset traffic prediction model p The traffic data of each time step, specifically, the first historical traffic data is convolved through a first convolution layer, and the convolved first historical traffic data is intersected through a first iteration RNN operatorOutputting a traffic prediction result of the first time step through prediction, and inputting the traffic prediction result of the first time step into a next iteration RNN operator for traffic prediction until the T-th p T of the iterative RNN operator output p Traffic prediction results of the time steps are compared with T through a splicing module p And splicing the traffic prediction results of the time steps, carrying out convolution processing on the spliced traffic prediction results through a second convolution layer, and outputting a final traffic prediction result.
Further, the first historical traffic data is input to a first convolution layer, T p The iterative RNN operator, the splicing module and the preset traffic prediction model of the second convolution layer further comprise: the first historical traffic data is preprocessed. Specifically, a linear difference method can be used for complementing the missing null value in the first historical traffic data, and a Z-Score method is used for removing outliers.
Further, the iterative RNN operator in embodiments of the present application includes a gated linear unit, a diffusion convolution layer, and a fully connected layer. Referring to fig. 6, the specific steps of outputting the traffic prediction result of the first time step by performing traffic prediction on the convolved first historical traffic data by the first iterative RNN operator include:
s1031, the gating linear unit extracts the time dependence of the convolved first historical traffic data and outputs a first feature.
In the embodiment of the application, a gating linear unit performs one-dimensional convolution processing on first historical traffic data subjected to convolution processing to obtain a first convolution feature and a second convolution feature; the gating linear unit performs activation processing on the second convolution characteristic through an activation function, and calculates a Hadamard product of the first convolution characteristic and the activated second convolution characteristic; and the gating linear unit performs residual connection on the convolved first historical traffic data and the Hadamard product and outputs a first characteristic.
The gating linear unit in the embodiment of the present application captures dynamic information of the first historical traffic data in the time dimension through a convolution structure, and reference may be made to fig. 4. Specifically, one-dimensional convolution processing is performed through one-dimensional causal convolution layerThe causal convolution layer has a convolution kernel width K t Under the action of convolution kernel, the time sequence length of input data after convolution processing is shortened by K t -1. For the input convolved first historical traffic data
C i For the dimension of the first historical traffic data after convolution processing, the size of the convolution kernel of the causal convolution layer in the embodiment of the application is
The gating linear unit carries out one-dimensional convolution processing on the convolved first historical traffic data, and the convolution result is +.>
Dividing the convolution result into a front part and a rear part to obtain a first convolution feature P and a second convolution feature Q, wherein the feature numbers of the first convolution feature P and the second convolution feature Q are consistent with the input feature numbers before convolution.
In the embodiment of the application, the second convolution feature Q is preferably activated by using a sigmoid activation function to gate the first convolution feature P, i.e. the Hadamard product P sigma (Q) of the first convolution feature and the activated second convolution feature is calculated, as indicated by the Hadamard product. And the gating linear unit further performs residual connection on the convolved first historical traffic data and the Hadamard product obtained through calculation, and outputs a first characteristic.
S1032, the diffusion convolution layer extracts the spatial dependency relation of the first feature and outputs the second feature.
The diffusion convolution layer in the embodiment of the application is a diffusion convolution layer with an adaptive matrix, and the diffusion convolution layer performs diffusion convolution feature extraction with the adaptive matrix on the first feature and outputs the second feature.
In given node structure information, K finite steps can be adopted to simulate the diffusion process of signals, node characteristics are extracted, and diffusion convolution characteristics obtained by diffusion convolution processing are as follows:
wherein ,Pk The transition matrix is calculated from the weight adjacency matrix a for the power series of the transition matrix. In the embodiment of the present application, since the traffic road network structure is a directed graph, the diffusion process has two directions, i.e. the forward transfer matrix P f =a/rowsum (a), backward transfer matrix P b ＝A T /rowsum(A T ) The diffusion convolution characteristic can thus be expressed as:
where X is the first feature of the input and W is the parameter matrix of the diffusion convolution layer.
In given node structure information, hidden space information can be obtained through an adaptive matrix, wherein the adaptive matrix is as follows:
in the adaptive matrix
Comprising two randomly initialized learnable parameters E 1 、E 2 ∈R N×c Specifically, E 1 Embedded for source node, E 2 Embedded for the target node, through E 1 、E 2 Multiplying to obtain a space dependent weight between a source node and a target node, eliminating weak connection of the weight through a ReLU activation function, normalizing the weight processed by the ReLU activation function through a softMax function, and enabling the self-adaptive matrix to be regarded as a transfer matrix of a hidden diffusion process.
In the embodiment of the application, explicit graph structure association is captured by adopting diffusion convolution, implicit graph structure association is captured by adopting an adaptive matrix, and spatial dependency relationship, namely graph structure information of a given node, is captured jointly by the two, and finally the second output characteristic is as follows:
s1033, the full-connection layer predicts the traffic of the second feature and outputs a traffic prediction result of the first time step.
The full connection layer predicts the traffic of the second input feature and outputs the traffic prediction result y of the first time step 1 。
The specific processing procedure of the other iterative RNN operators is similar to that of the first iterative RNN operator, and the difference is that the input data and the output data of different iterative RNN operators are different.
Further, the configuration process of the preset traffic prediction model in the embodiment of the application is as follows: acquiring second historical traffic data of nodes in the traffic network diagram; and training the traffic prediction network through the second historical traffic data to obtain a preset traffic prediction model.
After the second historical traffic data is obtained, the time interval may be set to 5 minutes, i.e. there will be 12 historical traffic data in one hour, and 288 historical traffic data in total in one day. The second historical traffic data can also be preprocessed, specifically, a linear interpolation method is used for complementing null values missing in the second historical traffic data, and a Z-Score method is used for removing outliers.
Wherein the second historical traffic data is longer in time period from the current time than the first historical traffic data. For example, a sensor in a certain traffic road network records historical traffic data of 62 days, and the historical traffic data of the first 50 days can be used as a training set to obtain second historical traffic data; the historical traffic data from the 51 st day to the 56 th day can be used as a verification set to obtain third historical traffic data; the historical traffic data of the last 6 days can be used as a test set to obtain first historical traffic data. Of course, those skilled in the art can flexibly divide according to the actual situation, and the specific limitation is not limited herein.
Training the traffic prediction network through the second historical traffic data until the preset iteration times are reached, stopping training to obtain a preset traffic prediction model, and performing model verification by adopting the third historical traffic data in the training process.
During training, the loss value can be calculated through the loss function, and parameters of each layer of the traffic prediction network are updated through back propagation of the loss value. The loss function of the traffic prediction network may be:
wherein ,
for predicted T p Traffic data of individual time steps +.>
For future T p Real traffic data of individual time steps, W θ Training parameters for the traffic prediction network.
The long-short-period traffic prediction method combines the advantages of iterative prediction and one-time prediction, and ensures the accuracy of long-short-period traffic prediction results; in addition, the preset traffic prediction model in the embodiment of the application can simultaneously give consideration to the accuracy of long-term and short-term prediction, and can reduce the accumulation of iterative prediction errors, because the prediction in each step can adjust the error of the prediction result according to the true value during training.
According to the traffic prediction method, the obtained first historical traffic data is input into the preset traffic prediction model, one time step of prediction is carried out through the iteration RNN operator, the traffic prediction result of each time is input into the next iteration RNN operator to carry out the prediction of the next time step, meanwhile, as a part of the final output result, one time step of prediction is carried out through each iteration RNN operator, effective long-period traffic space-time information can be extracted, the phenomenon of error accumulation is reduced, and the short-period traffic prediction precision is improved; the traffic prediction results of all time steps are spliced through the splicing module, and the final traffic prediction result is output through convolution processing of the spliced traffic prediction results, so that the long-short-term traffic prediction result is obtained, the long-short-term traffic prediction precision is considered, and the technical problems that the existing traffic prediction method is high in prediction error accumulation and cannot simultaneously consider the long-short-term prediction precision are solved.
In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions to execute all or part of the steps of the methods described in the embodiments of the present application by a computer device (which may be a personal computer, a server, or a network device, etc.). And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.
The above embodiments are merely for illustrating the technical solution of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (8)
1. A long-short-term traffic prediction method, comprising:
constructing a traffic road network into a graph structure to obtain a traffic road network graph;
acquiring first historical traffic data of nodes in the traffic road network graph;
inputting the first historical traffic data into a computer system comprising a first convolution layer, T p The first iteration RNN operator carries out traffic prediction on the convolved first historical traffic data, outputs a traffic prediction result of a first time step, and outputs the traffic prediction result of the first time stepTraffic prediction is carried out by entering the next iteration RNN operator until the T-th p T of the iterative RNN operator output p Traffic prediction results of the time steps, the splicing module pairs T p Splicing the traffic prediction results of the time steps, and carrying out convolution processing on the spliced traffic prediction results by the second convolution layer to output final traffic prediction results;
the iterative RNN operator comprises a gating linear unit, a diffusion convolution layer and a full connection layer;
the first iterative RNN operator performs traffic prediction on the convolved first historical traffic data, and outputs a traffic prediction result of a first time step, including:
the gating linear unit extracts the time dependence of the convolved first historical traffic data and outputs a first feature;
the diffusion convolution layer extracts the space dependence relation of the first feature and outputs a second feature;
and the full-connection layer predicts the traffic of the second feature and outputs a traffic prediction result of the first time step.
2. The long-term traffic prediction method according to claim 1, wherein the gating linear unit extracts a time dependency relationship of the convolved first historical traffic data, outputs a first feature, and includes:
the gating linear unit carries out one-dimensional convolution processing on the first historical traffic data after the convolution processing to obtain a first convolution characteristic and a second convolution characteristic,
the gating linear unit performs activation processing on the second convolution feature through an activation function, and calculates a Hadamard product of the first convolution feature and the activated second convolution feature;
and the gating linear unit carries out residual connection on the convolved first historical traffic data and the Hadamard product and outputs a first characteristic.
3. The long-term traffic prediction method according to claim 2, wherein the diffusion convolution layer extracts a spatial dependency relationship of the first feature, outputs a second feature, and includes:
and the diffusion convolution layer performs diffusion convolution feature extraction with an adaptive matrix on the first feature and outputs a second feature.
4. The long-term traffic prediction method according to claim 3, characterized in that the second feature is:
wherein Z is a second feature, P f =a/rowsum (a) is the forward transfer matrix, a is the weight adjacency matrix; p (P) b ＝A T /rowsum(A T ) For backward transfer matrix, A T The transpose matrix is the weighting adjacency matrix; x is a first characteristic, W is a parameter matrix of the diffusion convolution layer, and K is a constant;
is an adaptive matrix, E 1 、E 2 Respectively embedding source node and target node, W k1 Parameter matrix, W, of forward transfer matrix at kth finite step for diffusion convolution layer k2 Parameter matrix, W, of backward transfer matrix at kth finite step for diffusion convolution layer k3 The parameter matrix of the adaptive matrix at the kth finite step is the diffusion convolution layer.
5. The long-term traffic prediction method according to claim 4, characterized in that the method further comprises:
calculating the weight adjacency matrix based on the traffic road network graph, wherein the calculation formula of the weight in the weight adjacency matrix is as follows:
wherein ,aij Is the weight of the edge between the neighbor nodes i and j in the traffic road network diagram, d ij Delta is a threshold parameter and epsilon is a super parameter for the distance between the neighbor nodes i and j.
6. The long-term traffic prediction method according to claim 1, wherein the input of the first historical traffic data comprises a first convolution layer, T p The iterative RNN operator, the splicing module and the preset traffic prediction model of the second convolution layer further include:
and preprocessing the first historical traffic data.
7. The long-short-term traffic prediction method according to any one of claims 1 to 6, wherein the configuration process of the preset traffic prediction model is as follows:
acquiring second historical traffic data of nodes in the traffic road network graph;
and training a traffic prediction network through the second historical traffic data to obtain the preset traffic prediction model.
8. The long-term traffic prediction method according to claim 7, wherein the loss function of the traffic prediction network is:
wherein ,
for predicted T p Traffic data of individual time steps +.>
For future T p Real traffic data of individual time steps, W θ Training parameters for the traffic prediction network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011641479.9A CN112668797B (en) | 2020-12-31 | 2020-12-31 | Long-short-period traffic prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011641479.9A CN112668797B (en) | 2020-12-31 | 2020-12-31 | Long-short-period traffic prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112668797A CN112668797A (en) | 2021-04-16 |
| CN112668797B true CN112668797B (en) | 2023-06-16 |
Family
ID=75413867
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011641479.9A Active CN112668797B (en) | 2020-12-31 | 2020-12-31 | Long-short-period traffic prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112668797B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113505536A (en) * | 2021-07-09 | 2021-10-15 | 兰州理工大学 | Optimized traffic flow prediction model based on space-time diagram convolution network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020040411A1 (en) * | 2018-08-21 | 2020-02-27 | 한국과학기술정보연구원 | Device for predicting traffic state information, method for predicting traffic state information, and storage medium for storing program for predicting traffic state information |
| CN111899510A (en) * | 2020-07-28 | 2020-11-06 | 南京工程学院 | Intelligent traffic system flow short-term prediction method and system based on divergent convolution and GAT |
| CN111968375A (en) * | 2020-08-27 | 2020-11-20 | 北京嘀嘀无限科技发展有限公司 | Traffic flow prediction method and device, readable storage medium and electronic equipment |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109410575B (en) * | 2018-10-29 | 2020-05-01 | 北京航空航天大学 | Road network state prediction method based on capsule network and nested long-time memory neural network |
| CN110674987A (en) * | 2019-09-23 | 2020-01-10 | 北京顺智信科技有限公司 | Traffic flow prediction system and method and model training method |
| 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 |
| CN110717627B (en) * | 2019-09-29 | 2022-01-25 | 浙江大学 | Full traffic prediction method based on dual graph framework |
| CN111292562B (en) * | 2020-05-12 | 2020-08-18 | 北京航空航天大学 | Aviation flow prediction method |
| CN112071065A (en) * | 2020-09-16 | 2020-12-11 | 山东理工大学 | Traffic flow prediction method based on global diffusion convolution residual error network |
-
2020
- 2020-12-31 CN CN202011641479.9A patent/CN112668797B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020040411A1 (en) * | 2018-08-21 | 2020-02-27 | 한국과학기술정보연구원 | Device for predicting traffic state information, method for predicting traffic state information, and storage medium for storing program for predicting traffic state information |
| CN111899510A (en) * | 2020-07-28 | 2020-11-06 | 南京工程学院 | Intelligent traffic system flow short-term prediction method and system based on divergent convolution and GAT |
| CN111968375A (en) * | 2020-08-27 | 2020-11-20 | 北京嘀嘀无限科技发展有限公司 | Traffic flow prediction method and device, readable storage medium and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 基于图卷积神经网络的城市交通态势预测算法;闫旭;范晓亮;郑传潘;臧彧;王程;程明;陈龙彪;;浙江大学学报(工学版)(第06期);第104-112页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112668797A (en) | 2021-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111047078B (en) | Traffic characteristic prediction method, system and storage medium | |
| CN112910711B (en) | Wireless service flow prediction method, device and medium based on self-attention convolutional network | |
| CN113591380B (en) | Traffic flow prediction method, medium and equipment based on graph Gaussian process | |
| CN112085161B (en) | Graph neural network method based on random information transmission | |
| CN110570035B (en) | People flow prediction system for simultaneously modeling space-time dependency and daily flow dependency | |
| CN111860787A (en) | Short-term prediction method and device for coupling directed graph structure flow data containing missing data | |
| CN112668797B (en) | Long-short-period traffic prediction method | |
| CN112527547B (en) | Mechanical intelligent fault prediction method based on automatic convolution neural network | |
| CN112489420A (en) | Road traffic state prediction method, system, terminal and storage medium | |
| CN116861262A (en) | Perception model training method and device, electronic equipment and storage medium | |
| CN117095460A (en) | Self-supervision group behavior recognition method and system based on long-short time relation predictive coding | |
| CN114399901B (en) | Method and equipment for controlling traffic system | |
| CN113342474B (en) | Method, equipment and storage medium for predicting customer flow and training model | |
| CN112581177B (en) | Marketing prediction method combining automatic feature engineering and residual neural network | |
| CN114444813A (en) | Traffic flow prediction method based on deep learning | |
| CN114566048A (en) | Traffic control method based on multi-view self-adaptive space-time diagram network | |
| CN117808040B (en) | Method and device for predicting low forgetting hot events based on brain map | |
| CN116798521B (en) | Abnormality monitoring method and abnormality monitoring system for immune cell culture control system | |
| CN118247966A (en) | Traffic flow prediction method and device based on time-space sequence, storage medium and computing equipment | |
| CN116434537A (en) | Traffic flow prediction method based on time attention cyclic graph convolutional neural network | |
| CN117171543A (en) | Space-time data prediction method and data acquisition monitoring system | |
| CN116563021A (en) | Stock market analysis and prediction method, system, device and storage medium | |
| CN117593877A (en) | Short-time traffic flow prediction method based on integrated graph convolution neural network | |
| CN117743852A (en) | Traffic missing data filling method based on diffusion neural network | |
| CN117828546A (en) | Method, device and equipment for predicting net load of power grid |
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 |