CN116151493A - Traffic jam prediction method and device based on head effect and cyclic neural network - Google Patents
Traffic jam prediction method and device based on head effect and cyclic neural network Download PDFInfo
- Publication number
- CN116151493A CN116151493A CN202310443476.1A CN202310443476A CN116151493A CN 116151493 A CN116151493 A CN 116151493A CN 202310443476 A CN202310443476 A CN 202310443476A CN 116151493 A CN116151493 A CN 116151493A
- Authority
- CN
- China
- Prior art keywords
- behavior
- track
- access
- intention
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000000694 effects Effects 0.000 title claims abstract description 56
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 title claims abstract description 38
- 125000004122 cyclic group Chemical group 0.000 title claims abstract description 18
- 230000006399 behavior Effects 0.000 claims abstract description 218
- 238000010586 diagram Methods 0.000 claims abstract description 52
- 239000013598 vector Substances 0.000 claims abstract description 33
- 238000005065 mining Methods 0.000 claims abstract description 11
- 238000004140 cleaning Methods 0.000 claims abstract description 9
- 230000000306 recurrent effect Effects 0.000 claims description 30
- 238000010276 construction Methods 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 9
- 238000012732 spatial analysis Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 description 7
- 230000009471 action Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000033001 locomotion Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000029305 taxis Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
-
- 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"
-
- G06Q50/40—
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a traffic jam prediction method and a device based on head effect and a cyclic neural network, belonging to the field of traffic prediction, and comprising the following steps: acquiring vehicle track point sequence data and sequentially cleaning and road network matching to obtain effective track point sequence data; constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data; mining real intention track data in the effective track point sequence data based on the head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram; embedding the intention enhancement behavior diagram to obtain an embedded vector; and predicting the access behaviors corresponding to the track points at the future moment by adopting a gating cyclic neural network according to the embedded vector, then carrying out density clustering on the access behaviors, and predicting the traffic jam condition according to the clustering result. The method and the device improve the prediction accuracy of the traffic jam by considering the head effect.
Description
Technical Field
The invention belongs to the technical field of combination of traffic congestion management and data mining, and particularly relates to a traffic congestion prediction method and device based on head effect and a circulating neural network.
Background
With rapid development of economy and rapid improvement of living standard of people, traffic demand and automobile conservation amount are continuously increased, meanwhile, the road management industry is faced with contradiction between road traffic capacity and rapidly-increased public traffic demand, urban core areas and expressways are frequently congested around urban areas, and great challenges are brought to road traffic efficiency and user experience. Many achievements exist in the research of vehicle flow evaluation, for example, many students study the traditional vehicle flow problem by establishing mathematical models such as differential equations, time prediction and BP neural networks, but the calculation process is often too complicated, and the required parameter variables are many, for example, the influence of random factors such as vehicle speed, weather, emergencies and the like on the vehicle flow is considered, and in the actual process, because the cost of obtaining the parameters is high, many values are often lost, so that the final calculation result is quite different from the actual situation. The method faces serious challenges in the popularization and application processes.
The comprehensive, timely and accurate acquisition of the traffic running state is to evaluate the road network state, analyze the running efficiency of the road network, pertinently formulate the basis of the fine traffic management and control measures, and is also the core of intelligent traffic construction. The fixed detector represented by the detection technologies of geomagnetic coils, microwave vehicle detectors, millimeter wave radars, cameras and the like can only cover part of road networks due to higher installation and maintenance cost, and cannot realize comprehensive acquisition of the running state of the urban road networks. In recent years, with the wide application of GPS (Global Positioning System) and the rising of LBS (Location-based service), track data represented by floating car data has the characteristics of abundant information, wide coverage range and the like, so that the track data hopefully solves the problem of insufficient space coverage of a road network of a fixed detector, and becomes a research hotspot of the industry.
The movement track data containing time information and position information is space-time data, which is simply called space-time track data, and specific track information such as human activity tracks, vehicle activity tracks, animal activity tracks, natural rule activity tracks, tracks collected by video monitoring and the like can be formed by aggregating specific space-time data points.
In the prior art, track data is adopted to conduct traffic prediction, for example Wen Meiling et al propose a long-short-term memory model method in 'traffic jam assessment and prediction based on track big data', and the result is compared with a support vector regression model and a cyclic neural network, so that the long-short-term memory model has a better prediction effect. Gu Changlong in the urban road network running state perception based on track data fusion, multi-source track data such as taxis, buses and bus IC cards are fused, and a Gaussian process model is adopted to predict the road network traffic running speed.
The method does not consider the interference of uncertain data when carrying out traffic prediction, so that the prediction accuracy is to be improved.
Disclosure of Invention
In view of the above, the present invention aims to provide a traffic congestion prediction method and apparatus based on the head effect and the recurrent neural network, which enhance the intention behavior by considering the head effect, thereby improving the accuracy of traffic congestion prediction.
In order to achieve the above object, an embodiment provides a traffic congestion prediction method based on head effect and recurrent neural network, including the following steps:
acquiring vehicle track point sequence data and sequentially cleaning and road network matching to obtain effective track point sequence data;
constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data;
mining real intention track data in the effective track point sequence data based on the head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram;
embedding the intention enhancement behavior diagram to obtain an embedded vector;
predicting access behaviors corresponding to future track points by using a gate control cyclic neural network according to the embedded vector;
and carrying out probability density clustering on the access behaviors corresponding to the track points at the future moment, and predicting the traffic jam condition according to the clustering result.
Preferably, the road network matching for the cleaned vehicle track point sequence data comprises:
at least one candidate point corresponding to each track point in the vehicle track point sequence data and a road section where the candidate point is located are matched in the road network database, a candidate point set corresponding to the vehicle track point sequence data is obtained, invalid candidate points in the candidate point set are deleted according to independent spatial analysis or simultaneous spatial analysis and time analysis, the remaining candidate points form an effective candidate point set, a plurality of candidate matching tracks are generated based on the effective candidate point set, and the candidate matching track with the highest quality score is screened from the candidate matching tracks to generate the effective track point sequence data.
Preferably, when space analysis is performed, if adjacent candidate points of the current candidate point are all on the corresponding road segments and the current candidate point is not on the corresponding road segments, deleting the current candidate point as an invalid candidate point;
when time analysis is carried out, calculating the average speed of the shortest path according to the distance between the current candidate point and the previous adjacent candidate point and the corresponding time interval, calculating the similarity between the average speed of the shortest path and the speed limit requirement of each road section included in the shortest path, and deleting the current candidate point as an invalid candidate point if the similarity is smaller than a set threshold value.
Preferably, the mining the real intention track data in the effective track point sequence data based on the head effect, and the intention enhancement on the original behavior graph based on the real intention track data comprises the following steps:
setting a maximum time window, and searching the effective track point sequence data in the maximum time window for a second-order concurrency mode < head access behavior and tail access behavior > representing head effect;
counting all frequent sequence modes (head access behaviors, at least 1 middle access behaviors) starting from head access behaviors and ending from tail access behaviors from the effective track point sequence data, wherein the tail access behaviors are taken as real intention track data;
calculating a probability value of the middle access behavior between the head access behavior and the tail access behavior by counting the frequency of each middle access behavior;
and calculating the relation strength of connection between the head access behavior and the tail access behavior based on the probability values of all the intermediate access behaviors, taking the relation strength as a connecting edge weight, and establishing a connecting edge between the head access behavior and the tail access behavior in the original behavior diagram so as to enhance the intention.
Preferably, the calculating the probability value that the intermediate access behavior occurs between the head access behavior and the tail access behavior by counting the frequency of each intermediate access behavior includes:
the probability value that an intermediate access behavior appears between a head access behavior and a tail access behavior is taken as the ratio of the frequency of each intermediate access behavior to the total frequency of all intermediate access behaviors in the recurrent sequence pattern.
Preferably, the calculating the relation strength of the connection between the head access behavior and the tail access behavior based on the probability values of all the intermediate access behaviors includes:
and calculating an information entropy value according to the probability values of all the intermediate access behaviors, and taking the information entropy value as the relation strength of connection between the head access behavior and the tail access behavior.
Preferably, the LINE graph embedding learning algorithm is adopted to embed the intention enhancement behavior graph to obtain an embedded vector.
Preferably, the predicting, by using a gated recurrent neural network, the access behavior corresponding to the future time track point according to the embedded vector includes:
sequencing the embedded vectors of the track points corresponding to the access behaviors according to the sequence of the track points in the effective track point sequence data to obtain an embedded vector sequence;
and (3) inputting the embedded vector sequence into a gating cyclic neural network, and calculating each future moment track point and the corresponding access behavior thereof through forward propagation.
In order to achieve the above object, an embodiment further provides a traffic congestion prediction device based on head effect and a recurrent neural network, which comprises a data acquisition and preprocessing module, an original behavior diagram construction module, an intention enhancement behavior diagram construction module, an embedded representation module, an access behavior prediction module and a traffic congestion prediction module,
the data acquisition and preprocessing module is used for acquiring vehicle track point sequence data and sequentially carrying out cleaning and road network matching to obtain effective track point sequence data;
the original behavior diagram construction module is used for constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data;
the intention enhancement behavior diagram construction module is used for mining real intention track data in the effective track point sequence data based on head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram;
the embedded representation module is used for carrying out embedded representation on the intention enhancement behavior diagram to obtain an embedded vector;
the access behavior prediction module is used for predicting the access behavior corresponding to the track point at the future moment according to the embedded vector by adopting a gating cyclic neural network;
the traffic jam prediction module is used for carrying out probability density clustering on the access behaviors corresponding to the track points at the future moment and predicting the traffic jam condition according to the clustering result.
To achieve the above object, an embodiment further provides a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the above method for predicting traffic congestion based on head effect and recurrent neural network when executing the computer program.
Compared with the prior art, the invention has the beneficial effects that at least the following steps are included:
the vehicle track point sequence data are cleaned and road network matched to remove invalid track points, so that the data quantity is reduced, real intention track data are mined based on head effect, intention enhancement is carried out on an original behavior diagram, access behaviors corresponding to future track points are predicted by using the obtained intention enhancement behavior diagram, uncertain behavior interference can be reduced, the prediction accuracy of the access behaviors is improved, on the basis, traffic congestion is predicted by means of density clustering results of the access behaviors corresponding to future track points, and the prediction accuracy of the traffic congestion is greatly improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a traffic congestion prediction method based on head effect and recurrent neural network provided by an embodiment;
FIG. 2 is a flow diagram of a method for traffic congestion prediction based on head effect and recurrent neural network provided by an embodiment;
FIG. 3 is a flow chart providing mining of real intent trace data;
FIG. 4 is a build intent enhancement behavior illustration intent provided by an embodiment;
FIG. 5 provides a schematic representation of an embedded representation of an embodiment;
FIG. 6 is a flowchart of an embodiment providing a prediction of access behavior for future time trace points using a gated recurrent neural network;
fig. 7 is a schematic structural diagram of a traffic congestion prediction device based on head effect and recurrent neural network provided by the embodiment;
FIG. 8 is a schematic diagram of a computing device provided by an embodiment.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the detailed description is presented by way of example only and is not intended to limit the scope of the invention.
In order to solve the problem that the accuracy of the existing traffic jam prediction is to be improved, the embodiment of the invention provides a traffic jam prediction method and device based on a head effect and a circulating neural network, and the intention behavior is enhanced by considering the head effect, so that the accuracy of the traffic jam prediction is improved.
As shown in fig. 1 and fig. 2, the traffic congestion prediction method based on the head effect and the recurrent neural network provided by the embodiment includes the following steps:
s110, acquiring vehicle track point sequence data, and sequentially cleaning and road network matching to obtain effective track point sequence data.
In an embodiment, the obtained vehicle track point sequence data is a sequence obtained by arranging track point positions according to a time sequence, and each track point position corresponds to an access behavior. Because the track data acquisition equipment generates errors in the acquisition process, the data is in a problem in the transmission process or the storage process, so that the problems of data key field deletion, repeated storage, error storage and the like are caused. Therefore, after the vehicle track point sequence data is obtained, the data needs to be cleaned to primarily filter out the missing key fields, repeated storage and wrong track point data.
In the embodiment, the traffic congestion prediction based on the track data needs to consider the time characteristic and the space characteristic, so that the cleaned track point data needs to be matched with the road network, and the matched track point data is distributed to different time slices. Specifically, a hidden Markov matching algorithm is adopted to perform road network matching and the matched data is projected to a corresponding time slice.
In an embodiment, performing road network matching on the cleaned vehicle track point sequence data includes:
preparing a road network database containing index edges and inflection point information, and matching at least one corresponding candidate point and a road section where the candidate point is located for each track point in the vehicle track point sequence data in the road network database by adopting a space index based on grid division of the road network database to obtain a candidate point set corresponding to the vehicle track point sequence data.
And deleting invalid candidate points in the candidate point set according to the independent spatial analysis or the simultaneous spatial analysis and the time analysis, wherein the rest candidate points form the valid candidate point set. When space analysis is carried out, if adjacent candidate points of the current candidate point are all on the corresponding road segments, but the current candidate point is not on the corresponding road segments, deleting the current candidate point as an invalid candidate point, and when the rest candidate points can be restored to obtain a track route meeting the route requirement, forming an effective candidate point set by the rest candidate points;
when the trajectory route obtained by spatially analyzing the reduction of the remaining candidate points does not meet the route requirement, time analysis is also required. When the time analysis is carried out, for the current candidate point, calculating the average speed of the shortest path according to the distance between the current candidate point and the previous adjacent candidate point and the corresponding time interval, and then calculating the similarity between the average speed of the shortest path and the speed limit requirement of each road section included in the shortest path, if the similarity is smaller than a set threshold value, deleting the current candidate point as an invalid candidate point, and analyzing the rest candidate points through time to form an effective candidate point set.
It should be noted that the route requirement that the trajectory route satisfies means that the trajectory route is on a real road section.
After the effective candidate point set is obtained, a plurality of candidate matching tracks are generated based on the effective candidate point set, and the candidate matching track with the highest quality score is screened from the candidate matching tracks to generate effective track point sequence data. In an embodiment, the similarity between the candidate matching track and the road shape of the corresponding road section is used as the quality score of the candidate matching track.
S120, constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data.
In an embodiment, after obtaining valid track point sequence data, an original behavior graph g= (V, E) is constructed, where V represents a node set, each node represents an access behavior corresponding to a track point, E represents a conjoined set, each conjoined between nodes represents a relationship strength between two access behaviors, and the relationship strength is used as a weight of the conjoined.
S130, mining real intention track data in the effective track point sequence data based on the head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram.
In an embodiment, the key to the intent enhancement of the original behavior graph is to identify intermediate uncertainty access behaviors and measure the connection strength between the true intents. Therefore, the method for mining the frequent pattern based on the head effect is provided, the interference of the uncertain access behavior of the vehicle is reduced, and the real intention of the vehicle is captured.
In an embodiment, the recurring patterns are mined in the valid trajectory point sequence data of all vehicles. In order to ensure timeliness between behaviors, a maximum time window W is set, and the time span of the mined frequency mode must be constrained within the maximum time window W. Counting the effective track point sequence data of all vehicles within the maximum time window W to obtain the type and the corresponding number of the intermediate access behaviors of the frequent patterns, specifically, searching to obtain a plurality of second-order frequent patterns representing head effect<Head access behavior act i Tail access behavior act j >As a true intent behavior, for each second order concurrency mode<Head access behavior act i Tail access behavior act j >Counting all head-by-head access actions act from valid trace point sequence data i Starting with a tail access behavior act j Ending burst sequence pattern<Head access behavior act i At least 1 intermediate access behaviour act k Tail access behavior act j >As real intent trace data, whereThe intermediate access behavior is an ambiguous behavior.
For example, as shown in fig. 3, when the maximum time window w=5 is set, searching of the real intention behavior < a, b > corresponding to the frequent sequence pattern is performed within the maximum time window w=5, so as to obtain the frequent sequence pattern < a, c, d, e, b >, < a, c, e, f, g, b >, < a, c, d, b >, wherein a is the head access behavior, b is the tail access behavior, c, d, e, f, g is the middle access behavior, and the middle access behavior is the uncertainty behavior.
After each frequent sequence pattern is obtained, as shown in FIG. 4, the types and the corresponding frequencies of the intermediate access behaviors are counted, and act is reflected by the types and the frequencies of the intermediate access behaviors i And act j The strength of the connection between them. Based on this, act with each intermediate access behavior k The ratio of the frequency of (a) to the total frequency of all intermediate access actions in the recurrent pattern is taken as the intermediate access action act k Occurs in the head access behavior act i Act with tail access behavior j Probability values between
。
After obtaining the probability values of all the intermediate access behaviors in the intermediate access behavior set S, calculating information entropy values according to the probability values of all the intermediate access behaviors, and taking the information entropy values as the relation strength of connection between the head access behaviors and the tail access behaviors
Expressed by the formula:
after obtaining the relationship strength, taking the relationship strength as a link weight, and accessing the behavior act by the head in the original behavior diagram i Act with tail access behavior j And establishing a connecting edge between the two to perform intention enhancement so as to obtain an intention enhancement behavior diagram.
And S140, embedding the intention enhancement behavior diagram to obtain an embedded vector.
The LINE graph embedded learning algorithm has the characteristics of high learning speed and suitability for directional weighted graphs, so as shown in fig. 5, the embodiment adopts the LINE graph embedded learning algorithm to carry out embedded representation on the intention enhancement behavior graph to obtain an embedded vector, and the learned behavior embedded representation is enabled to simultaneously retain local structure information and global structure information of the graph through optimizing an objective function.
S150, predicting the access behavior corresponding to the track point at the future moment according to the embedded vector by adopting a gating cyclic neural network.
In an embodiment, the embedded vector V is obtained based on the embedded representation of the intent enhancement behavior diagram act ∈R d D represents the dimension of the embedded vector. And predicting the access behavior corresponding to the track point at the future moment by adopting a gating circulating neural network. Specifically, as shown in fig. 6, the track point order actseq=in the effective track point sequence data<act 1 , act 2 , ⋯ , act n-1 , act n >Corresponding the track points to the access behaviors act ti Is embedded vector V of (a) act,i Sequencing to obtain an embedded vector sequence Vseq=<V act,1 , V act,2 , ⋯ , V act,n-1 , V act,n >Belonging to R (n×d) . The embedded vector sequence Vreq is input into a gate-controlled cyclic neural network, and each future moment track point and the corresponding access behavior thereof are calculated through forward propagation.
S160, probability density clustering is carried out on the access behaviors corresponding to the track points at the future moment, and traffic jam conditions are predicted according to clustering results.
After the access behaviors corresponding to the track points at all future moments are obtained, probability density clustering is carried out based on the density degree of the track points in space distribution, a plurality of cluster clusters are obtained, the accessed intensity of each cluster is used as the congestion degree, and accordingly the congestion condition is predicted.
According to the traffic congestion prediction method based on the head effect and the cyclic neural network, invalid track points are filtered out by cleaning and road network matching on vehicle track point sequence data, so that the data quantity is reduced, real intention track data are mined based on the head effect, the intention enhancement is carried out on an original behavior graph, the obtained intention enhancement behavior graph is used for predicting access behaviors corresponding to track points at future moments, uncertain behavior interference can be reduced, the accuracy of predicting the access behaviors is improved, on the basis, the traffic congestion situation is predicted through density clustering results of the access behaviors corresponding to the track points at the future moments, and the accuracy of traffic congestion situation prediction is greatly improved.
Based on the same inventive concept, as shown in fig. 7, the embodiment further provides a traffic congestion prediction device 700 based on head effect and a recurrent neural network, which comprises a data acquisition and preprocessing module 710, an original behavior graph construction module 720, an intention enhancement behavior graph construction module 730, an embedded representation module 740, an access behavior prediction module 750 and a traffic congestion prediction module 760.
The data acquisition and preprocessing module 710 is configured to acquire vehicle track point sequence data and sequentially perform cleaning and road network matching to obtain effective track point sequence data;
the original behavior diagram construction module 720 is configured to construct an original behavior diagram of the relationship strength between the node representation track points corresponding access behaviors and the continuous edge representation behaviors based on the valid track point sequence data;
the intention enhancement behavior diagram construction module 730 is configured to mine real intention track data in the effective track point sequence data based on the head effect, and perform intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram;
the embedding representation module 740 is configured to perform embedding representation on the intent enhancement behavior diagram to obtain an embedding vector;
the access behavior prediction module 750 is used for predicting the access behavior corresponding to the track point at the future moment according to the embedded vector by adopting a gated cyclic neural network;
the traffic congestion prediction module 760 is configured to perform probability density clustering on access behaviors corresponding to track points at future time points, and predict traffic congestion according to a clustering result.
It should be noted that, when the traffic congestion prediction device based on the head effect and the recurrent neural network provided in the foregoing embodiment performs traffic congestion prediction, the division of the foregoing functional modules should be used to illustrate, and the foregoing functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the terminal or the server is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the traffic congestion prediction device based on the head effect and the recurrent neural network provided in the above embodiment belongs to the same concept as the traffic congestion prediction method based on the head effect and the recurrent neural network, and the specific implementation process of the traffic congestion prediction device based on the head effect and the recurrent neural network is detailed in the traffic congestion prediction method based on the head effect and the recurrent neural network, which is not repeated here.
Based on the same inventive concept, as shown in fig. 8, the embodiment also provides a computing device, one or more processors (CPUs), an input/output interface, a network interface, and a memory. The memory stores a computer program executable on a processor, and the processor executes the computer program to implement the traffic congestion prediction method based on the head effect and the recurrent neural network, and the method comprises the following steps:
s110, acquiring vehicle track point sequence data, and sequentially cleaning and road network matching to obtain effective track point sequence data;
s120, constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data;
s130, mining real intention track data in the effective track point sequence data based on head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram;
s140, carrying out embedding representation on the intention enhancement behavior diagram to obtain an embedded vector;
s150, predicting access behaviors corresponding to future track points by using a gate control cyclic neural network according to the embedded vector;
s160, probability density clustering is carried out on the access behaviors corresponding to the track points at the future moment, and traffic jam conditions are predicted according to clustering results.
In practical applications, the memory may be a near-end volatile memory, such as a RAM, or a non-volatile memory, such as a ROM, a FLASH, a solid state disk, a mechanical hard disk, or a remote storage cloud. The processor may be a Central Processing Unit (CPU), a Microprocessor (MPU), a Digital Signal Processor (DSP), or a Field Programmable Gate Array (FPGA), i.e. the steps of the traffic congestion prediction method based on head effect and recurrent neural network may be implemented by these processors.
The foregoing detailed description of the preferred embodiments and advantages of the invention will be appreciated that the foregoing description is merely illustrative of the presently preferred embodiments of the invention, and that no changes, additions, substitutions and equivalents of those embodiments are intended to be included within the scope of the invention.
Claims (10)
1. The traffic jam prediction method based on the head effect and the cyclic neural network is characterized by comprising the following steps of:
acquiring vehicle track point sequence data and sequentially cleaning and road network matching to obtain effective track point sequence data;
constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data;
mining real intention track data in the effective track point sequence data based on the head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram;
embedding the intention enhancement behavior diagram to obtain an embedded vector;
predicting access behaviors corresponding to future track points by using a gate control cyclic neural network according to the embedded vector;
and carrying out probability density clustering on the access behaviors corresponding to the track points at the future moment, and predicting the traffic jam condition according to the clustering result.
2. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 1, wherein the road network matching of the cleaned vehicle track point sequence data comprises:
at least one candidate point corresponding to each track point in the vehicle track point sequence data and a road section where the candidate point is located are matched in the road network database, a candidate point set corresponding to the vehicle track point sequence data is obtained, invalid candidate points in the candidate point set are deleted according to independent spatial analysis or simultaneous spatial analysis and time analysis, the remaining candidate points form an effective candidate point set, a plurality of candidate matching tracks are generated based on the effective candidate point set, and the candidate matching track with the highest quality score is screened from the candidate matching tracks to generate the effective track point sequence data.
3. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 2, wherein when spatial analysis is performed, if adjacent candidate points of the current candidate point are all on the corresponding road segments, and the current candidate point is not on the corresponding road segments, the current candidate point is an invalid candidate point and is deleted;
when time analysis is carried out, calculating the average speed of the shortest path according to the distance between the current candidate point and the previous adjacent candidate point and the corresponding time interval, calculating the similarity between the average speed of the shortest path and the speed limit requirement of each road section included in the shortest path, and deleting the current candidate point as an invalid candidate point if the similarity is smaller than a set threshold value.
4. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 1, wherein the mining of real intention trace data in the valid trace point sequence data based on head effect, the intention enhancement of the original behavior graph based on the real intention trace data, comprises:
setting a maximum time window, and searching the effective track point sequence data in the maximum time window for a second-order concurrency mode < head access behavior and tail access behavior > representing head effect;
counting all frequent sequence modes (head access behaviors, at least 1 middle access behaviors) starting from head access behaviors and ending from tail access behaviors from the effective track point sequence data, wherein the tail access behaviors are taken as real intention track data;
calculating a probability value of the middle access behavior between the head access behavior and the tail access behavior by counting the frequency of each middle access behavior;
and calculating the relation strength of connection between the head access behavior and the tail access behavior based on the probability values of all the intermediate access behaviors, taking the relation strength as a connecting edge weight, and establishing a connecting edge between the head access behavior and the tail access behavior in the original behavior diagram so as to enhance the intention.
5. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 4, wherein the calculating the probability value that the intermediate access behavior occurs between the head access behavior and the tail access behavior by counting the frequency of each intermediate access behavior comprises:
the probability value that an intermediate access behavior appears between a head access behavior and a tail access behavior is taken as the ratio of the frequency of each intermediate access behavior to the total frequency of all intermediate access behaviors in the recurrent sequence pattern.
6. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 4, wherein the calculating the relationship strength of connection between head access behavior and tail access behavior based on the probability values of all intermediate access behaviors comprises:
and calculating an information entropy value according to the probability values of all the intermediate access behaviors, and taking the information entropy value as the relation strength of connection between the head access behavior and the tail access behavior.
7. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 1, wherein a LINE graph embedding learning algorithm is adopted to embed and represent the intention enhancement behavior graph to obtain an embedded vector.
8. The traffic congestion prediction method based on head effect and recurrent neural network according to claim 1, wherein predicting the access behavior corresponding to the future time trajectory point by using the gated recurrent neural network according to the embedded vector comprises:
sequencing the embedded vectors of the track points corresponding to the access behaviors according to the sequence of the track points in the effective track point sequence data to obtain an embedded vector sequence;
and (3) inputting the embedded vector sequence into a gating cyclic neural network, and calculating each future moment track point and the corresponding access behavior thereof through forward propagation.
9. The traffic jam prediction device based on the head effect and the cyclic neural network is characterized by comprising a data acquisition and preprocessing module, an original behavior diagram construction module, an intention enhancement behavior diagram construction module, an embedded representation module, an access behavior prediction module and a traffic jam prediction module,
the data acquisition and preprocessing module is used for acquiring vehicle track point sequence data and sequentially carrying out cleaning and road network matching to obtain effective track point sequence data;
the original behavior diagram construction module is used for constructing an original behavior diagram of the relationship strength between the corresponding access behaviors of the node representation track points and the behavior represented by the continuous edges based on the effective track point sequence data;
the intention enhancement behavior diagram construction module is used for mining real intention track data in the effective track point sequence data based on head effect, and carrying out intention enhancement on the original behavior diagram based on the real intention track data to obtain an intention enhancement behavior diagram;
the embedded representation module is used for carrying out embedded representation on the intention enhancement behavior diagram to obtain an embedded vector;
the access behavior prediction module is used for predicting the access behavior corresponding to the track point at the future moment according to the embedded vector by adopting a gating cyclic neural network;
the traffic jam prediction module is used for carrying out probability density clustering on the access behaviors corresponding to the track points at the future moment and predicting the traffic jam condition according to the clustering result.
10. A computing device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the head effect and recurrent neural network based traffic congestion prediction method of any of the preceding claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310443476.1A CN116151493B (en) | 2023-04-24 | 2023-04-24 | Traffic jam prediction method and device based on head effect and cyclic neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310443476.1A CN116151493B (en) | 2023-04-24 | 2023-04-24 | Traffic jam prediction method and device based on head effect and cyclic neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116151493A true CN116151493A (en) | 2023-05-23 |
| CN116151493B CN116151493B (en) | 2023-07-25 |
Family
ID=86360412
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310443476.1A Active CN116151493B (en) | 2023-04-24 | 2023-04-24 | Traffic jam prediction method and device based on head effect and cyclic neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116151493B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110260870A (en) * | 2019-07-18 | 2019-09-20 | 北京百度网讯科技有限公司 | Map-matching method, device, equipment and medium based on hidden Markov model |
| CN110929841A (en) * | 2019-10-11 | 2020-03-27 | 浙江大学城市学院 | Method for predicting pedestrian track based on circular neural network model with enhanced intention |
| US20200111348A1 (en) * | 2017-06-09 | 2020-04-09 | Sistema - Soluzioni Per L'ingegneria Dei Sistemi Di Trasporto E L'infomobilità S.R.L. | Computer system and method for state prediction of a traffic system |
| CN111275225A (en) * | 2018-12-04 | 2020-06-12 | 北京嘀嘀无限科技发展有限公司 | Empty vehicle track prediction method, prediction device, server and readable storage medium |
| CN112419710A (en) * | 2020-10-22 | 2021-02-26 | 深圳云天励飞技术股份有限公司 | Traffic congestion data prediction method, traffic congestion data prediction device, computer equipment and storage medium |
| CN113704318A (en) * | 2021-07-22 | 2021-11-26 | 浙大城市学院 | Recurrent neural network FPGNN behavior trajectory prediction method based on frequent pattern graph embedding |
| CN115830866A (en) * | 2022-11-18 | 2023-03-21 | 同济大学 | Traffic jam inference method, system, device and medium based on time sequence dynamic graph |
| US20230093546A1 (en) * | 2021-09-17 | 2023-03-23 | Toyota Motor Engineering & Manufacturing North America, Inc. | Vehicle communication sender identification via hyper-graph matching |
-
2023
- 2023-04-24 CN CN202310443476.1A patent/CN116151493B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200111348A1 (en) * | 2017-06-09 | 2020-04-09 | Sistema - Soluzioni Per L'ingegneria Dei Sistemi Di Trasporto E L'infomobilità S.R.L. | Computer system and method for state prediction of a traffic system |
| CN111275225A (en) * | 2018-12-04 | 2020-06-12 | 北京嘀嘀无限科技发展有限公司 | Empty vehicle track prediction method, prediction device, server and readable storage medium |
| CN110260870A (en) * | 2019-07-18 | 2019-09-20 | 北京百度网讯科技有限公司 | Map-matching method, device, equipment and medium based on hidden Markov model |
| CN110929841A (en) * | 2019-10-11 | 2020-03-27 | 浙江大学城市学院 | Method for predicting pedestrian track based on circular neural network model with enhanced intention |
| CN112419710A (en) * | 2020-10-22 | 2021-02-26 | 深圳云天励飞技术股份有限公司 | Traffic congestion data prediction method, traffic congestion data prediction device, computer equipment and storage medium |
| CN113704318A (en) * | 2021-07-22 | 2021-11-26 | 浙大城市学院 | Recurrent neural network FPGNN behavior trajectory prediction method based on frequent pattern graph embedding |
| US20230093546A1 (en) * | 2021-09-17 | 2023-03-23 | Toyota Motor Engineering & Manufacturing North America, Inc. | Vehicle communication sender identification via hyper-graph matching |
| CN115830866A (en) * | 2022-11-18 | 2023-03-21 | 同济大学 | Traffic jam inference method, system, device and medium based on time sequence dynamic graph |
Non-Patent Citations (2)
| Title |
|---|
| EMMANOUIL PH. KAMPITAKIS 等: "Shared space multi-modal traffic modeling using LSTM networks with repulsion map and an intention-based multi-loss function", 《TRANSPORTATION RESEARCH PART C: EMERGING TECHNOLOGIES》, vol. 150, pages 1 - 18 * |
| 赵庆: "面向短时交通流分析与预测的循环神经网络算法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, no. 08, pages 034 - 229 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116151493B (en) | 2023-07-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fang et al. | FTPG: A fine-grained traffic prediction method with graph attention network using big trace data | |
| WO2018122804A1 (en) | Road traffic anomaly detection method using non-isometric time/space division | |
| CN110836675B (en) | Decision tree-based automatic driving search decision method | |
| EP2590151A1 (en) | A framework for the systematic study of vehicular mobility and the analysis of city dynamics using public web cameras | |
| CN107766636A (en) | A kind of urban intersection safe evaluation method based on extreme value theory and microscopic simulation | |
| Zou et al. | Air quality prediction based on a spatiotemporal attention mechanism | |
| CN115148019A (en) | Early warning method and system based on holiday congestion prediction algorithm | |
| CN112365708A (en) | Scenic spot traffic volume prediction model establishing and predicting method based on multi-graph convolution network | |
| Xu et al. | Utilizing artificial neural network in GPS-equipped probe vehicles data-based travel time estimation | |
| Dai et al. | Spatio-temporal deep learning framework for traffic speed forecasting in IoT | |
| CN114802303A (en) | Obstacle trajectory prediction method, obstacle trajectory prediction device, electronic device, and storage medium | |
| Zhang et al. | Automatic traffic anomaly detection on the road network with spatial-temporal graph neural network representation learning | |
| CN113159371B (en) | Unknown target feature modeling and demand prediction method based on cross-modal data fusion | |
| Zhang et al. | Off-deployment traffic estimation—a traffic generative adversarial networks approach | |
| CN112101132B (en) | Traffic condition prediction method based on graph embedding model and metric learning | |
| Bing et al. | Integrating semantic zoning information with the prediction of road link speed based on taxi GPS data | |
| CN116151493B (en) | Traffic jam prediction method and device based on head effect and cyclic neural network | |
| Salamanis et al. | Evaluating the effect of time series segmentation on STARIMA-based traffic prediction model | |
| Dong et al. | An identification model of urban critical links with macroscopic fundamental diagram theory | |
| Qin et al. | Spatiotemporal K-Nearest Neighbors Algorithm and Bayesian Approach for Estimating Urban Link Travel Time Distribution From Sparse GPS Trajectories | |
| Wang et al. | The prediction of urban road traffic congestion by using a deep stacked long short-term memory network | |
| CN115456238A (en) | Urban trip demand prediction method based on dynamic multi-view coupling graph convolution | |
| Tikka et al. | Convolutional neural networks in estimating the spatial distribution of electric vehicles to support electricity grid planning | |
| Chen et al. | Examine the prediction error of ride-hailing travel demands with various ignored sparse demand effects | |
| Wu et al. | Excavation of Attractive Areas for Car-Share Travel and Prediction of Car-Share Usage |
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 |