CN114971093A - Method, system, equipment and medium for predicting urban road traffic flow attribute - Google Patents

Method, system, equipment and medium for predicting urban road traffic flow attribute Download PDF

Info

Publication number
CN114971093A
CN114971093A CN202210915226.9A CN202210915226A CN114971093A CN 114971093 A CN114971093 A CN 114971093A CN 202210915226 A CN202210915226 A CN 202210915226A CN 114971093 A CN114971093 A CN 114971093A
Authority
CN
China
Prior art keywords
traffic flow
flow attribute
matrix
road
prediction model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202210915226.9A
Other languages
Chinese (zh)
Other versions
CN114971093B (en
Inventor
谌恺祺
邓敏
石岩
雷凯媛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to CN202210915226.9A priority Critical patent/CN114971093B/en
Publication of CN114971093A publication Critical patent/CN114971093A/en
Application granted granted Critical
Publication of CN114971093B publication Critical patent/CN114971093B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic 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
    • G06N3/048—Activation functions
    • 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
    • G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • General Business, Economics & Management (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • Probability & Statistics with Applications (AREA)
  • Educational Administration (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Analytical Chemistry (AREA)
  • Primary Health Care (AREA)
  • Chemical & Material Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)

Abstract

The embodiment of the disclosure provides a method, a system, equipment and a medium for predicting urban road traffic flow attributes, which belong to the technical field of data processing, and specifically comprise the following steps: collecting urban road network data, vehicle GPS data and road section traffic flow data; modeling the spatial dependence by means of a Markov chain; the conventional spectrogram convolution structure is modified to
Figure 538235DEST_PATH_IMAGE002
Weighted directed graph for edge weights
Figure 127479DEST_PATH_IMAGE004
Learning and fitting spatial dependence characteristics of traffic flow; carrying out time-dependent modeling by constructing a traffic flow key frame sequence; constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer; updating the learnable parameters in the traffic flow attribute prediction model until the traffic flow attribute prediction model has the lowest prediction error in the verification set; and (4) predicting the traffic flow attribute in the target scene by using the traffic flow attribute prediction model updated in the step (7). By the scheme, the prediction accuracy and robustness are improved.

Description

Method, system, equipment and medium for predicting urban road traffic flow attribute
Technical Field
The disclosed embodiments relate to the technical field of data processing, and in particular, to a method, a system, a device, and a medium for predicting urban road traffic flow attributes.
Background
At present, traffic flow prediction is a technology for estimating traffic flow attributes including traffic speed, traffic volume and the like at a future moment by establishing a proper model based on a historical traffic flow state or an internal dynamic mechanism of a traffic flow space-time process. The traffic flow prediction technology is used as a core component for efficient operation and deployment of an Intelligent Transportation System (ITS), plays a direct and important information support role in downstream service type traffic analysis applications such as travel time estimation, path planning, unmanned driving, congestion detection and the like, and obviously influences urban traffic management efficiency and daily travel experience of residents.
In recent years, in the background of the era of big data, various sensors including a GPS, a ground induction coil and the like are widely laid, so that a large amount of traffic flow historical data is generated, and a solid data basis is provided for realizing accurate traffic flow prediction. Meanwhile, with the rise of artificial intelligence data mining paradigm represented by various deep neural networks, a strong intelligent model greatly inspires a design idea of a more accurate and more stable traffic flow prediction technical framework. Under the guidance of the opportunity in this era, most of the existing advanced traffic flow prediction methods are based on a data-driven framework, and by improving various neural networks to fit hidden complex traffic flow spatio-temporal features and evolution mechanisms in data in a black box manner, representative models are as follows: T-GCN, DCRNN, etc. Although the method can realize higher prediction precision by means of strong complex function approximation capability of a deep neural network, the method ignores the internal evolution mechanism of a traffic flow process which is a complex system product in the modeling process, so that the prediction result is difficult to interpret and has low credibility. Moreover, under the conditions of low large data value density and poor information quality, the brute force fitting mode based on pure data driving is easily influenced by noise in data, so that the robustness and the upper limit of precision of the method are greatly reduced.
Therefore, an urban road traffic flow attribute prediction method with high accuracy and robustness is urgently needed.
Disclosure of Invention
In view of this, embodiments of the present disclosure provide a method, a system, a device, and a medium for predicting an urban road traffic flow attribute, which at least partially solve the problem in the prior art that the accuracy and the robustness are poor.
In a first aspect, an embodiment of the present disclosure provides a method for predicting an urban road traffic flow attribute, including:
step 1, collecting urban road network data, vehicle GPS data and road section traffic flow data;
step 2, constructing the urban road network into a directed graph structure with the right
Figure 985356DEST_PATH_IMAGE001
In which
Figure 123076DEST_PATH_IMAGE002
Representing cities for node sets
Figure 850861DEST_PATH_IMAGE003
A collection of road segments to be joined together,
Figure 808452DEST_PATH_IMAGE004
the weight matrix represents the strength of the correlation between every two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 778420DEST_PATH_IMAGE005
representing intersections for a matrix of traffic flow attributesThrough-flow in
Figure 770647DEST_PATH_IMAGE006
Within a time stamp
Figure 403754DEST_PATH_IMAGE003
Attribute information on the road segment;
step 3, every road section in the road network
Figure 114221DEST_PATH_IMAGE007
As one state, its entire road network structure
Figure 920503DEST_PATH_IMAGE008
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
step 4, transforming the conventional spectrogram convolution structure to ensure that the conventional spectrogram convolution structure is modified
Figure 767236DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 836823DEST_PATH_IMAGE008
Learning and fitting the spatial dependence characteristics of the traffic flow;
step 5, based on the prior cognition of trend and periodicity of the traffic flow time-space process, building a traffic flow key frame sequence to carry out time-dependent modeling;
step 6, after the modeling of the space dependency and the time dependency is completed, utilizing the weighted directed graph defined in the step 4
Figure 769007DEST_PATH_IMAGE008
Convolution operation on
Figure 582242DEST_PATH_IMAGE010
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer;
step 7, calculating the real traffic flow attribute value on the training set
Figure 47596DEST_PATH_IMAGE011
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 756926DEST_PATH_IMAGE012
Error between, updating learnable parameters in traffic flow attribute prediction model
Figure 441986DEST_PATH_IMAGE013
、
Figure 793333DEST_PATH_IMAGE014
、
Figure 614658DEST_PATH_IMAGE015
And
Figure 557206DEST_PATH_IMAGE016
until the traffic flow attribute prediction model has the lowest prediction error in the verification set;
and 8, predicting the traffic flow attribute in the target scene by using the traffic flow attribute prediction model updated in the step 7.
According to a specific implementation manner of the embodiment of the present disclosure, the step 3 specifically includes:
step 3.1, for the binary road section pair formed by all road sections in the road network
Figure 463982DEST_PATH_IMAGE017
And calculating the times of occurrence of the GPS track and recording the times as
Figure 353441DEST_PATH_IMAGE018
And form a matrix
Figure 29273DEST_PATH_IMAGE019
：
Figure 104240DEST_PATH_IMAGE021
；
Step 3.2, passing throughLarge likelihood method for calculating Markov chain transfer matrix
Figure 763891DEST_PATH_IMAGE022
；
Step 3.3, assigning the value of the Markov transfer matrix to the weight matrix of the road network
Figure 191462DEST_PATH_IMAGE023
And completing the modeling of the spatial dependency.
According to a specific implementation manner of the embodiment of the present disclosure, the step 4 specifically includes:
step 4.1, to the weight matrix
Figure 987379DEST_PATH_IMAGE009
Performing characteristic decomposition to obtain the final product by the following formula
Figure 740572DEST_PATH_IMAGE009
Maximum eigenvalue
Figure 356361DEST_PATH_IMAGE024
Corresponding feature vector
Figure 853201DEST_PATH_IMAGE025
：
Figure 503625DEST_PATH_IMAGE026
；
Step 4.2, define the matrix
Figure 926254DEST_PATH_IMAGE027
Is a diagonal matrix whose diagonal elements are formed by feature vectors
Figure 294919DEST_PATH_IMAGE028
Is composed of, i.e.
Figure 329871DEST_PATH_IMAGE030
；
Step 4.3, calculating the weighted directed graph
Figure 834801DEST_PATH_IMAGE031
Laplacian matrix of
Figure 726534DEST_PATH_IMAGE032
：
Figure 316915DEST_PATH_IMAGE034
；
Step 4.4, the Laplace matrix
Figure 155558DEST_PATH_IMAGE035
Carrying out symmetry to form a symmetrical Laplace matrix
Figure 780575DEST_PATH_IMAGE036
：
Figure 780892DEST_PATH_IMAGE038
；
Step 4.5, based on
Figure 622684DEST_PATH_IMAGE039
Defining weighted direction
Figure 733859DEST_PATH_IMAGE040
The convolution operator of
Figure 213382DEST_PATH_IMAGE041
The convolution operator acts on the traffic flow attribute matrix
Figure 650180DEST_PATH_IMAGE042
Slicing at a certain time
Figure 215153DEST_PATH_IMAGE043
Is operated as
Figure 395599DEST_PATH_IMAGE044
Wherein, in the step (A),
Figure 995207DEST_PATH_IMAGE045
is a symmetric Laplace matrix
Figure 399644DEST_PATH_IMAGE046
A matrix of feature vectors is formed from the feature vectors,
Figure 888132DEST_PATH_IMAGE047
the product of the Hadamard is used as the target,
Figure 872268DEST_PATH_IMAGE048
in order to calculate the operators for the convolution,
Figure 591963DEST_PATH_IMAGE049
parameters which can be learned in the convolution operator;
step 4.6, simplification by utilizing Chebyshev polynomial
Figure 370563DEST_PATH_IMAGE050
Comprises the following steps:
Figure 644549DEST_PATH_IMAGE051
；
step 4.7, based on the weighted directed graph in step 4.6
Figure 166797DEST_PATH_IMAGE008
The above convolution formula is used for calculating the traffic flow attribute matrix
Figure 475419DEST_PATH_IMAGE052
The convolution operation above:
Figure 424920DEST_PATH_IMAGE053
wherein the content of the first and second substances,
Figure 956177DEST_PATH_IMAGE054
represents the network parameters of the convolutional layer, and
Figure 750957DEST_PATH_IMAGE055
and
Figure 445244DEST_PATH_IMAGE056
representing the dimensions of the convolutional layer input features and the dimensions of the output features, respectively.
According to a specific implementation manner of the embodiment of the present disclosure, the step 5 specifically includes:
step 5.1, aiming at the periodic characteristics of the traffic flow and the predicted time
Figure 300068DEST_PATH_IMAGE057
Selecting
Figure 548646DEST_PATH_IMAGE057
Front side
Figure 412697DEST_PATH_IMAGE058
The traffic flow attributes at corresponding time in each period form a periodic frame sequence
Figure 695911DEST_PATH_IMAGE059
I.e. by
Figure 783953DEST_PATH_IMAGE060
Wherein
Figure 690466DEST_PATH_IMAGE061
Indicating the number of time stamps included in one period;
step 5.2, aiming at the trend characteristics of the traffic flow and the predicted time
Figure 358208DEST_PATH_IMAGE057
Selecting
Figure 495928DEST_PATH_IMAGE057
Front part
Figure 754871DEST_PATH_IMAGE062
Trend frame sequence composed by traffic flow attribute of each moment
Figure 978042DEST_PATH_IMAGE063
I.e. by
Figure 449475DEST_PATH_IMAGE064
。
According to a specific implementation manner of the embodiment of the present disclosure, the step 6 specifically includes:
step 6.1, in the hidden layer, for periodic frame sequences
Figure 441702DEST_PATH_IMAGE065
Trending frame sequences
Figure 809229DEST_PATH_IMAGE063
Using two convolutions respectively
Figure 18231DEST_PATH_IMAGE066
And
Figure 27776DEST_PATH_IMAGE067
learning is carried out to respectively obtain corresponding hidden layer characteristics
Figure 874509DEST_PATH_IMAGE068
And
Figure 944096DEST_PATH_IMAGE069
the formula is as follows:
Figure 673018DEST_PATH_IMAGE071
Figure 486253DEST_PATH_IMAGE073
wherein
Figure 187493DEST_PATH_IMAGE074
And
Figure 427981DEST_PATH_IMAGE075
respectively learning periodic convolution
Figure 847461DEST_PATH_IMAGE076
Convolution with learning tendency
Figure 431764DEST_PATH_IMAGE077
The learnable parameter of (1);
step 6.2, hiding layer characteristics related to trend in the aggregation layer
Figure 518669DEST_PATH_IMAGE078
Periodicity dependent hidden layer features
Figure 930059DEST_PATH_IMAGE079
Splicing on characteristic dimension to form a matrix
Figure 571256DEST_PATH_IMAGE080
；
Step 6.3, matrix
Figure 726293DEST_PATH_IMAGE081
Are scrambled in a characteristic dimension to form
Figure 402125DEST_PATH_IMAGE082
；
Step 6.4, by a convolution
Figure 984416DEST_PATH_IMAGE083
To pair
Figure 175226DEST_PATH_IMAGE084
Performing deep feature extraction to form
Figure 868376DEST_PATH_IMAGE085
:
Figure 180407DEST_PATH_IMAGE087
；
Step 6.5, through a fully connected network, will
Figure 668020DEST_PATH_IMAGE088
Is compressed into 1-dimension to form
Figure 283809DEST_PATH_IMAGE089
As a model to
Figure 780649DEST_PATH_IMAGE090
At the time of day, the user may,
Figure 227811DEST_PATH_IMAGE091
a traffic flow attribute prediction model of a traffic flow attribute prediction result of a link:
Figure 151905DEST_PATH_IMAGE092
wherein
Figure 254990DEST_PATH_IMAGE093
Is a learnable parameter of a fully connected network.
According to a specific implementation manner of the embodiment of the present disclosure, the step 7 specifically includes:
step 7.1, borrowing root mean square error, defining real traffic flow attribute value
Figure 289942DEST_PATH_IMAGE094
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 60452DEST_PATH_IMAGE095
An error of
Figure 388403DEST_PATH_IMAGE096
：
Figure 978784DEST_PATH_IMAGE097
；
Step 7.2, based on calculation
Figure 817427DEST_PATH_IMAGE098
Iteratively updating learnable parameters in traffic flow attribute prediction model by adopting SGD optimizer
Figure 442444DEST_PATH_IMAGE099
、
Figure 708340DEST_PATH_IMAGE100
、
Figure 786017DEST_PATH_IMAGE101
And
Figure 959510DEST_PATH_IMAGE102
until the traffic flow attribute prediction model has the lowest prediction error in the validation set.
In a second aspect, an embodiment of the present disclosure provides an urban road traffic flow attribute prediction system, including:
the acquisition module is used for acquiring urban road network data, vehicle GPS data and road section traffic flow data;
a first construction module for constructing the city road network into a weighted directed graph structure
Figure 439033DEST_PATH_IMAGE103
Wherein
Figure 610251DEST_PATH_IMAGE104
Representing cities for node sets
Figure 673760DEST_PATH_IMAGE105
A collection of road segments to be joined together,
Figure 854205DEST_PATH_IMAGE106
the weight matrix represents the strength of the correlation between every two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 188235DEST_PATH_IMAGE107
representing the traffic flow in the traffic flow attribute matrix
Figure 795934DEST_PATH_IMAGE108
Is within a time stamp
Figure 848203DEST_PATH_IMAGE109
Attribute information on the road segment;
a first modeling module for modeling each road segment in the road network
Figure 629077DEST_PATH_IMAGE110
As one state, its entire road network structure
Figure 817613DEST_PATH_IMAGE111
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
a transformation module for transforming the conventional spectrogram convolution structure
Figure 596213DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 135779DEST_PATH_IMAGE008
Learning and fitting spatial dependence characteristics of traffic flow;
the second modeling module is used for modeling the time dependence by constructing a traffic flow key frame sequence based on the prior cognition of the trend and the periodicity of the traffic flow time-space process;
a second construction module, configured to utilize the weighted directed graph defined in step 4 after completing modeling of the spatial dependency and the time dependency
Figure 890983DEST_PATH_IMAGE111
Convolution operation on
Figure 199605DEST_PATH_IMAGE112
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer;
a training module for calculating real traffic flow attribute value on the training set
Figure 414686DEST_PATH_IMAGE113
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 441547DEST_PATH_IMAGE114
Error between, updating learnable parameters in traffic flow attribute prediction model
Figure 501907DEST_PATH_IMAGE115
、
Figure 399456DEST_PATH_IMAGE100
、
Figure 785438DEST_PATH_IMAGE116
And
Figure 34017DEST_PATH_IMAGE117
until the traffic flow attribute prediction model has the lowest prediction error in the verification set;
and a prediction module, configured to predict the traffic flow attribute in the target scene using the traffic flow attribute prediction model updated in step 7.
In a third aspect, an embodiment of the present disclosure further provides an electronic device, where the electronic device includes:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to cause the at least one processor to perform the method of predicting an attribute of urban road traffic flow in the first aspect or any implementation manner of the first aspect.
In a fourth aspect, the disclosed embodiments also provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the urban road traffic flow attribute prediction method in the first aspect or any implementation manner of the first aspect.
In a fifth aspect, the disclosed embodiments also provide a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the urban road traffic flow attribute prediction method of the first aspect or any implementation manner of the first aspect.
The urban road traffic flow attribute prediction scheme in the embodiment of the disclosure comprises the following steps: step 1, collecting urban road network data, vehicle GPS data and road section traffic flow data; step 2, constructing the urban road network into a directed graph structure with the right
Figure 402462DEST_PATH_IMAGE001
Wherein
Figure 685676DEST_PATH_IMAGE002
Representing cities for node sets
Figure 976980DEST_PATH_IMAGE003
A collection of road segments to be joined together,
Figure 712855DEST_PATH_IMAGE004
the weight matrix represents the strength of the correlation between every two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 380597DEST_PATH_IMAGE005
representing the traffic flow in the traffic flow attribute matrix
Figure 315055DEST_PATH_IMAGE006
Within a time stamp
Figure 980522DEST_PATH_IMAGE003
Attribute information on the road segment; step 3, every road section in the road network
Figure 203693DEST_PATH_IMAGE007
As one state, its entire road network structure
Figure 206284DEST_PATH_IMAGE008
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
step 4, transforming the conventional spectrogram convolution structure to ensure that the conventional spectrogram convolution structure is modified
Figure 697046DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 330153DEST_PATH_IMAGE008
Learning and fitting spatial dependence characteristics of traffic flow; step 5, based on the prior cognition of trend and periodicity of the traffic flow time-space process, building a traffic flow key frame sequence to carry out time-dependent modeling; step 6, after the modeling of the space dependency and the time dependency is completed, utilizing the weighted directed graph defined in the step 4
Figure 40620DEST_PATH_IMAGE008
Convolution operation on
Figure 784585DEST_PATH_IMAGE010
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer; step 7, calculating the real traffic flow attribute value on the training set
Figure 896897DEST_PATH_IMAGE011
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 497643DEST_PATH_IMAGE012
Updating learnable parameters in the traffic flow attribute prediction model
Figure 695406DEST_PATH_IMAGE013
、
Figure 446325DEST_PATH_IMAGE014
、
Figure 911679DEST_PATH_IMAGE015
And
Figure 152167DEST_PATH_IMAGE016
until the traffic flow attribute prediction model has the lowest prediction error in the verification set; and 8, predicting the traffic flow attribute in the target scene by using the traffic flow attribute prediction model updated in the step 7.
The beneficial effects of the embodiment of the disclosure are: according to the scheme, the traffic flow space dependency of the road section level is modeled in a fine-grained mode through the Markov matrix and abstracted to a weighted directed graph structure, so that the traditional spectrogram convolution is improved in a targeted mode, the weighted directed graph can be embedded, and the space dependency is fitted in the learning process. In the aspect of time dependence fitting, a traffic flow key frame sequence is constructed by recognizing the trend and periodicity of the time evolution of the traffic flow, so that the fitting of the time dependence of the traffic flow under the driving of prior knowledge is realized, and the accuracy and the robustness of urban road traffic flow attribute prediction are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required to be used in the embodiments will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of an urban road traffic flow attribute prediction method according to an embodiment of the disclosure;
fig. 2 is a schematic view of another data processing flow of a method for predicting an urban road traffic flow attribute according to an embodiment of the present disclosure;
fig. 3 is a symmetric laplacian matrix provided in an embodiment of the present disclosure
Figure 837226DEST_PATH_IMAGE046
A visualization result diagram of (a);
fig. 4 is a schematic structural diagram of an urban road traffic flow attribute prediction system according to an embodiment of the present disclosure;
fig. 5 is a schematic view of an electronic device provided in an embodiment of the present disclosure.
Detailed Description
The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
The embodiments of the present disclosure are described below with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure in the specification. It is to be understood that the described embodiments are merely illustrative of some, and not restrictive, of the embodiments of the disclosure. The disclosure may be embodied or carried out in various other specific embodiments, and various modifications and changes may be made in the details within the description without departing from the spirit of the disclosure. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
It is noted that various aspects of the embodiments are described below within the scope of the appended claims. It should be apparent that the aspects described herein may be embodied in a wide variety of forms and that any specific structure and/or function described herein is merely illustrative. Based on the disclosure, one skilled in the art should appreciate that one aspect described herein may be implemented independently of any other aspects and that two or more of these aspects may be combined in various ways. For example, an apparatus may be implemented and/or a method practiced using any number of the aspects set forth herein. Additionally, such an apparatus may be implemented and/or such a method may be practiced using other structure and/or functionality in addition to one or more of the aspects set forth herein.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present disclosure, and the drawings only show the components related to the present disclosure rather than the number, shape and size of the components in actual implementation, and the type, amount and ratio of the components in actual implementation may be changed arbitrarily, and the layout of the components may be more complicated.
In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects may be practiced without these specific details.
Most of the existing intelligent models for traffic flow prediction are based on a pure data-driven framework, and most of the design of the method structure directly migrates general models in the field of computer vision or natural language processing, and the unique characteristics of a traffic evolution mechanism can not be effectively coupled in the modeling and learning process of a traffic flow spatio-temporal process, so that the intelligent models are easily misled by data noise and incomplete problems, and the accuracy and interpretability of prediction results are obviously influenced. The concrete points are as follows:
in the aspect of modeling spatial dependence of a traffic flow evolution process, an existing intelligent prediction model considers inherent limitation of a convolutional neural network or graph convolutional neural network structure, the spatial dependence of the traffic flow is mostly abstracted into an undirected graph or a picture with attributes according to a road network structure or spatial layout, and inherent property structures of the traffic flow are greatly lost;
in the aspect of modeling time dependence of a traffic flow evolution process, the conventional intelligent prediction model is mostly based on an RNN structure, learns through data of adjacent time segments, ignores trend and periodicity modeling of a traffic flow space-time process, and has certain bias.
The embodiment of the disclosure provides an urban road traffic flow attribute prediction method, which can be applied to an urban road traffic flow prediction process of an intelligent traffic scene.
Referring to fig. 1, a schematic flow chart of a method for predicting an urban road traffic flow attribute according to an embodiment of the present disclosure is shown. As shown in fig. 1 and 2, the method mainly includes the following steps:
step 1, collecting urban road network data, vehicle GPS data and road section traffic flow data;
for example, the implementation process of the present invention may be explained based on the total taxi GPS track data between 1 month 1 day 2012 and 2 month 29 2012 of the city a and the core segment 672 road segments of the lake region as the actual data set of the specific invention implementation. The traffic flow attribute is road section passing speed calculated by GPS track data, and the interval of time stamps is 15 minutes.
Step 2, constructing the urban road network into a directed graph structure with the right
Figure 188573DEST_PATH_IMAGE001
Wherein
Figure 9899DEST_PATH_IMAGE002
Representing cities for node sets
Figure 624551DEST_PATH_IMAGE003
A collection of road segments to be joined together,
Figure 531327DEST_PATH_IMAGE004
the weight matrix represents the strength of the correlation between every two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 732370DEST_PATH_IMAGE005
representing the traffic flow in the traffic flow attribute matrix
Figure 673781DEST_PATH_IMAGE006
Is within a time stamp
Figure 14327DEST_PATH_IMAGE003
Attribute information on the road segment;
in specific implementation, the urban road network can be constructed into a weighted directed graph structure
Figure 408400DEST_PATH_IMAGE118
Which isIn
Figure 304811DEST_PATH_IMAGE119
Representing cities for node sets
Figure 835150DEST_PATH_IMAGE003
A collection of road segments.
Figure 962244DEST_PATH_IMAGE004
The weight matrix represents the strength of the correlation between every two road segments in the road network and also represents the spatial dependency relationship of the traffic flow in the road network.
Figure 312454DEST_PATH_IMAGE005
For the traffic flow attribute matrix, the traffic flow is represented
Figure 809294DEST_PATH_IMAGE006
Is within a time stamp
Figure 928560DEST_PATH_IMAGE003
Attribute information on the road segment.
Step 3, every road section in the road network
Figure 852653DEST_PATH_IMAGE120
As one state, its entire road network structure
Figure 267323DEST_PATH_IMAGE008
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
further, the step 3 specifically includes:
step 3.1, for the binary road section pair formed by all road sections in the road network
Figure 567854DEST_PATH_IMAGE017
And calculating the times of occurrence of the GPS track and recording the times as
Figure 40162DEST_PATH_IMAGE018
And are combinedForm a matrix
Figure 338419DEST_PATH_IMAGE019
：
Figure 194380DEST_PATH_IMAGE020
；
Step 3.2, computing Markov chain transfer matrix by maximum likelihood method
Figure 501864DEST_PATH_IMAGE022
；
Step 3.3, assigning the value of the Markov transfer matrix to the weight matrix of the road network
Figure 861301DEST_PATH_IMAGE023
And completing the modeling of the spatial dependency.
When the method is specifically implemented, each road section in the road network is divided into two sections
Figure 100434DEST_PATH_IMAGE120
As one state, its entire road network structure
Figure 443690DEST_PATH_IMAGE121
Forming a state space, wherein the spatial dependency expression of traffic flow on a road network is a random process in the state space, and the modeling is carried out by a Markov chain mode, and the method mainly comprises the following steps:
3.1) binary road segment pairs for all road segments in the road network
Figure 86024DEST_PATH_IMAGE017
And calculating the times of occurrence of the GPS track and recording the times as
Figure 565547DEST_PATH_IMAGE018
And form a matrix
Figure 205607DEST_PATH_IMAGE019
：
Figure 770581DEST_PATH_IMAGE123
3.2) further computing the transition matrix of the Markov chain by means of the maximum likelihood method
Figure 715141DEST_PATH_IMAGE124
I.e. by
Figure 783591DEST_PATH_IMAGE022
；
3.3) assigning the values of the Markov transfer matrix to the weight matrix of the road network, i.e.
Figure 391290DEST_PATH_IMAGE023
. Modeling of spatial dependencies is done.
Step 4, transforming the conventional spectrogram convolution structure to ensure that the conventional spectrogram convolution structure is modified
Figure 443559DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 162117DEST_PATH_IMAGE008
Learning and fitting spatial dependence characteristics of traffic flow;
on the basis of the above embodiment, the step 4 specifically includes:
step 4.1, to the weight matrix
Figure 85073DEST_PATH_IMAGE009
Performing characteristic decomposition to obtain the final product by the following formula
Figure 863673DEST_PATH_IMAGE009
Maximum eigenvalue
Figure 839457DEST_PATH_IMAGE024
Corresponding feature vector
Figure 158443DEST_PATH_IMAGE025
：
Figure 467065DEST_PATH_IMAGE026
；
Step 4.2, define the matrix
Figure 619829DEST_PATH_IMAGE027
Is a diagonal matrix whose diagonal elements are formed by feature vectors
Figure 381111DEST_PATH_IMAGE028
Is formed by
Figure 674427DEST_PATH_IMAGE029
；
Step 4.3, calculating the weighted directed graph
Figure 103134DEST_PATH_IMAGE031
Laplacian matrix of
Figure 426799DEST_PATH_IMAGE032
：
Figure 940957DEST_PATH_IMAGE125
；
Step 4.4, the Prasise matrix is divided
Figure 805008DEST_PATH_IMAGE035
Carrying out symmetry to form a symmetrical Laplace matrix
Figure 88222DEST_PATH_IMAGE036
：
Figure 645105DEST_PATH_IMAGE126
；
Step 4.5, based on
Figure 569197DEST_PATH_IMAGE039
Defining weighted direction
Figure 502518DEST_PATH_IMAGE040
The convolution operator of
Figure 843501DEST_PATH_IMAGE041
The convolution operator acts on the traffic flow attribute matrix
Figure 571285DEST_PATH_IMAGE042
Slicing at a certain time
Figure 794456DEST_PATH_IMAGE043
Is operated as
Figure 498845DEST_PATH_IMAGE044
Wherein, in the step (A),
Figure 491072DEST_PATH_IMAGE045
is a symmetric Laplace matrix
Figure 593020DEST_PATH_IMAGE046
A matrix of feature vectors is formed from the feature vectors,
Figure 37908DEST_PATH_IMAGE047
the product of the Hadamard is used as the target,
Figure 844190DEST_PATH_IMAGE048
in order to calculate the operators for the convolution,
Figure 159764DEST_PATH_IMAGE049
parameters which can be learned in the convolution operator;
step 4.6, simplification by utilizing Chebyshev polynomial
Figure 196728DEST_PATH_IMAGE050
Comprises the following steps:
Figure 394491DEST_PATH_IMAGE128
；
step 4.7, based on the weighted directed graph in step 4.6
Figure 942147DEST_PATH_IMAGE008
The above convolution formula is used for calculating the traffic flow attribute matrix
Figure 220551DEST_PATH_IMAGE052
Convolution operation above:
Figure 195460DEST_PATH_IMAGE129
wherein the content of the first and second substances,
Figure 614940DEST_PATH_IMAGE054
represents the network parameters of the convolutional layer, and
Figure 966287DEST_PATH_IMAGE055
and
Figure 787613DEST_PATH_IMAGE056
representing the dimensions of the convolutional layer input features and the dimensions of the output features, respectively.
In specific implementation, the convolution structure of the traditional spectrogram can be modified to ensure that the traditional spectrogram can be used for
Figure 199002DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 105779DEST_PATH_IMAGE121
The effective learning and fitting traffic flow space dependency characteristics mainly comprise the following steps:
4.1) Pair weight matrix
Figure 791975DEST_PATH_IMAGE009
Performing characteristic decomposition to obtain the product by formula (2)
Figure 237781DEST_PATH_IMAGE009
Maximum eigenvalue
Figure 554492DEST_PATH_IMAGE130
Corresponding feature vector
Figure 948565DEST_PATH_IMAGE025
。
Figure 313818DEST_PATH_IMAGE131
4.2) definition matrix
Figure 375315DEST_PATH_IMAGE132
Is a diagonal matrix whose diagonal elements are formed by feature vectors
Figure 830305DEST_PATH_IMAGE025
Is formed by
Figure 977252DEST_PATH_IMAGE133
。
4.3) on the basis, calculating a weighted directed graph through the formula (3)
Figure 208513DEST_PATH_IMAGE121
Laplacian matrix of
Figure 593358DEST_PATH_IMAGE134
：
Figure 783031DEST_PATH_IMAGE136
Wherein the content of the first and second substances,
Figure 151696DEST_PATH_IMAGE137
representing an identity matrix.
4.4) further, the Laplace matrix can be divided into
Figure 921069DEST_PATH_IMAGE134
Carrying out symmetry to form a symmetrical Laplace matrix by the formula (4)
Figure 425999DEST_PATH_IMAGE138
：
Figure 285109DEST_PATH_IMAGE140
Its Laplace matrix
Figure 875490DEST_PATH_IMAGE138
The visualization results are shown in fig. 3, i.e. the symmetric laplacian matrix in example step 4.4. Wherein, the horizontal and vertical coordinates respectively represent each road section in the embodiment
Figure 448554DEST_PATH_IMAGE142
Number of
Figure 73570DEST_PATH_IMAGE144
Each color block in the graph represents its laplacian matrix
Figure 136204DEST_PATH_IMAGE138
The value of the corresponding element also indicates the strength of the correlation between the corresponding road sections.
4.5) based on
Figure 213882DEST_PATH_IMAGE138
Defining weighted direction
Figure 856215DEST_PATH_IMAGE121
Above convolution operator of
Figure 335738DEST_PATH_IMAGE145
The convolution operator acts on the traffic flow attribute matrix
Figure 506957DEST_PATH_IMAGE042
Slicing at a certain time
Figure 570465DEST_PATH_IMAGE043
Is operated as
Figure 954173DEST_PATH_IMAGE044
. Wherein the content of the first and second substances,
Figure 553782DEST_PATH_IMAGE045
is a symmetric Laplace matrix
Figure 161481DEST_PATH_IMAGE046
A matrix of feature vectors is formed, the feature vectors,
Figure 10488DEST_PATH_IMAGE047
the product of the Hadamard is used as the target,
Figure 729045DEST_PATH_IMAGE048
in order to be a convolution operator, the convolution operator,
Figure 183160DEST_PATH_IMAGE049
parameters which can be learned in the convolution operator;
4.6) borrowing Chebyshev polynomial simplifications similar to those in GCN
Figure 696181DEST_PATH_IMAGE146
Is a formula (5)
Figure 728423DEST_PATH_IMAGE148
4.7) finally, based on the weighted directed graph in step 4.6)
Figure 985092DEST_PATH_IMAGE121
The above convolution formula can be calculated in the traffic flow attribute matrix by the formula (5)
Figure 293714DEST_PATH_IMAGE042
The convolution operation above:
Figure 977636DEST_PATH_IMAGE150
（6）
wherein the content of the first and second substances,
Figure 112820DEST_PATH_IMAGE054
represents the network parameters of the convolutional layer, and
Figure 438759DEST_PATH_IMAGE055
and with
Figure 601887DEST_PATH_IMAGE056
Representing the dimensions of the convolutional layer input features and the dimensions of the output features, respectively, e.g. if the input is of formula (6)
Figure 722290DEST_PATH_IMAGE042
Then, then
Figure 236448DEST_PATH_IMAGE151
。
Figure 366078DEST_PATH_IMAGE152
Represents a non-linear activation function, here a ReLU function is chosen.
Step 5, based on the prior cognition of trend and periodicity of the traffic flow time-space process, building a traffic flow key frame sequence to carry out time-dependent modeling;
further, the step 5 specifically includes:
step 5.1, aiming at the periodic characteristics of the traffic flow and the predicted time
Figure 23193DEST_PATH_IMAGE057
Selecting
Figure 376814DEST_PATH_IMAGE057
Front part
Figure 315951DEST_PATH_IMAGE058
The traffic flow attributes at corresponding time in each period form a periodic frame sequence
Figure 718113DEST_PATH_IMAGE059
I.e. by
Figure 855834DEST_PATH_IMAGE060
Wherein
Figure 583618DEST_PATH_IMAGE061
Indicating the number of time stamps included in one period;
step 5.2, aiming at the trend characteristic of the traffic flow and the predicted time
Figure 39745DEST_PATH_IMAGE057
Selecting
Figure 511178DEST_PATH_IMAGE057
Front side
Figure 300142DEST_PATH_IMAGE062
Traffic flow attribute composition trend frame sequence at each moment
Figure 933249DEST_PATH_IMAGE063
I.e. by
Figure 846978DEST_PATH_IMAGE064
。
In specific implementation, based on the prior cognition of trend and periodicity of a traffic flow space-time process, the modeling of time dependence is realized by constructing a traffic flow key frame sequence, and the method mainly comprises the following steps:
5.1) periodic characteristics of the traffic flow, for the predicted time
Figure 856523DEST_PATH_IMAGE057
Selecting
Figure 499994DEST_PATH_IMAGE057
Front part
Figure 569581DEST_PATH_IMAGE058
The traffic flow attributes at corresponding time in each period form a periodic frame sequence
Figure 501765DEST_PATH_IMAGE059
I.e. by
Figure 22657DEST_PATH_IMAGE060
Wherein
Figure 989476DEST_PATH_IMAGE061
Indicating the number of timestamps contained in a cycle.
5.2) Trend characteristics of traffic flow, for the predicted time
Figure 964385DEST_PATH_IMAGE057
Selecting
Figure 649444DEST_PATH_IMAGE057
Front side
Figure 905851DEST_PATH_IMAGE062
Traffic flow attribute composition trend frame sequence at each moment
Figure 727176DEST_PATH_IMAGE063
I.e. by
Figure 138566DEST_PATH_IMAGE064
。
Step 6, after the modeling of the space dependency and the time dependency is completed, utilizing the weighted directed graph defined in the step 4
Figure 45342DEST_PATH_IMAGE040
Convolution operation on
Figure 934801DEST_PATH_IMAGE041
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer;
further, the step 6 specifically includes:
step 6.1, in the hidden layer, for periodic frame sequences
Figure 876212DEST_PATH_IMAGE065
Trending frame sequences
Figure 458503DEST_PATH_IMAGE063
Using two convolutions respectively
Figure 586996DEST_PATH_IMAGE066
And
Figure 44260DEST_PATH_IMAGE067
learning is carried out to respectively obtain corresponding hidden layer characteristics
Figure 981123DEST_PATH_IMAGE068
And
Figure 999895DEST_PATH_IMAGE069
the formula is as follows:
Figure 818946DEST_PATH_IMAGE153
Figure 79901DEST_PATH_IMAGE155
wherein
Figure 464746DEST_PATH_IMAGE074
And with
Figure 654419DEST_PATH_IMAGE075
Respectively learning periodic convolution
Figure 23083DEST_PATH_IMAGE076
Convolution with learning tendency
Figure 58035DEST_PATH_IMAGE077
The learnable parameter of (1);
step 6.2, hiding layer characteristics related to trends in the aggregation layer
Figure 297387DEST_PATH_IMAGE078
Periodicity dependent hidden layer features
Figure 126803DEST_PATH_IMAGE079
Splicing on characteristic dimension to form a matrix
Figure 233297DEST_PATH_IMAGE080
；
Step 6.3, the matrix is processed
Figure 212886DEST_PATH_IMAGE081
Is scrambled in the feature dimension to form
Figure 634640DEST_PATH_IMAGE082
；
Step 6.4, by a convolution
Figure 369377DEST_PATH_IMAGE083
To pair
Figure 181476DEST_PATH_IMAGE084
Performing deep feature extraction to form
Figure 56766DEST_PATH_IMAGE085
:
Figure 536288DEST_PATH_IMAGE156
；
Step 6.5, through a fully connected network, will
Figure 441928DEST_PATH_IMAGE088
Is compressed into 1-dimension to form
Figure 803639DEST_PATH_IMAGE089
As a model to
Figure 984084DEST_PATH_IMAGE090
At the moment of time, the time of day,
Figure 318114DEST_PATH_IMAGE091
a traffic-flow attribute prediction model of a traffic-flow attribute prediction result of a link section:
Figure 394654DEST_PATH_IMAGE157
wherein
Figure 414301DEST_PATH_IMAGE093
Is a learnable parameter of a fully connected network.
In specific implementation, after the modeling of the spatial dependency and the time dependency is completed, the weighted directed graph defined in the step 4 is utilized
Figure 132858DEST_PATH_IMAGE158
Convolution operation on
Figure 586973DEST_PATH_IMAGE159
The method for constructing the traffic flow attribute intelligent prediction model comprising the hidden layer, the aggregation layer and the output layer mainly comprises the following steps of:
6.1) in the hidden layer, for two types of key frame sequences, i.e. periodic frame sequences
Figure 365573DEST_PATH_IMAGE160
Trending frame sequences
Figure 639560DEST_PATH_IMAGE161
Using two convolutions respectively
Figure 427387DEST_PATH_IMAGE076
And
Figure 736009DEST_PATH_IMAGE077
learning is carried out to respectively obtain corresponding hidden layer characteristics
Figure 184045DEST_PATH_IMAGE068
And
Figure 210907DEST_PATH_IMAGE069
the formula is as follows:
Figure 271267DEST_PATH_IMAGE163
（7）
Figure 699974DEST_PATH_IMAGE165
（8）
wherein
Figure 23639DEST_PATH_IMAGE074
And
Figure 272218DEST_PATH_IMAGE075
respectively learning periodic convolution
Figure 136269DEST_PATH_IMAGE076
Convolution with learning tendency
Figure 923877DEST_PATH_IMAGE077
The learnable parameter of (1);
6.2) in the aggregate layer, trend-related hidden layer features are first identified
Figure 215181DEST_PATH_IMAGE078
Periodicity dependent hidden layer features
Figure 623160DEST_PATH_IMAGE079
Splicing on characteristic dimension to form a matrix
Figure 25322DEST_PATH_IMAGE080
。
6.3) then the matrix
Figure 163043DEST_PATH_IMAGE081
Are scrambled in a characteristic dimension to form
Figure 123783DEST_PATH_IMAGE082
。
6.4) and then convolving by a convolution
Figure 612533DEST_PATH_IMAGE083
To pair
Figure 83966DEST_PATH_IMAGE084
Performing deep feature extraction to form
Figure 279455DEST_PATH_IMAGE085
:
Figure 709299DEST_PATH_IMAGE167
（9）
Wherein the content of the first and second substances,
Figure 91870DEST_PATH_IMAGE168
is a convolution of
Figure 68791DEST_PATH_IMAGE169
Of the learning parameters.
6.5) finally, via a fully connected network, will
Figure 181104DEST_PATH_IMAGE088
Is compressed into 1-dimension to form
Figure 985112DEST_PATH_IMAGE089
As a model to
Figure 386137DEST_PATH_IMAGE090
At the moment of time, the time of day,
Figure 199373DEST_PATH_IMAGE091
and predicting the traffic flow attribute of the road segment, namely:
Figure 697350DEST_PATH_IMAGE171
（10）
wherein
Figure 639636DEST_PATH_IMAGE093
Is a learnable parameter of a fully connected network.
Step 7, calculating the real traffic flow attribute value on the training set
Figure 59116DEST_PATH_IMAGE113
Pertaining to traffic flowTraffic flow attribute value predicted by sexual prediction model
Figure 410463DEST_PATH_IMAGE114
Error between, updating learnable parameters in traffic flow attribute prediction model
Figure 231788DEST_PATH_IMAGE115
、
Figure 908757DEST_PATH_IMAGE100
、
Figure 549954DEST_PATH_IMAGE116
And with
Figure 704992DEST_PATH_IMAGE117
Until the traffic flow attribute prediction model has the lowest prediction error in the verification set;
on the basis of the above embodiment, the step 7 specifically includes:
step 7.1, defining real traffic flow attribute value by using root mean square error
Figure 380824DEST_PATH_IMAGE094
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 455791DEST_PATH_IMAGE095
An error of
Figure 849863DEST_PATH_IMAGE096
：
Figure 543013DEST_PATH_IMAGE172
；
Step 7.2, based on calculation
Figure 604510DEST_PATH_IMAGE098
Iteratively updating learnable parameters in traffic flow attribute prediction model by adopting SGD optimizer
Figure 826543DEST_PATH_IMAGE099
、
Figure 973491DEST_PATH_IMAGE100
、
Figure 470331DEST_PATH_IMAGE101
And
Figure 822553DEST_PATH_IMAGE102
until the traffic flow attribute prediction model has the lowest prediction error in the validation set.
When the method is implemented, the real traffic flow attribute value can be calculated on the training set
Figure 543384DEST_PATH_IMAGE094
Model predicted traffic flow attribute value
Figure 646470DEST_PATH_IMAGE095
Error therebetween to guide updating of learnable parameters in the model
Figure 681422DEST_PATH_IMAGE099
、
Figure 592877DEST_PATH_IMAGE100
、
Figure 937140DEST_PATH_IMAGE101
And
Figure 760477DEST_PATH_IMAGE102
so that the model has the lowest prediction error in the validation set, the method mainly comprises the following steps:
7.1) defining real traffic flow attribute value by borrowing root mean square error
Figure 67961DEST_PATH_IMAGE094
Model predicted traffic flow attribute value
Figure 161819DEST_PATH_IMAGE095
An error of
Figure 427716DEST_PATH_IMAGE096
：
Figure 505393DEST_PATH_IMAGE173
（11）
7.2) based on calculation
Figure 793067DEST_PATH_IMAGE096
Iteratively updating learnable parameters in the model using an SGD optimizer
Figure 475852DEST_PATH_IMAGE099
、
Figure 912650DEST_PATH_IMAGE100
、
Figure 477623DEST_PATH_IMAGE101
And
Figure 156604DEST_PATH_IMAGE102
until the model has the lowest prediction error in the validation set.
And 8, predicting the traffic flow attribute in the target scene by using the traffic flow attribute prediction model updated in the step 7.
In specific implementation, when the traffic flow attribute prediction model has the lowest prediction error in the verification set, which means that the prediction accuracy of the current traffic flow attribute prediction model is the highest, the updated traffic flow attribute prediction model can be used for predicting the traffic flow attribute in the target scene which actually needs to be predicted.
According to the urban road traffic flow attribute prediction method provided by the embodiment, by introducing the space-time prior knowledge of the traffic flow evolution process and pertinently transforming a neural network learning module, the model can be ensured to fit a real traffic flow space-time process mode under the influence of noise, specifically, the traffic flow space dependency of a road section level is modeled at a fine granularity by a Markov matrix and is abstracted to a weighted directed graph structure, so that the traditional spectrogram convolution is pertinently transformed, the traditional spectrogram convolution can be embedded into the weighted directed graph, and the space dependency is fitted in the learning process. In the aspect of time dependence fitting, a traffic flow key frame sequence is constructed by recognizing the trend and periodicity of the time evolution of the traffic flow, so that the fitting of the time dependence of the traffic flow under the driving of prior knowledge is realized, and the accuracy and the robustness of urban road traffic flow attribute prediction are improved.
The method of the embodiment of the present disclosure will be described with reference to a specific embodiment, and to verify the advancement of the present disclosure, a traffic flow prediction method of the mainstream at home and abroad is selected: HA. The ARIMA, VAR, SVR, FC-GRU, T-GCN and the method for predicting the accuracy are compared and analyzed, and the method mainly comprises the following steps:
9.1) to achieve a comprehensive accuracy comparison, a variety of accuracy evaluation indicators are selected, including the root mean square error
Figure 287371DEST_PATH_IMAGE096
Mean absolute value error
Figure 567174DEST_PATH_IMAGE174
And accuracy
Figure 822706DEST_PATH_IMAGE175
. Wherein, the first and the second end of the pipe are connected with each other,
Figure 774219DEST_PATH_IMAGE096
and
Figure 166017DEST_PATH_IMAGE174
the lower the value of (a), the better the performance of the model-representative prediction, and conversely,
Figure 413459DEST_PATH_IMAGE175
the higher the value of (a), the better the model behaves.
Figure 953025DEST_PATH_IMAGE177
Wherein the content of the first and second substances,
Figure 740852DEST_PATH_IMAGE178
representing the average of all traffic flow attribute truth values.
9.2) Table 1 shows the results of the accuracy comparisons between the method of the invention and the comparison method at different prediction step sizes (15 minutes, 30 minutes, 45 minutes into the future of the prediction):
TABLE 1
Figure 49474DEST_PATH_IMAGE179
It can be clearly found that the method of the present invention has a significant advantage in prediction accuracy compared to the comparative method.
In correspondence with the above method embodiment, referring to fig. 4, an embodiment of the present disclosure further provides an urban road traffic flow attribute prediction system 40, including:
the acquisition module 401 is used for acquiring urban road network data, vehicle GPS data and road section traffic flow data;
a first constructing module 402, configured to construct the urban road network into a weighted directed graph structure
Figure 497510DEST_PATH_IMAGE103
Wherein
Figure 524372DEST_PATH_IMAGE104
Representing cities for node sets
Figure 584732DEST_PATH_IMAGE105
A collection of road segments to be joined together,
Figure 747860DEST_PATH_IMAGE106
the weight matrix represents the strength of the correlation between every two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 337104DEST_PATH_IMAGE107
attribute moments for traffic flowArray, representing the flow of traffic
Figure 585683DEST_PATH_IMAGE108
Is within a time stamp
Figure 449734DEST_PATH_IMAGE109
Attribute information on the road segment;
a first modeling module 403 for modeling each road segment in the road network
Figure 249061DEST_PATH_IMAGE110
As one state, its entire road network structure
Figure 274786DEST_PATH_IMAGE111
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
a reconstruction module 404 for reconstructing the conventional spectrogram convolution structure
Figure 10661DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 350506DEST_PATH_IMAGE008
Learning and fitting spatial dependence characteristics of traffic flow;
a second modeling module 405, configured to perform time-dependent modeling by constructing a traffic flow key frame sequence based on prior knowledge of trend and periodicity of a traffic flow spatio-temporal process;
a second building module 406, configured to utilize the weighted directed graph defined in step 4 after completing the modeling of the spatial dependency and the temporal dependency
Figure 488227DEST_PATH_IMAGE111
Convolution operation on
Figure 714546DEST_PATH_IMAGE112
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer;
trainingA module 407 for calculating a true traffic flow attribute value by computing a true traffic flow attribute value on a training set
Figure 875400DEST_PATH_IMAGE113
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 81254DEST_PATH_IMAGE114
Error between, updating learnable parameters in traffic flow attribute prediction model
Figure 73481DEST_PATH_IMAGE115
、
Figure 975096DEST_PATH_IMAGE100
、
Figure 768785DEST_PATH_IMAGE116
And
Figure 716012DEST_PATH_IMAGE117
until the traffic flow attribute prediction model has the lowest prediction error in the verification set;
and a prediction module 408, configured to predict the traffic flow attribute in the target scene by using the traffic flow attribute prediction model updated in step 7.
The system shown in fig. 4 can correspondingly execute the content in the above method embodiment, and details of the part not described in detail in this embodiment refer to the content described in the above method embodiment, which is not described again here.
Referring to fig. 5, an embodiment of the present disclosure also provides an electronic device 50, including: at least one processor and a memory communicatively coupled to the at least one processor. Wherein the memory stores instructions executable by the at least one processor, the instructions being executable by the at least one processor to enable the at least one processor to perform the urban road traffic flow attribute prediction method in the aforementioned method embodiments.
The disclosed embodiments also provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the urban road traffic flow attribute prediction method in the foregoing method embodiments.
The disclosed embodiments also provide a computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the urban road traffic flow attribute prediction method in the aforementioned method embodiments.
Referring now to FIG. 5, a schematic diagram of an electronic device 50 suitable for use in implementing embodiments of the present disclosure is shown. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., car navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 5 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 5, electronic device 50 may include a processing means (e.g., central processing unit, graphics processor, etc.) 501 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM) 502 or a program loaded from a storage means 508 into a Random Access Memory (RAM) 503. In the RAM 503, various programs and data necessary for the operation of the electronic apparatus 50 are also stored. The processing device 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input/output (I/O) interface 505 is also connected to bus 504.
Generally, the following devices may be connected to the I/O interface 505: input devices 506 including, for example, a touch screen, touch pad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 507 including, for example, a Liquid Crystal Display (LCD), speakers, vibrators, and the like; storage devices 508 including, for example, magnetic tape, hard disk, etc.; and a communication device 509. The communication means 509 may allow the electronic device 50 to communicate with other devices wirelessly or by wire to exchange data. While the figures illustrate an electronic device 50 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication means 509, or installed from the storage means 508, or installed from the ROM 502. The computer program performs the above-described functions defined in the methods of the embodiments of the present disclosure when executed by the processing device 501.
It should be noted that the computer readable medium in the present disclosure can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In contrast, in the present disclosure, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, either in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device.
The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform the steps associated with the method embodiments.
Alternatively, the computer readable medium carries one or more programs which, when executed by the electronic device, enable the electronic device to perform the steps associated with the method embodiments.
Computer program code for carrying out operations for aspects of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present disclosure may be implemented by software or hardware.
It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.
The above description is only for the specific embodiments of the present disclosure, but the scope of the present disclosure is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present disclosure should be covered within the scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims (9)

1. An urban road traffic flow attribute prediction method is characterized by comprising the following steps:
step 1, collecting urban road network data, vehicle GPS data and road section traffic flow data;
in the step 2, the step of mixing the raw materials,constructing urban road network into a directed graph structure with rights
Figure 490673DEST_PATH_IMAGE001
Wherein
Figure 781715DEST_PATH_IMAGE002
Representing cities for node sets
Figure 499135DEST_PATH_IMAGE003
A collection of road segments to be joined together,
Figure 832028DEST_PATH_IMAGE004
is a weight matrix which represents the correlation strength of two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 584083DEST_PATH_IMAGE005
representing the traffic flow in the traffic flow attribute matrix
Figure 406546DEST_PATH_IMAGE006
Within a time stamp
Figure 673579DEST_PATH_IMAGE003
Attribute information on the road segment;
step 3, every road section in the road network
Figure 308697DEST_PATH_IMAGE007
As one state, its entire road network structure
Figure 977576DEST_PATH_IMAGE008
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
step 4, transforming the conventional spectrogram convolution structure to ensure that the conventional spectrogram convolution structure is modified
Figure 970940DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 662952DEST_PATH_IMAGE008
Learning and fitting spatial dependence characteristics of traffic flow;
step 5, based on the prior cognition of trend and periodicity of the traffic flow time-space process, building a traffic flow key frame sequence to carry out time-dependent modeling;
step 6, after the modeling of the space dependency and the time dependency is completed, utilizing the weighted directed graph defined in the step 4
Figure 399964DEST_PATH_IMAGE008
Convolution operation on
Figure 861032DEST_PATH_IMAGE010
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer;
step 7, calculating the real traffic flow attribute value on the training set
Figure 290877DEST_PATH_IMAGE011
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 532502DEST_PATH_IMAGE012
Error between, updating learnable parameters in traffic flow attribute prediction model
Figure 745309DEST_PATH_IMAGE013
、
Figure 824998DEST_PATH_IMAGE014
、
Figure 160164DEST_PATH_IMAGE015
And
Figure 889086DEST_PATH_IMAGE016
until the traffic flow attribute prediction model has the lowest prediction error in the verification set;
and 8, predicting the traffic flow attribute in the target scene by using the traffic flow attribute prediction model updated in the step 7.
2. The method according to claim 1, wherein step 3 specifically comprises:
step 3.1, for the binary road section pair formed by all road sections in the road network
Figure 905584DEST_PATH_IMAGE017
And calculating the times of occurrence of the GPS track and recording the times as
Figure 403561DEST_PATH_IMAGE018
And form a matrix
Figure 112891DEST_PATH_IMAGE019
：
Figure 63530DEST_PATH_IMAGE020
；
Step 3.2, computing Markov chain transfer matrix by maximum likelihood method
Figure 946035DEST_PATH_IMAGE021
；
Step 3.3, assigning the value of the Markov transfer matrix to the weight matrix of the road network
Figure 236202DEST_PATH_IMAGE022
And completing the modeling of the spatial dependency.
3. The method according to claim 2, wherein the step 4 specifically comprises:
step 4.1, to the weight matrix
Figure 178750DEST_PATH_IMAGE009
Performing characteristic decomposition to obtain the final product by the following formula
Figure 52903DEST_PATH_IMAGE009
Maximum eigenvalue
Figure 473520DEST_PATH_IMAGE023
Corresponding feature vector
Figure 883773DEST_PATH_IMAGE024
：
Figure 997222DEST_PATH_IMAGE025
；
Step 4.2, define the matrix
Figure 922453DEST_PATH_IMAGE026
Is a diagonal matrix whose diagonal elements are formed by feature vectors
Figure 84444DEST_PATH_IMAGE027
Is formed by
Figure 145941DEST_PATH_IMAGE028
；
Step 4.3, calculating the weighted directed graph
Figure 430292DEST_PATH_IMAGE029
Laplacian matrix of
Figure 780502DEST_PATH_IMAGE030
：
Figure 808501DEST_PATH_IMAGE031
；
Step (ii) of4.4, the Laplace matrix
Figure 426301DEST_PATH_IMAGE032
Carrying out symmetry to form a symmetrical Laplace matrix
Figure 615974DEST_PATH_IMAGE033
：
Figure 515797DEST_PATH_IMAGE034
；
Step 4.5, based on
Figure 19591DEST_PATH_IMAGE035
Defining weighted direction
Figure 258942DEST_PATH_IMAGE036
Above convolution operator of
Figure 619516DEST_PATH_IMAGE037
The convolution operator of which acts on the traffic flow attribute matrix
Figure 944319DEST_PATH_IMAGE038
Slicing at a certain time
Figure 314120DEST_PATH_IMAGE039
Is operated as a convolution of
Figure 204716DEST_PATH_IMAGE040
Wherein, in the step (A),
Figure 437989DEST_PATH_IMAGE041
is a symmetric Laplace matrix
Figure 46825DEST_PATH_IMAGE042
A matrix of feature vectors is formed from the feature vectors,
Figure 158000DEST_PATH_IMAGE043
the product of the Hadamard is used as the target,
Figure 168681DEST_PATH_IMAGE044
in order to calculate the operators for the convolution,
Figure 136637DEST_PATH_IMAGE045
parameters which can be learned in the convolution operator;
step 4.6, simplification by utilizing Chebyshev polynomial
Figure 170452DEST_PATH_IMAGE046
Comprises the following steps:
Figure 882056DEST_PATH_IMAGE047
；
step 4.7, based on the weighted directed graph in step 4.6
Figure 747244DEST_PATH_IMAGE008
The above convolution formula is used for calculating the traffic flow attribute matrix
Figure 260003DEST_PATH_IMAGE048
The convolution operation above:
Figure 843431DEST_PATH_IMAGE049
wherein the content of the first and second substances,
Figure 93147DEST_PATH_IMAGE050
represents the network parameters of the convolutional layer, and
Figure 16103DEST_PATH_IMAGE051
and
Figure 325862DEST_PATH_IMAGE052
representing the dimensions of the convolutional layer input features and the dimensions of the output features, respectively.
4. The method according to claim 3, wherein the step 5 specifically comprises:
step 5.1, aiming at the periodic characteristics of the traffic flow and the predicted time
Figure 334269DEST_PATH_IMAGE053
Selecting
Figure 387676DEST_PATH_IMAGE053
Front side
Figure 227456DEST_PATH_IMAGE054
The traffic flow attributes at corresponding time in each period form a periodic frame sequence
Figure 645799DEST_PATH_IMAGE055
I.e. by
Figure 938240DEST_PATH_IMAGE056
Wherein
Figure 231556DEST_PATH_IMAGE057
Indicating the number of time stamps included in one period;
step 5.2, aiming at the trend characteristics of the traffic flow and the predicted time
Figure 925842DEST_PATH_IMAGE053
Selecting
Figure 577404DEST_PATH_IMAGE053
Front side
Figure 560403DEST_PATH_IMAGE058
Traffic flow attribute composition trend frame sequence at each moment
Figure 955612DEST_PATH_IMAGE059
I.e. by
Figure 442089DEST_PATH_IMAGE060
。
5. The method according to claim 4, wherein the step 6 specifically comprises:
step 6.1, in the hidden layer, for periodic frame sequences
Figure 530130DEST_PATH_IMAGE061
Trending frame sequences
Figure 797164DEST_PATH_IMAGE059
Using two convolutions respectively
Figure 933747DEST_PATH_IMAGE062
And with
Figure 773265DEST_PATH_IMAGE063
Learning to obtain corresponding hidden layer characteristics
Figure 32208DEST_PATH_IMAGE064
And
Figure 724220DEST_PATH_IMAGE065
the formula is as follows:
Figure 461232DEST_PATH_IMAGE066
Figure 859983DEST_PATH_IMAGE067
wherein
Figure 726046DEST_PATH_IMAGE068
And
Figure 577458DEST_PATH_IMAGE069
respectively learning periodic convolution
Figure 118161DEST_PATH_IMAGE070
Convolution with learning tendency
Figure 699315DEST_PATH_IMAGE071
The learnable parameter of (1);
step 6.2, hiding layer characteristics related to trends in the aggregation layer
Figure 34482DEST_PATH_IMAGE072
Periodicity dependent hidden layer features
Figure 223059DEST_PATH_IMAGE073
Splicing on characteristic dimension to form a matrix
Figure 301873DEST_PATH_IMAGE074
；
Step 6.3, matrix
Figure 799851DEST_PATH_IMAGE075
Is scrambled in the feature dimension to form
Figure 243602DEST_PATH_IMAGE076
；
Step 6.4, by a convolution
Figure 459819DEST_PATH_IMAGE077
To pair
Figure 217691DEST_PATH_IMAGE078
Performing deep feature extraction to form
Figure 570175DEST_PATH_IMAGE079
:
Figure 948941DEST_PATH_IMAGE080
；
Step 6.5, through a fully connected network, will
Figure 386876DEST_PATH_IMAGE081
Is compressed into 1-dimension to form
Figure 807493DEST_PATH_IMAGE082
As a model to
Figure 217746DEST_PATH_IMAGE083
At the moment of time, the time of day,
Figure 3299DEST_PATH_IMAGE084
a traffic flow attribute prediction model of a traffic flow attribute prediction result of a link:
Figure 928530DEST_PATH_IMAGE085
wherein
Figure 152838DEST_PATH_IMAGE086
Is a learnable parameter of a fully connected network.
6. The method according to claim 5, wherein the step 7 specifically comprises:
step 7.1, defining real traffic flow attribute value by using root mean square error
Figure 417597DEST_PATH_IMAGE087
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 701948DEST_PATH_IMAGE088
An error of
Figure 550693DEST_PATH_IMAGE089
：
Figure 313112DEST_PATH_IMAGE090
；
Step 7.2, based on calculation
Figure 494695DEST_PATH_IMAGE091
Iteratively updating learnable parameters in traffic flow attribute prediction model by adopting SGD optimizer
Figure 887630DEST_PATH_IMAGE092
、
Figure 787453DEST_PATH_IMAGE093
、
Figure 291247DEST_PATH_IMAGE094
And
Figure 265019DEST_PATH_IMAGE095
until the traffic flow attribute prediction model has the lowest prediction error in the validation set.
7. An urban road traffic flow attribute prediction system characterized by comprising:
the acquisition module is used for acquiring urban road network data, vehicle GPS data and road section traffic flow data;
a first constructing module for constructing the urban road network into a weighted directed graph structure
Figure 891173DEST_PATH_IMAGE096
Wherein
Figure 714510DEST_PATH_IMAGE097
Representing cities for node sets
Figure 818732DEST_PATH_IMAGE098
A collection of road segments that are to be joined,
Figure 974907DEST_PATH_IMAGE099
the weight matrix represents the strength of the correlation between every two road segments in the road network and represents the spatial dependency relationship of the traffic flow in the road network,
Figure 709645DEST_PATH_IMAGE100
representing the traffic flow in the traffic flow attribute matrix
Figure 256164DEST_PATH_IMAGE101
Within a time stamp
Figure 429656DEST_PATH_IMAGE102
Attribute information on the road segment;
a first modeling module for modeling each road segment in the road network
Figure 378020DEST_PATH_IMAGE103
As one state, the whole road network structure
Figure 345976DEST_PATH_IMAGE104
Forming a state space, and carrying out modeling of spatial dependence in a Markov chain mode;
a transformation module for transforming the conventional spectrogram convolution structure
Figure 176529DEST_PATH_IMAGE009
Weighted directed graph for edge weights
Figure 199718DEST_PATH_IMAGE008
Learning and fitting spatial dependence characteristics of traffic flow;
the second modeling module is used for modeling the time dependence by constructing a traffic flow key frame sequence based on the prior cognition of the trend and the periodicity of the traffic flow time-space process;
a second construction module, configured to utilize the weighted directed graph defined in step 4 after completing modeling of the spatial dependency and the time dependency
Figure 330485DEST_PATH_IMAGE104
Convolution operation on
Figure 407025DEST_PATH_IMAGE105
Constructing a traffic flow attribute prediction model comprising a hidden layer, a polymerization layer and an output layer;
a training module for calculating real traffic flow attribute value on the training set
Figure 990453DEST_PATH_IMAGE106
Traffic flow attribute value predicted by traffic flow attribute prediction model
Figure 676387DEST_PATH_IMAGE107
Error between, updating learnable parameters in traffic flow attribute prediction model
Figure 396082DEST_PATH_IMAGE108
、
Figure 705840DEST_PATH_IMAGE093
、
Figure 917510DEST_PATH_IMAGE109
And
Figure 908600DEST_PATH_IMAGE110
until the traffic flow attribute prediction model has the lowest prediction error in the verification set;
and a prediction module, configured to predict the traffic flow attribute in the target scene using the traffic flow attribute prediction model updated in step 7.
8. An electronic device, characterized in that the electronic device comprises:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the urban road traffic flow attribute prediction method of any one of the preceding claims 1-6.
9. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the urban road traffic flow attribute prediction method according to any one of claims 1 to 6.
CN202210915226.9A 2022-08-01 2022-08-01 Method, system, equipment and medium for predicting urban road traffic flow attribute Active CN114971093B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210915226.9A CN114971093B (en) 2022-08-01 2022-08-01 Method, system, equipment and medium for predicting urban road traffic flow attribute

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210915226.9A CN114971093B (en) 2022-08-01 2022-08-01 Method, system, equipment and medium for predicting urban road traffic flow attribute

Publications (2)

Publication Number Publication Date
CN114971093A true CN114971093A (en) 2022-08-30
CN114971093B CN114971093B (en) 2022-11-25

Family

ID=82968863

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210915226.9A Active CN114971093B (en) 2022-08-01 2022-08-01 Method, system, equipment and medium for predicting urban road traffic flow attribute

Country Status (1)

Country Link
CN (1) CN114971093B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116029459A (en) * 2023-02-28 2023-04-28 速度时空信息科技股份有限公司 Extraction method of TMGCN traffic flow prediction model combined with graph Fourier transform
CN117292551A (en) * 2023-11-27 2023-12-26 辽宁邮电规划设计院有限公司 Urban traffic situation adjustment system and method based on Internet of Things

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109754605A (en) * 2019-02-27 2019-05-14 中南大学 A kind of traffic forecast method based on attention temporal diagram convolutional network
CN112085123A (en) * 2020-09-25 2020-12-15 北方民族大学 Point cloud data classification and segmentation method based on salient point sampling
CN113158738A (en) * 2021-01-28 2021-07-23 中南大学 Port environment target detection method, system, terminal and readable storage medium based on attention mechanism
CN113487061A (en) * 2021-05-28 2021-10-08 山西云时代智慧城市技术发展有限公司 Long-time-sequence traffic flow prediction method based on graph convolution-Informer model
AU2021106420A4 (en) * 2021-08-22 2021-12-09 A, Arun DR Time and topology structure based traffic flow prediction with artificial intelligence and neural network
CN114037889A (en) * 2021-11-12 2022-02-11 泰康保险集团股份有限公司 Image identification method and device, electronic equipment and storage medium

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109754605A (en) * 2019-02-27 2019-05-14 中南大学 A kind of traffic forecast method based on attention temporal diagram convolutional network
CN112085123A (en) * 2020-09-25 2020-12-15 北方民族大学 Point cloud data classification and segmentation method based on salient point sampling
CN113158738A (en) * 2021-01-28 2021-07-23 中南大学 Port environment target detection method, system, terminal and readable storage medium based on attention mechanism
CN113487061A (en) * 2021-05-28 2021-10-08 山西云时代智慧城市技术发展有限公司 Long-time-sequence traffic flow prediction method based on graph convolution-Informer model
AU2021106420A4 (en) * 2021-08-22 2021-12-09 A, Arun DR Time and topology structure based traffic flow prediction with artificial intelligence and neural network
CN114037889A (en) * 2021-11-12 2022-02-11 泰康保险集团股份有限公司 Image identification method and device, electronic equipment and storage medium

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
KAIQI CHEN等: "A Temporal Directed Graph Convolution Network for Traffic Forecasting Using Taxi Trajectory Data", 《ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116029459A (en) * 2023-02-28 2023-04-28 速度时空信息科技股份有限公司 Extraction method of TMGCN traffic flow prediction model combined with graph Fourier transform
CN117292551A (en) * 2023-11-27 2023-12-26 辽宁邮电规划设计院有限公司 Urban traffic situation adjustment system and method based on Internet of Things
CN117292551B (en) * 2023-11-27 2024-02-23 辽宁邮电规划设计院有限公司 Urban traffic situation adjustment system and method based on Internet of things

Also Published As

Publication number Publication date
CN114971093B (en) 2022-11-25

Similar Documents

Publication Publication Date Title
CN114971093B (en) Method, system, equipment and medium for predicting urban road traffic flow attribute
JP7262503B2 (en) Method and apparatus, electronic device, computer readable storage medium and computer program for detecting small targets
CN108805348B (en) Method and device for controlling and optimizing intersection signal timing
Wilkie et al. Flow reconstruction for data-driven traffic animation
CN112863180B (en) Traffic speed prediction method, device, electronic equipment and computer readable medium
WO2023273611A1 (en) Speech recognition model training method and apparatus, speech recognition method and apparatus, medium, and device
CN116795720A (en) Unmanned driving system credibility evaluation method and device based on scene
CN110555861B (en) Optical flow calculation method and device and electronic equipment
CN115359663B (en) Mountain road disaster section disaster-resistant toughness calculation method and device and electronic equipment
CN115543638A (en) Uncertainty-based edge calculation data collection and analysis method, system and equipment
CN115773744A (en) Model training and road network processing method, device, equipment, medium and product
CN113111860B (en) Road mobile source emission calculation method, device, equipment and medium
CN111582456B (en) Method, apparatus, device and medium for generating network model information
CN111581455B (en) Text generation model generation method and device and electronic equipment
CN113516315B (en) Wind power generation power interval prediction method, device and medium
CN113723712B (en) Wind power prediction method, system, equipment and medium
CN111986243A (en) Road shoulder extraction method and device, electronic equipment and computer readable medium
CN115601960B (en) Multi-mode traffic flow prediction method and system based on graph comparison learning
CN116167894B (en) Water resource shortage warning method, device, electronic equipment and computer readable medium
CN116032985B (en) Uniform channel changing method, system, equipment and medium based on intelligent network-connected vehicle
CN111738416B (en) Model synchronous updating method and device and electronic equipment
CN116934557B (en) Behavior prediction information generation method, device, electronic equipment and readable medium
CN111582482B (en) Method, apparatus, device and medium for generating network model information
CN112069635B (en) Method and device for deploying battery changing cabinet, medium and electronic equipment
CN115527163A (en) Target detection model optimization method and device, electronic equipment and readable storage medium

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