CN115115094B - Traffic flow prediction method combining sequence local information and multi-sequence association relation - Google Patents
Traffic flow prediction method combining sequence local information and multi-sequence association relation Download PDFInfo
- Publication number
- CN115115094B CN115115094B CN202210588239.XA CN202210588239A CN115115094B CN 115115094 B CN115115094 B CN 115115094B CN 202210588239 A CN202210588239 A CN 202210588239A CN 115115094 B CN115115094 B CN 115115094B
- Authority
- CN
- China
- Prior art keywords
- sequence
- information
- flow
- association relation
- matrix
- 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 35
- 230000007246 mechanism Effects 0.000 claims abstract description 9
- 230000008569 process Effects 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 42
- 239000013598 vector Substances 0.000 claims description 14
- 230000004927 fusion Effects 0.000 claims description 6
- 238000005096 rolling process Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 abstract description 6
- 230000003068 static effect Effects 0.000 abstract description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000007787 long-term memory Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000006403 short-term memory Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000011144 upstream manufacturing Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000006872 improvement 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
- 238000012545 processing Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- General Physics & Mathematics (AREA)
- Educational Administration (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a traffic flow prediction method combining sequence local information and a multi-sequence association relation. The method comprises the steps of firstly extracting local dynamic change information of a historical flow sequence through a long-period memory network, then extracting global association relation information among different sequences through a graph convolution network, and finally combining the local dynamic change information with multi-sequence association relation information through a gating mechanism. In the process of extracting global relation information, the method further considers the dynamic relation of different sequences changing along with time on the basis of the static association relation so as to realize more accurate traffic flow prediction.
Description
Technical Field
The invention belongs to the field of data mining and intelligent traffic, and particularly relates to a traffic flow prediction method combining sequence local information and multi-sequence association relation.
Background
With the improvement of living standard and the development of urban traffic, the rapidly increased travel demands bring a series of traffic problems, and the problems of road congestion, traffic accidents and the like frequently occur. In this context, the establishment of an efficient intelligent traffic system is an important content for assisting traffic institutions in making scientific management decisions, wherein how to achieve accurate predictions of traffic flow is an important component of intelligent traffic systems. Accurate traffic flow prediction can assist in formulating a real-time control strategy, and has important significance for scientific management planning of traffic and safe and efficient travel of residents.
Traffic flow prediction refers to predicting future traffic flow by analytical mining of historical traffic flow data. At present, the prediction methods of traffic flow performed by researchers at home and abroad can be mainly divided into a statistical learning-based method, a traditional machine learning-based method and a deep learning-based method.
A typical representative based on a statistical learning method is a History Average (HA) method which calculates an Average value of History contemporaneous traffic as a current predicted value, but which is not suitable for traffic flow data that dynamically changes; differential average moving autoregressive (AutoRegressive Integrated Moving Strategy, ARIMA) is another typical statistical-based method that predicts by converting an unstable sequence into a stable sequence by difference; however, the effect of the above-described statistical learning-based predictive model is severely dependent on data quality.
Because of the significant non-linearity and uncertainty of traffic flow, machine learning methods are also widely used for traffic flow predictions, such as K-nearest neighbor (K Nearest Neighbors, KNN) algorithms, bayesian models, support vector regression (Support Vector Regression, SVR) algorithms, and the like. Although machine learning methods can model nonlinear relationships between data, they have high requirements on data characteristics, often requiring complex feature processing.
Deep neural networks are widely used in traffic flow prediction in recent years because they can effectively model high-dimensional spatiotemporal data relationships and do not require complex artificial feature engineering. For example, long and Short-term Memory Network (LSTM) class methods consider the dependency relationship between traffic, and convolutional neural network (Convolutional Neural Network, CNN) class methods primarily consider the space-time dependency relationship between different traffic data, but model the space dependency relationship more coarsely, so that it is difficult to realize effective expression of traffic road network data. The graph neural network (Graph Neural Network, GNN) can realize effective modeling of flow data space-time characteristics based on road network data by virtue of strong non-European space modeling capability, and has become one of the main means for current traffic flow prediction, and the graph rolling network (Graph Convolution Network, GCN) in the GNN is mainly used for current traffic flow prediction.
However, the current traffic flow prediction method generally carries out overall modeling on different traffic sequences to consider global relations among the sequences, or only considers dynamic changes of the traffic sequences, and lacks the work of combining the traffic sequences and mining potential information contained in traffic data to improve traffic flow prediction accuracy; in addition, the static graph on which the existing GCN-based method depends is difficult to accurately reflect the dynamic correlation relationship between different nodes along with time change.
Disclosure of Invention
In view of the problem that the complex mode and dynamic association relation in the data are difficult to effectively utilize in the traffic flow information extracted by the existing traffic flow prediction work, the invention provides a traffic flow prediction method combining sequence local information and multi-sequence association relation by combining the dynamic change information of the flow history sequence and the global relation information among a plurality of flow sequences, and realizes accurate traffic flow prediction by considering the dynamic relation of different sequences along with time in the extraction process of the global relation information.
The method comprises the following specific steps:
and (1) acquiring historical traffic flow of N flow sensors at T moments to form a traffic flow data set.
Forming a directed road network graph G (S, E) according to the geographical position of the sensor, wherein S is a vertex set in the directed graph, each vertex represents one flow sensor, and E is a directed edge set in the directed graph.
Step (3) extracting sequence local information, inputting the historical flow data into Long and Short-term memory network (LSTM) to obtain sequence local information C L 。
Extracting multi-sequence association relation information, which comprises the following sub-steps:
step (4.1) determining a geographical location adjacency matrix A between different sensors based on road network G (S, E) G ;
Step (4.2) forming a historical traffic correlation matrix A according to the historical traffic data of each node on the graph T ;
Step (4.3) the geographic position adjacency matrix A G And a historical traffic adjacency matrix A T Combining to form dynamic adjacency matrix A D ;
Step (4.4) dynamic adjacency matrix A D And the historical flow data are input into a graph rolling network (Graph Convolution Network, GCN) to obtain multi-sequence association relation information C G ;
Step (5) extracting the sequence local information C based on LSTM L And multi-sequence association relation information C extracted based on GCN G Information fusion is carried out based on a gating mechanism to obtain a final information representation C;
step (6), multi-scale traffic flow prediction, comprising the following sub-steps:
step (6.1) determining the number of times K at which flow prediction is to be performed, based on which K1 x 1 convolution kernels are selected;
and (6.2) applying K different convolution kernels to the C to obtain traffic flow prediction results at K moments.
The invention has the beneficial effects that: in the extraction process of the multi-sequence association relation, considering that most of the existing GCN-based methods only model a single static diagram, the invention combines the historical flow relation correlation matrix which dynamically changes along with time to form a dynamic adjacent matrix on the basis of the traditional geographic position adjacent matrix, and the dynamic sequence relation diagram is constructed through the dynamic adjacent matrix so as to realize more accurate and effective traffic flow prediction.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
fig. 2 is a schematic view of LSTM.
Detailed Description
The invention designs a traffic flow prediction method combining sequence local information and multi-sequence association relation aiming at the defect that the current flow prediction method only considers the global relation among different sequences (nodes and sensors) and regards the global relation as a static relation or only considers the dynamic change of the flow sequence.
The traffic flow prediction method based on the modeling of the sequence and the sequence relation, which is designed by the invention, is specifically described below, and the implementation process of the method is shown in fig. 1.
For convenience of description, the relevant symbols are defined as follows:
flow matrixA flow matrix formed by historical traffic flows of the N sensors at T time points.
Road network G (S, E) is a directed graph, where S is the set of vertices in the directed graph, vertex v ε S, representing different sensors. E is a set of directed edges in the directed graph, and the directed edge E E is generated by connecting the upstream sensor with the downstream sensor when the two sensors are within a certain distance threshold d and have a relationship between the upstream and the downstream.
At time t, all sensorsVector of flow components at current moment of device Represents a column in the flow matrix V, where +.>The flow rate of the i-th sensor at time t is shown.
The flow sequence formed by the sensor i at all times isRepresenting one row of the flow matrix V, intercepting a section with the length w during each training, wherein the matrix Q formed by the intercepted flows of all the sensors is one of the submatrices of V.
The specific steps of the invention are as follows:
step (1), data acquisition: the historical traffic flow of the N sensors at T time points is acquired to form a historical flow matrix V.
And (2) forming a road network G (S, E) according to the geographic position of the sensor, and setting a time window as w.
Step (3) extracting sequence local information, and obtaining flow x of all sensors of w time steps according to a time window w t-w+1 ,…,x t . As shown in fig. 2, the vector x of the flow of all the sensors at the current moment of each moment is formed t Sequentially inputting Long and Short-term memory network (LSTM) units, wherein the LSTM units are combined with a hidden vector h at the previous moment t-1 And cell state vector c t-1 Obtaining a hidden vector h output at the current moment t And cell state vector c t . The calculation steps of the whole process are as follows:
i t =σ(W xi x t +W hi h t-1 +b i )#(3.1)
f t =σ(W xf x t +W hf h t-1 +b f )#(3.2)
o t =σ(W xo x t +W ho h t-1 +b o )#(3.3)
c t =f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c )#(3.4)
h t =o t ⊙tanh(c t )#(3.5)
wherein i is t ,f t ,o t Input gate, forget gate and output gate at time t respectively, W represents a parameter matrix which can be learned, b i ,b f ,b o ,b c Represents the bias vector, σ represents the sigmoid activation function, tan h represents the hyperbolic tangent activation function, and as such, it represents the multiplication of parity elements.
The hidden vector h output at all moments t Splicing to obtain sequence local information C L 。
Step (4) extracting multi-sequence association relation information
Step (4.1) obtaining a geographic position neighbor matrix based on the directed graph G (S, E)Specifically, if there is a directional edge between sensor i and sensor j that points from i to j, then element A in row i and column j in the adjacent matrix G [i][j]Set to 1, otherwise set to 0.
Step (4.2) modeling a dynamic historical traffic correlation matrix A based on historical traffic data T . Specifically, A T [i][j]Representing the historical flow similarity between sensor i and sensor j, calculated by
Wherein the method comprises the steps ofRepresenting the ith transmissionThe sensor intercepts the average flow of the historical flow sequence, A T [i][j]The closer to 1 the absolute value of (c) indicates a higher historical flow similarity between sensor i and sensor j.
Step (4.3) existing methods of traffic prediction based on graph rolling networks (Graph Convolution Network, GCN) typically use distances between nodes to build a static graph, but ignore the case where there is a change in correlation between nodes at different points in time. The invention combines the dynamic historical traffic flow correlation matrix established based on the historical traffic flow data of different nodes on the basis of the distance-based adjacency matrix to form the dynamic adjacency matrix with different moments
A D =αA G +βA T #(4.2)
Where α, β represent weight coefficients. In particular, due to A G The element with the upper 0 indicates that two nodes are not adjacent in distance or there is no flow direction relationship of traffic, thus for A G Element with upper 0, in the final dynamic adjacency matrix A D Is also kept at 0 at the corresponding position of (c).
Step (4.4) the dynamic adjacency matrix A is obtained D And the historical flow vector q of each node are input into GCN to obtain multi-sequence association relation information C G The method comprises the steps of carrying out a first treatment on the surface of the In the GCN, the update formula of the node is:
wherein the method comprises the steps ofRepresentation matrix->Representing an identity matrix>Matrix representing information of all nodes of the first layer, < >>Representing a layer I learnable parameter matrix, in particular H (0) =Q。
The whole graph is updated L times to obtain H (L) Representing the updated information set of all nodes, namely the final sequence association relation information C G 。
Step (5) extracting the sequence local information C based on LSTM L And multi-sequence association relation information C extracted based on GCN G And information fusion is carried out based on a gating mechanism to obtain a final information representation C. The information fusion process comprises the following steps:
first, the weight g is obtained through a gating mechanism c :
g c =σ(W G C G +W L C L +b c )#(4.4)
Wherein W is G ,B is a parameter matrix which can be learned in a gating mechanism c Is a bias vector.
Then according to the learned weight g c Sequence local information C L And multi-sequence association relation information C G Fusion is carried out to obtain a final information representation C:
C=g c ⊙C G +(1-g c )⊙C L #(4.5)
and (6) after the final information representation C is obtained, carrying out multi-scale traffic flow prediction according to the requirements.
Step (6.1) assuming the current time is t and the goal is to predict traffic flow at K times after t times, K different 1×1 convolution kernels K are selected 1 ,…,k K ;
And (6.2) applying K different convolution kernels to the C to obtain the traffic flow prediction results of K future moments.
Claims (1)
1. The traffic flow prediction method combining sequence local information and multi-sequence association relation is characterized by comprising the following specific steps:
step (1), acquiring historical traffic flow of N flow sensors at T moments to form a traffic flow data set;
forming a directed road network graph g (S, E) according to the geographic position of the sensor, wherein S is a vertex set in the directed graph, each vertex represents one flow sensor, and E is a set of directed edges in the directed graph;
step (3) extracting sequence local information, inputting the historical flow data into a long-short-period memory network to obtain sequence local information C L ;
Extracting multi-sequence association relation information, which comprises the following sub-steps:
step (4.1) determining a geographical location adjacency matrix A between different sensors based on road network G (S, E) G ;
Step (4.2) forming a historical traffic correlation matrix A according to the historical traffic data of each node on the graph T ;
Step (4.3) the geographic position adjacency matrix A G And a historical traffic adjacency matrix A T Combining to form dynamic adjacency matrix A D ;
Step (4.4) dynamic adjacency matrix A D And the historical flow data are input into a graph convolution network to obtain multi-sequence association relation information C G ;
Step (5) sequence local information C L And multi-sequence association relation information C G Information fusion is carried out based on a gating mechanism to obtain a final information representation C;
step (6), multi-scale traffic flow prediction, comprising the following sub-steps:
step (6.1) determining the number of times K at which flow prediction is to be performed, based on which K1 x 1 convolution kernels are selected;
step (6.2) K different convolution kernels are applied to C, and traffic flow prediction results at K moments are obtained;
the step (3) is specifically as follows:
acquiring the flow of all the sensors of w time steps according to a time window w;
vector formed by current time flow of all sensors at each timex t Sequentially inputting a long-period memory network and a short-period memory network;
splicing hidden vectors output by all-time long-short-term memory networks to obtain sequence local information C L ;
The step (4.1) is specifically as follows:
obtaining a geographic position neighbor matrix A based on the directed graph G (S, E) G If there is a directed edge between sensor i and sensor j pointing from i to j, then element A in row i and column j in the adjacency matrix G [i][j]Set to 1, otherwise set to 0;
the step (4.2) is specifically as follows: with A T [i][j]Representing the historical flow similarity between sensor i and sensor j, then:
wherein the method comprises the steps ofMean flow representing the sequence of i-th sensor intercepted historical flow,/->The flow of the ith sensor at the moment ti is represented, and w is a time window;
the step (4.3) is specifically as follows: dynamic adjacency matrix A D The calculation is as follows:
A D =αA G +βA T
wherein alpha and beta represent weight coefficients;
in the step (4.4), the update formula of the nodes in the graph rolling network is as follows:
wherein D represents a degree matrix, I N Representing the identity matrix, H (l) Representing all nodes of layer IMatrix of information, W (l) Representing a layer I learnable parameter matrix, and representing an activation function by a ReLU;
the whole graph is updated L times to obtain H (L) Representing the updated information set of all nodes, namely the final sequence association relation information C G ;
In the step (5), the process of information fusion through a gating mechanism is as follows:
first, the weight g is obtained through a gating mechanism c :
g c =σ(W G C G +W L C L +b c )
Wherein W is G ,W L B is a parameter matrix which can be learned in a gating mechanism c Is a bias vector;
then according to the weight g c Sequence local information C L And multi-sequence association relation information C G Fusion is carried out to obtain a final information representation C:
C=g c ⊙C G +(1-g c )⊙C L 。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210588239.XA CN115115094B (en) | 2022-05-26 | 2022-05-26 | Traffic flow prediction method combining sequence local information and multi-sequence association relation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210588239.XA CN115115094B (en) | 2022-05-26 | 2022-05-26 | Traffic flow prediction method combining sequence local information and multi-sequence association relation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115115094A CN115115094A (en) | 2022-09-27 |
CN115115094B true CN115115094B (en) | 2024-02-06 |
Family
ID=83326655
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210588239.XA Active CN115115094B (en) | 2022-05-26 | 2022-05-26 | Traffic flow prediction method combining sequence local information and multi-sequence association relation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115115094B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112418547A (en) * | 2020-12-03 | 2021-02-26 | 北京工业大学 | Bus stop passenger flow prediction method based on GCN-LSTM combined model |
CN113268916A (en) * | 2021-04-07 | 2021-08-17 | 浙江工业大学 | Traffic accident prediction method based on space-time graph convolutional network |
CN113971479A (en) * | 2021-10-01 | 2022-01-25 | 浙江省机电设计研究院有限公司 | Daily traffic flow parameter prediction method integrating holiday and festival effects |
US11238729B1 (en) * | 2020-09-11 | 2022-02-01 | Toyota Motor Engineering & Manufacturing North America, Inc. | Systems and methods for traffic flow prediction |
CN114169647A (en) * | 2022-01-07 | 2022-03-11 | 重庆大学 | Traffic prediction method and system of continuous memory self-adaptive heterogeneous space-time diagram convolution |
CN114202122A (en) * | 2021-12-13 | 2022-03-18 | 河北师范大学 | Urban traffic flow prediction method based on Markov cluster map attention network |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070208498A1 (en) * | 2006-03-03 | 2007-09-06 | Inrix, Inc. | Displaying road traffic condition information and user controls |
CN111130839B (en) * | 2019-11-04 | 2021-07-16 | 清华大学 | Flow demand matrix prediction method and system |
-
2022
- 2022-05-26 CN CN202210588239.XA patent/CN115115094B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11238729B1 (en) * | 2020-09-11 | 2022-02-01 | Toyota Motor Engineering & Manufacturing North America, Inc. | Systems and methods for traffic flow prediction |
CN112418547A (en) * | 2020-12-03 | 2021-02-26 | 北京工业大学 | Bus stop passenger flow prediction method based on GCN-LSTM combined model |
CN113268916A (en) * | 2021-04-07 | 2021-08-17 | 浙江工业大学 | Traffic accident prediction method based on space-time graph convolutional network |
CN113971479A (en) * | 2021-10-01 | 2022-01-25 | 浙江省机电设计研究院有限公司 | Daily traffic flow parameter prediction method integrating holiday and festival effects |
CN114202122A (en) * | 2021-12-13 | 2022-03-18 | 河北师范大学 | Urban traffic flow prediction method based on Markov cluster map attention network |
CN114169647A (en) * | 2022-01-07 | 2022-03-11 | 重庆大学 | Traffic prediction method and system of continuous memory self-adaptive heterogeneous space-time diagram convolution |
Non-Patent Citations (3)
Title |
---|
Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting;Bing Yu 等;https://arxiv.org/pdf/arXiv:1709.04875v4;1-7 * |
一种融合局部和全局空间相关性的交通预测混合模型;stranger_Ss;https://blog.csdn.net/stranger_Ss/article/details/124716999;2-10 * |
基于时空相关性的短时交通流预测模型;高小婷;;电脑知识与技术(第18期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN115115094A (en) | 2022-09-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112801404B (en) | Traffic prediction method based on self-adaptive space self-attention force diagram convolution | |
Wang et al. | Long-term traffic prediction based on lstm encoder-decoder architecture | |
CN107529651B (en) | Urban traffic passenger flow prediction method and equipment based on deep learning | |
CN111612243A (en) | Traffic speed prediction method, system and storage medium | |
CN112270355B (en) | Active safety prediction method based on big data technology and SAE-GRU | |
Jiang et al. | Attention-LSTM architecture combined with Bayesian hyperparameter optimization for indoor temperature prediction | |
CN110675632B (en) | Vehicle short-time trajectory prediction control method aiming at multi-feature space and data sparseness | |
CN113538910A (en) | Self-adaptive full-chain urban area network signal control optimization method | |
CN113554060B (en) | LSTM neural network track prediction method integrating DTW | |
Faiq et al. | Prediction of energy consumption in campus buildings using long short-term memory | |
CN115346372B (en) | Multi-component fusion traffic flow prediction method based on graph neural network | |
CN113112791A (en) | Traffic flow prediction method based on sliding window long-and-short term memory network | |
CN112766603A (en) | Traffic flow prediction method, system, computer device and storage medium | |
Yu et al. | 3D CNN-based accurate prediction for large-scale traffic flow | |
Zhang et al. | A local semi-supervised ensemble learning strategy for the data‐driven soft sensor of the power prediction in wind power generation | |
CN115376317A (en) | Traffic flow prediction method based on dynamic graph convolution and time sequence convolution network | |
CN115755219A (en) | Flood forecast error real-time correction method and system based on STGCN | |
Han et al. | Short-term 4D trajectory prediction based on LSTM neural network | |
Ren et al. | A prediction model for surface deformation caused by underground mining based on spatio-temporal associations | |
CN112927507B (en) | Traffic flow prediction method based on LSTM-Attention | |
CN115115094B (en) | Traffic flow prediction method combining sequence local information and multi-sequence association relation | |
CN114065996A (en) | Traffic flow prediction method based on variational self-coding learning | |
CN117556133A (en) | Neural time gate time sequence enhanced session recommendation method based on graph neural network | |
Su et al. | Trajectory prediction based on machine learning | |
Goswami et al. | Traffic Flow Prediction Using Deep Learning Techniques |
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 |