CN114548572A - Method, device, equipment and medium for predicting urban road network traffic state - Google Patents

Method, device, equipment and medium for predicting urban road network traffic state Download PDF

Info

Publication number
CN114548572A
CN114548572A CN202210177092.5A CN202210177092A CN114548572A CN 114548572 A CN114548572 A CN 114548572A CN 202210177092 A CN202210177092 A CN 202210177092A CN 114548572 A CN114548572 A CN 114548572A
Authority
CN
China
Prior art keywords
speed
sequence
network
subsequence
road section
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.)
Pending
Application number
CN202210177092.5A
Other languages
Chinese (zh)
Inventor
廖娴静
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.)
Agricultural Bank of China
Original Assignee
Agricultural Bank of China
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 Agricultural Bank of China filed Critical Agricultural Bank of China
Priority to CN202210177092.5A priority Critical patent/CN114548572A/en
Publication of CN114548572A publication Critical patent/CN114548572A/en
Pending legal-status Critical Current

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • 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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The embodiment of the invention discloses a method, a device, equipment and a medium for predicting urban road network traffic states. Wherein, the method comprises the following steps: generating a historical speed sequence of a target road network under a plurality of historical time slots according to road network traffic data; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section. According to the embodiment of the invention, the problem of information prediction of road congestion conditions is solved, the accurate prediction of traffic conditions is realized, and the waste of time caused by road congestion is avoided.

Description

Method, device, equipment and medium for predicting urban road network traffic state
Technical Field
The embodiment of the invention relates to a computer data processing technology, in particular to a method, a device, equipment and a medium for predicting urban road network traffic states.
Background
In recent years, the government of China vigorously promotes the development and construction of intelligent traffic systems in cities so as to relieve a series of social problems caused by the increase of the usage amount of automobiles, such as traffic jam, traffic accidents, excessive energy consumption, carbon emission and the like. The traffic prediction, as an important component of an intelligent traffic system, is a process of analyzing urban road network traffic conditions, excavating traffic modes and predicting urban road network traffic trends, and has significant scientific research value and commercial value.
Currently, in recent years, deep learning models have shown excellent ability in traffic prediction, and researchers often use convolutional neural networks or graph-convolutional neural networks to capture spatial dependence and cyclic neural networks or time-convolutional networks to model non-linear time dependence. Researchers combine the basic modules and combine different types of technologies to jointly model complex and dynamic space-time association, and the most advanced traffic prediction performance is achieved. However, these efforts are deficient in that training of the model requires the use of a large amount of historical traffic data, and the predictive performance of the model is poor when the data used for training is insufficient. In addition, training of deep learning models requires the use of large amounts of historical data, which is inconvenient to collect due to problems related to privacy security of individuals or enterprises.
Disclosure of Invention
The embodiment of the invention provides a method, a device, equipment and a medium for predicting urban road network traffic states, which are used for realizing the problem of information prediction of road congestion conditions, improving the accuracy of traffic condition prediction and reducing the waste of time caused by road congestion.
In a first aspect, an embodiment of the present invention provides a method for predicting a traffic state of an urban road network, where the method includes:
according to the road network traffic data, generating a historical speed sequence of a target road network under a plurality of historical time slots, wherein each sequence object of the historical speed sequence comprises: historical traffic speed of each road section in the target road network under a set historical time slot;
generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network;
extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence;
and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
In a second aspect, an embodiment of the present invention further provides an apparatus for predicting traffic states of an urban road network, where the apparatus for predicting traffic states of an urban road network includes:
the historical speed sequence generation module is used for generating a historical speed sequence of a target road network under a plurality of historical time slots according to road network traffic data, and each sequence object of the historical speed sequence comprises: historical traffic speed of each road section in the target road network under a set historical time slot;
the space fusion speed sequence generation module is used for generating a space fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network;
the spatial fusion speed subsequence extraction module is used for extracting a spatial fusion speed subsequence of a target road section under the plurality of historical time slots from the spatial fusion speed sequence;
and the predicted speed subsequence prediction module is used for predicting to obtain the predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
In a third aspect, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor, when executing the computer program, implements the method for predicting the traffic state of the urban road network according to any embodiment of the present invention.
In a fourth aspect, an embodiment of the present invention further provides a storage medium readable by a computer, and having a computer program stored thereon, where the computer program is executed by a processor to implement the method for predicting traffic status of urban road network according to any embodiment of the present invention.
According to the technical scheme provided by the embodiment of the invention, a historical speed sequence of a target road network under a plurality of historical time slots is generated according to road network traffic data; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section. According to the embodiment of the invention, the problem of road congestion condition information prediction is solved, the accurate traffic condition prediction is realized, the road congestion phenomenon is reduced, the time waste and traffic accidents caused by road congestion are avoided, and the travel of people is facilitated.
Drawings
Fig. 1 is a flowchart of a method for predicting traffic states of an urban road network according to an embodiment of the present invention;
fig. 2a is a flowchart of another method for predicting traffic status of an urban road network according to a second embodiment of the present invention;
fig. 2b is a schematic structural diagram of a traffic prediction deep learning model of the method according to the second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a prediction apparatus for traffic status of an urban road network according to a third embodiment of the present invention;
fig. 4 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention;
fig. 5 is a schematic structural diagram of a traffic prediction platform according to a sixth embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a flowchart of a method for predicting traffic states of an urban road network according to an embodiment of the present invention. The embodiment can be suitable for accurately predicting the traffic state of the urban road network. The method of the embodiment may be executed by a device for predicting traffic conditions of a city road network, the device may be implemented by software and/or hardware, and the device may be configured in a server or a terminal device.
Correspondingly, the method specifically comprises the following steps:
s110, according to road network traffic data, generating a historical speed sequence of a target road network under a plurality of historical time slots, wherein each sequence object of the historical speed sequence comprises: and historical traffic speed of each road section in the target road network under the set historical time slot.
The road network traffic data may be traffic data of various types of vehicles in a road network in the traffic field. The target road network may be a road segment for which road condition prediction is performed. The historical speed sequence may be a sequence formed by average speeds of roads corresponding to different road segments at historical time, wherein the target road network corresponds to the different road segments. The historical traffic speed can be the average speed of different types of vehicles at the historical moment calculated under the current road section.
Illustratively, according to the road network traffic data, it is assumed that the target road network has 5 road segments, i.e. road segment 1, road segment 2, road segment 3, road segment 4 and road segment 5, which may be named as L1、L2、L3、L4And L5Assume that the plurality of historical time slots may be t-2, t-1, and t. Furthermore, historical speed sequences corresponding to t-2, t-1 and t at a plurality of historical time slots of the target road network can be generated, and the historical speed sequences are shown as the following formula:
Vgeneral (1)＝{(L1,Vt-2,1；L2,Vt-2,2；L3,Vt-2,3；L4,Vt-2,4；L5,Vt-2,5)；(L1,Vt-1,1；L2,Vt-1,2；L3,Vt-1,3；L4,Vt-1,4；L5,Vt-1,5)；(L1,Vt,1；L2,Vt,2；L3,Vt,3；L4,Vt,4；L5,Vt,5)}
Wherein, each sequence object of the historical speed sequence comprises: and historical traffic speed of each road section in the target road network under the set historical time slot. Specifically, the historical traffic speed of the road section 1 in the t-2 th time slot, the t-1 th time slot and the t-1 th time slot is Vt-2,1、Vt-1,1And Vt,1(ii) a The historical traffic speed of the road section 2 in the t-2 th time slot, the t-1 th time slot and the t time slot is Vt-2,2、Vt-1,2And Vt,2(ii) a The historical traffic speed of the road section 3 in the t-2 th time slot, the t-1 th time slot and the t time slot is Vt-2,3、Vt-1,3And Vt,3(ii) a The historical traffic speed of the road section 4 under the t-2 th, t-1 th and t-th time slots is Vt-2,4、Vt-1,4And Vt,4(ii) a And the historical traffic speed V corresponding to the road section 5 under the t-2 th time slot, the t-1 th time slot and the t time slot respectivelyt-2,5、Vt-1,5And Vt,5。
And S120, generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network.
The network topology may refer to modeling a target road network as a directed graph G (N, L) according to a driving direction of a vehicle, where a node set N represents a set of intersections (intersections of roads or selected demarcation points on roads), and a link set L represents a set of road segments. The link speed sequence may be a sequence of road average speeds corresponding to the target link at the historical time. The spatial fusion speed sequence may be a speed sequence obtained by fusing the current road speed sequence with road condition factors of other road segments in the target road network corresponding to the current road segment.
Optionally, generating a spatial fusion speed sequence corresponding to the road segment speed sequence according to the network topology corresponding to the target road network includes: inputting the road section speed sequence into a graph convolution network trained in advance; and generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the graph convolution network, pre-trained network parameters and a K-order neighborhood matrix determined by the network topology, wherein K is the quantity value of the adjacent road sections.
The graph convolution network may be a neural network for processing graph data, and implements convolution operation on the graph by means of graph theory. The graph data processed by the graph convolution network refers to a topological graph which uses a vertex and an edge to establish a corresponding relation in graph theory, such as an urban road network. The neighborhood matrix may be a data structure used to delineate vertex-to-edge relationships. Its essence is a two-dimensional array suitable for handling the association between the smallest data units. There are two modes of adjacency matrices: undirected graphs and directed graphs. The undirected graph is mainly characterized in that the undirected graph does not indicate that two-way circulation can be realized between a direction point and a point, and the directed graph comprises the direction between the two points and can be unidirectional or bidirectional. Specifically, the K-th order neighborhood matrix may describe a relationship between the current road segment and other road segments in the target road network.
Specifically, a K-order neighborhood matrix (hereinafter referred to as K-hop) segment set of each segment in a network topology corresponding to a target road network is defined as follows:
Hi(K)＝{lj|dis(li,lj)≤K,lj∈L}
wherein, dis (l)i,lj) Representing a road section liTo road section ljThe number of segments required for the shortest path.
Thus, a common adjacency matrix is the 1-hop adjacency matrix A. We can get the K-hop adjacency matrix by computing a to the K power of a. Since the adjacency matrix a usually has a diagonal value a (i, i) of 0, we add diagonal elements to the adjacency matrix a in order to model the laplacian matrix, which also means that the road segment is reachable in the road network topology map itself. The resulting K-hop adjacency matrix is:
Figure BDA0003520709890000071
where A represents the adjacency matrix, I represents the identity matrix, and Ci (-) functions act to modify the non-zero elements of the matrix to 1.
Based on the K-hop adjacency matrix, we define the traffic speed processing formula of the graph convolutional layer as follows:
Figure BDA0003520709890000072
wherein, VtRepresenting the space velocity vector, W, in the t-th time slotGCIs a trainable parameter matrix with the same dimension as the adjacency matrix a, operator o denotes the hadamard product, i.e. element-level multiplication is performed on two matrices with the same dimension. By means of the multiplication between the elements of the element,
Figure BDA0003520709890000073
a new matrix will be generated with parameters trainable at K-hop neighbor positions and zero at the rest positions, so that
Figure BDA0003520709890000074
Can be understood as a velocity vector VtIs performed by the spatial discrete convolution of (a). Final result Vt(K) Representing the spatial fusion velocity vector in the t-th time slot, its i-th element Vt i(K) Is a section of road liSpatial fusion speed in t-th time slot, which fuses Hi(K) Speed information for all road segments within.
In the previous example, the speed sequence of the road section in the t-th time slot is V under each road section in the target road networkTotal, t＝{L1,Vt,1；L2,Vt,2；L3,Vt,3；L4,Vt,4；L5,Vt,5Therefore, the historical traffic speed V corresponding to the t-th time slot of the road section 1 can be obtainedt,1I.e. the space velocity vector is Vt,1. And obtaining a K-order neighborhood matrix of
Figure BDA0003520709890000075
Since K is the contiguous link quantity value, K is 5. Therefore, the spatial fusion speed vector V of the road section 1 in the t-th time slot can be calculatedt 1(K) .1. the In the same way, the spatial fusion velocity vector V of the road section 2 in the t-th time slot can be obtainedt 2(K) (ii) a Spatial fusion velocity vector V of road section 3 in t-th time slott 3(K) (ii) a Spatial fusion velocity vector V of road segment 4 in t-th time slott 4(K) (ii) a And the spatial fusion velocity vector V of the road section 5 in the t-th time slott 5(K) In that respect Thus, it can be seen that the spatial fusion velocity sequence is Vt(K)＝{Vt 1(K)；Vt 2(K)；Vt 3(K)；Vt 4(K)；Vt 5(K)}。
The advantages of such an arrangement are: inputting a road section speed sequence into a graph convolution network trained in advance; and generating a spatial fusion speed sequence corresponding to the road section speed sequence through a graph convolution network according to a pre-trained network parameter and a K-order neighborhood matrix determined by network topology. Therefore, the influence of the traffic conditions of other road sections on the current road section can be considered, and the accuracy of the traffic condition prediction of the current road section can be improved.
And S130, extracting the space fusion speed sub-sequence of the target road section under the plurality of historical time slots from the space fusion speed sequence.
The spatial fusion speed sub-sequence may be a speed sequence of the same road segment in different historical time slots extracted from the spatial fusion speed sequence.
For the previous example, the spatial fusion speed sequence of the target road network at the t-th time slot can be calculated as Vt(K)＝{Vt 1(K)；Vt 2(K)；Vt 3(K)；Vt 4(K)；Vt 5(K) }; in the same way, the spatial fusion speed sequence of the target road network in the t-1 time slot is obtained as
Figure BDA0003520709890000081
And the spatial fusion speed sequence of the target road network in the t-2 time slot is
Figure BDA0003520709890000082
That is, the spatial fusion speed sub-sequence of the road segment 1 in the t-2 th, t-1 th and t time slots can be determined as
Figure BDA0003520709890000083
The spatial fusion speed subsequences of the road section 2 under the t-2 th, t-1 th and t-th time slots can be determined to be
Figure BDA0003520709890000084
The spatial fusion speed subsequences of the road section 3 under the t-2 th, t-1 th and t-th time slots can be determined to be
Figure BDA0003520709890000085
The spatial fusion speed subsequences of the road section 4 under the t-2 th, t-1 th and t-th time slots can be determined as
Figure BDA0003520709890000086
The spatial fusion speed subsequences of the road section 5 under the t-2 th, t-1 th and t-th time slots can be determined as
Figure BDA0003520709890000087
And S140, predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
The predicted speed sub-sequence may be a speed sub-sequence composed of predicted speeds, where the speed corresponding to the current target road segment is predicted in different predicted time slots.
In the previous example, it is assumed that the speed is predicted for the link 1. Spatial fusion speed subsequence according to road section 1
Figure BDA0003520709890000088
The predicted speed sub-sequence of the road section 1 under a plurality of predicted time slots is obtained through prediction, and the plurality of predicted time slots can comprise t +1, t +2, t +3 and the like, namely, the predicted speed sub-sequence can be obtainedTo obtain a predicted velocity subsequence of
Figure BDA0003520709890000091
According to the technical scheme provided by the embodiment of the invention, a historical speed sequence of a target road network under a plurality of historical time slots is generated according to road network traffic data; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section. According to the embodiment of the invention, the problem of road congestion condition information prediction is solved, the accurate traffic condition prediction is realized, the road congestion phenomenon is reduced, the time waste and traffic accidents caused by road congestion are avoided, and the travel of people is facilitated.
Optionally, after obtaining the predicted speed sub-sequence of the target road segment in multiple predicted time slots by prediction, the method further includes: and performing road congestion early warning on the target road section according to the predicted speed subsequence.
Continuing on the example, road segment 1 is based on a predicted speed subsequence
Figure BDA0003520709890000092
And obtaining corresponding prediction speeds under t +1, t +2 and t +3 time slots. Therefore, the predicted speed can be fed back to relevant workers, and the workers can perform road congestion early warning on the current road section according to the predicted speed.
The advantages of such an arrangement are: according to the predicted speed subsequence, road congestion early warning is carried out on the target road section, so that the waste of user time caused by the congestion condition of the road section can be avoided, and traffic accidents are avoided.
Example two
Fig. 2a is a flowchart of a method for predicting traffic status of an urban road network according to a second embodiment of the present invention. In this embodiment, a predicted speed sub-sequence of the target link in a plurality of predicted time slots is predicted and obtained according to the spatial fusion speed sub-sequence of the target link, and is further refined.
Correspondingly, the method specifically comprises the following steps:
s210, according to road network traffic data, generating a historical speed sequence of a target road network under a plurality of historical time slots, wherein each sequence object of the historical speed sequence comprises: and historical traffic speed of each road section in the target road network under the set historical time slot.
And S220, generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network.
And S230, predicting to obtain a predicted speed subsequence of the target road section in a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section and the associated description characteristics of the target road section in the plurality of historical time slots.
The association description feature may be other factors affecting the predicted speed of the target road segment, and specifically, the association description feature may include a spatial correlation feature, a temporal correlation feature, and an external factor feature.
Optionally, the association description feature includes at least one of: spatial correlation features, temporal correlation features, and external factor features; wherein the spatial correlation features comprise: the entrance degree, the exit degree and the distribution characteristics of surrounding buildings of the road section; the temporal correlation features include: historical traffic speed and time information of each road section in the previous historical time slot; the external factor characteristics include: meteorological features and fact features.
The spatial correlation feature may be a feature related to the urban road network topology in the traffic speed prediction problem or a feature related to the geographic location where the road segment is located, such as the spatial correlation feature belonging to the incoming degree and the outgoing degree of the road segment. Besides the road network related features, the road network related features also include interest point features, namely features describing the distribution of buildings around the road section, such as the distribution of schools, shopping malls and residences around the road section. The time correlation characteristic can be a characteristic related to time dynamic change in traffic speed prediction, such as a traffic speed sequence of a time slot before a current time slot, and the time itself belonging to the time correlation characteristic. The external factor characteristics may be characteristics related to external factors, such as weather, etc., in the traffic speed prediction problem.
The advantage of this arrangement is that when the speed prediction of the target link is performed in multiple prediction time slots, the spatial fusion speed subsequence and the associated description feature need to be considered. Therefore, the spatial correlation characteristic, the time correlation characteristic and the external factor characteristic of the current road section are referred to, the predicted speed can be more accurate, and the traffic condition of the current road section can be more accurately mastered.
Optionally, the predicting the predicted speed sub-sequence of the target road segment in the multiple predicted time slots according to the spatial fusion speed sub-sequence of the target road segment and the associated description features of the target road segment in the multiple historical time slots includes: performing feature splicing on each space fusion speed included in the space fusion speed subsequence and the associated description features under the matched historical time slot to obtain a spliced feature subsequence; inputting the splicing characteristic subsequence into a sequence network trained in advance; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the splicing characteristic subsequence through the sequence-to-sequence network.
The spatial fusion speed may be a fusion speed of the target link in a certain time slot. The splicing feature subsequence may be a sequence obtained by feature splicing the spatial fusion speed of the target road segment in a certain time slot with the associated description feature, and specifically, the feature splicing is performed on the spatial fusion speed, the spatial correlation feature, the temporal correlation feature and the external factor feature to obtain a corresponding splicing feature subsequence. The sequence-to-sequence network may be an Encoder-Decoder (encode-decode) architecture network, whose input is a sequence and output may be a sequence. Wherein, the Encoder part and the Decoder part are both composed of a recurrent neural network module and the like. In the Encoder, a sequence is converted into a fixed-length vector, which is then converted into the desired sequence output by the Decoder.
In the previous example, the spatial fusion speed subsequence of the road section 1 under the t-2 th, t-1 th and t time slots is determined as
Figure BDA0003520709890000121
And performing feature splicing on each space fusion speed included in the space fusion speed subsequence and the associated description features under the matched historical time slot to obtain a spliced feature subsequence. Specifically, the splicing characteristic subsequence of the road section 1 in the t-2 time slot is
Figure BDA0003520709890000122
Wherein, SF1Representing the spatial correlation characteristic, TF, of the 1 st road segmentt-2The time correlation characteristic of the t-2 time slot is shown, and the time correlation characteristic of the future time slot can be obtained in advance due to the regularity of the time correlation characteristic,
Figure BDA0003520709890000123
showing the external factor characteristics of the 1 st road segment at the t-2 th time slot,
Figure BDA0003520709890000124
and (3) a splicing characteristic subsequence of the 1 st road section in the t-2 th time slot is represented.
In the same way, the splicing characteristic subsequence of the road section 1 in the t-1 time slot is
Figure BDA0003520709890000125
The splicing characteristic subsequence of the road section 1 in the t time slot is
Figure BDA0003520709890000126
Further, splicing the characteristic subsequences
Figure BDA0003520709890000127
Figure BDA0003520709890000128
And
Figure BDA0003520709890000129
inputting a pre-trained sequence into a sequence network; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots through the sequence-to-sequence network according to the splicing characteristic subsequence.
The advantages of such an arrangement are: and predicting to obtain a predicted speed subsequence of the target road section in a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section and the associated description characteristics of the target road section in a plurality of historical time slots. Therefore, the influence generated by the spatial correlation characteristic, the temporal correlation characteristic and the external factor characteristic can be considered when the speed is predicted, and the accuracy and the reliability of the predicted speed are improved.
Specifically, the encoder of the sequence-to-sequence network takes as input a designed multi-dimensional feature vector, which is the velocity generated by the above-mentioned map convolutional layer
Figure BDA00035207098900001210
The space correlation characteristic SF, the time correlation characteristic TF and the external factor characteristic EF are spliced, wherein the time correlation characteristic only selects the time characteristic and does not contain the traffic speed sequence characteristic, and the time sequence data is required to be constructed due to the input of the sequence to the sequence network. The process is as follows:
Figure BDA0003520709890000131
Figure BDA0003520709890000132
where i denotes the ith road segment, j denotes the jth time slot, Vt-jRoad section speed sequence for representing t-j time slot，
Figure BDA0003520709890000133
Represents the K-hop spatial fusion speed, SF, of the ith road section in the t-j time slotiRepresenting the spatial correlation characteristic, TF, of the ith road segmentt-jRepresenting the time dependency characteristics at the t-j time slots,
Figure BDA0003520709890000134
showing the external factor characteristics of the ith road section in the t-j time slot,
Figure BDA0003520709890000135
and showing the splicing characteristic subsequences of the ith road section in the t-j time slot.
Hidden state h in the encoder part for the t-j, j ∈ {0, Λ m } time slott-jFrom the hidden state h of the last time slott-j-1And input Xt-jJointly calculating to obtain:
Figure BDA0003520709890000136
wherein h is0For the initial hidden state, typically set to zero vector, Cellencoder(. cndot.) is a computational function of the encoder, determined by the recurrent neural network model structure employed.
The finally obtained context vector C is the hidden state h of the current time slottIt stores all information of the encoder, including the hidden state (h)t-m,ht-m+1,Λ,ht-1) And the input vector (X)t-m,Xt-m+1,Λ,Xt) And also the connector between the encoder and decoder modules.
C＝ht
In the decoder part, the context vector C is used as an initial concealment state, and then the output predicted future velocity sequence is decoded step by step. The input of the decoder is different from that of the encoder, and the feature vector is formed by splicing spatial correlation features, temporal correlation features and external factor features:
Figure BDA0003520709890000137
wherein, SFiRepresenting the spatial correlation characteristic, TF, of the ith road segmentt+jThe time correlation characteristic of the t + j time slot is shown, and the time correlation characteristic, EF, of the future time slot can be obtained in advance due to the regularity of the time correlation characteristict iShowing the external factor characteristics of the ith road segment in the t + j time slot,
Figure BDA0003520709890000141
and showing the splicing characteristic subsequence of the ith road section in the t + j time slot.
Similarly, the hidden state h of the t + j, j ∈ {1, Λ, n } th slott+jNot only contain the input information
Figure BDA0003520709890000142
Also including the previous hidden state (h)t+1,ht+2,Λ,ht+j-1) It is calculated as follows:
Figure BDA0003520709890000143
wherein, CelldecoderIs the computational function of the decoder, determined by the structure of the cyclic convolution network model employed.
Further, the cyclic convolution network model structure adopted by the sequence-to-sequence network framework uses a gated recursive unit as an internal structure of the encoder and the decoder in the embodiment. The internal operation of the gated recursion unit is mainly to combine the input x of the current time slottAnd hidden state h transmitted from last time slott-1Calculating to obtain the output y of the current time slottAnd a hidden state h passed to the next slott。
Specifically, input x of the current time slot is firstly passedtAnd hidden state h of last time slott-1To respectively meterCalculating two gating signals ztAnd rtWherein z istIs the gating of control updates, rtIs the gating to control the reset, the calculation formula is as follows:
zt＝σ(Wz·[ht-1；xt]+bz)
rt＝σ(Wr·[ht-1；xt]+br)
where σ (·) is a nonlinear activation function that maps data into a range of (0,1), and σ (x) is 1/(1+ e)-x)，Wz,WrRepresenting a weight matrix, bz,brAre corresponding offset vectors, all of which are trainable parameters.
After the gating signal is calculated, r is usedtGated reset ht-1And is combined with xtSplicing, and then scaling the data to the range of-1 to 1 through the tanh activation function to obtain a vector ctThe calculation is as follows:
ct＝tanh(Wc·[rtοht-1；xt]+bc)
c hereintMainly containing x of the current inputtAnd h after resett-1This means that the state of the current slot is memorized. WcRepresenting a weight matrix, bcAre corresponding offset vectors, all of which are trainable parameters. Finally, the updating memory stage of the gate control recursion unit forgets the h transferredt-1And adding to the current slot input xtObtaining the hidden state h of the current time slot from some dimension informationt：
ht＝(1-zt)οht-1+ztοct
Wherein z istοctIndicating that the pair contains current time slot informationtFor selective memory, (1-z)t)οht-1Indicates a hidden state h from the originalt-1Selective forgetting.
At htOn the basis of (2), the output y of the current time slot can be calculatedtIn the sequenceListed into the sequential network framework, only the gated recursion units in the decoder have outputs.
yt＝σ(Wo·ht+bo)
Wherein, WoRepresenting a weight matrix, boAre corresponding offset vectors, are trainable parameters.
And S240, predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
In this embodiment, a method for predicting a traffic state of an urban road network may be implemented by a deep learning model for traffic prediction. Fig. 2b is a schematic structural diagram of the traffic prediction deep learning model. The map is a road condition map of a target road network of the ith road section in the t-m, t-1 and t time slots. And the K-hop spatial fusion speed of the ith road section in the t-m time slots can be obtained
Figure BDA0003520709890000151
The same principle can be used for obtaining the K-hop spatial fusion speed of the ith road section in the t-1 time slot
Figure BDA0003520709890000152
And the fusion speed of the ith road section in the Tth time slot in the K-hop space
Figure BDA0003520709890000153
Further, feature splicing is carried out on each space fusion speed and the associated description features under the matched historical time slot, and a spliced feature subsequence is obtained. Namely, each space fusion speed is spliced with the space correlation characteristic SF, the time correlation characteristic TF and the external factor characteristic EF to obtain a corresponding spliced characteristic subsequence.
Correspondingly, the splicing characteristic subsequence is input into a sequence network trained in advance, and a predicted speed subsequence of the target road section under a plurality of predicted time slots is obtained through prediction according to the splicing characteristic subsequence through the sequence network. I.e. the predicted speed sub-sequence comprises that of figure 2b
Figure BDA0003520709890000161
And
Figure BDA0003520709890000162
that is to say
Figure BDA0003520709890000163
For the predicted speed of the ith road segment in the t +1 th time slot,
Figure BDA0003520709890000164
the predicted speed for the ith road segment in the t + n time slots.
According to the technical scheme provided by the embodiment of the invention, a historical speed sequence of a target road network under a plurality of historical time slots is generated according to road network traffic data; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section and the associated description characteristics of the target road section under the plurality of historical time slots; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section. According to the embodiment of the invention, by the method, the spatial correlation characteristic, the time correlation characteristic and the external factor characteristic of the current road section are referred, so that the prediction speed can be more accurate, the traffic condition of the current road section can be more accurately mastered, the accuracy and the reliability of the prediction speed are improved, and the accurate prediction of the traffic condition is realized.
EXAMPLE III
Fig. 3 is a structural diagram of a prediction apparatus of an urban road network traffic state according to a third embodiment of the present invention, where the prediction apparatus of an urban road network traffic state according to the third embodiment of the present invention may be implemented by software and/or hardware, and may be configured in a server or a terminal device to implement a prediction method of an urban road network traffic state according to the third embodiment of the present invention. As shown in fig. 3, the apparatus may specifically include: a historical velocity sequence generation module 310, a spatial fusion velocity sequence generation module 320, a spatial fusion velocity subsequence extraction module 330, and a predicted velocity subsequence prediction module 340.
The historical speed sequence generating module 310 is configured to generate a historical speed sequence of a target road network in a plurality of historical time slots according to road network traffic data, where each sequence object of the historical speed sequence includes: historical traffic speed of each road section in the target road network under a set historical time slot;
a spatial fusion speed sequence generation module 320, configured to generate a spatial fusion speed sequence corresponding to the road segment speed sequence according to the network topology corresponding to the target road network;
a spatial fusion speed sub-sequence extraction module 330, configured to extract a spatial fusion speed sub-sequence of a target road segment in the multiple historical time slots from the spatial fusion speed sequence;
and the predicted speed sub-sequence prediction module 340 is configured to predict, according to the spatial fusion speed sub-sequence of the target road segment, a predicted speed sub-sequence of the target road segment in multiple predicted time slots.
According to the technical scheme of the embodiment, a historical speed sequence of a target road network under a plurality of historical time slots is generated according to road network traffic data; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section. According to the embodiment of the invention, the problem of road congestion condition information prediction is solved, the accurate traffic condition prediction is realized, the road congestion phenomenon is reduced, the time waste and traffic accidents caused by road congestion are avoided, and the travel of people is facilitated.
On the basis of the foregoing embodiments, the spatial fusion velocity sequence generating module 320 may be specifically configured to: inputting the road section speed sequence into a graph convolution network trained in advance; and generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the graph convolution network, pre-trained network parameters and a K-order neighborhood matrix determined by the network topology, wherein K is the quantity value of the adjacent road sections.
On the basis of the foregoing embodiments, the spatial fusion speed subsequence extraction module 330 may specifically include: and the predicted speed subsequence prediction unit is used for predicting and obtaining the predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section and the associated description characteristics of the target road section under the plurality of historical time slots.
On the basis of the above embodiments, the following embodiments may be specifically used: the association description features include at least one of: spatial correlation features, temporal correlation features, and external factor features; wherein the spatial correlation features comprise: the entrance degree, the exit degree and the distribution characteristics of surrounding buildings of the road section; the temporal correlation features include: historical traffic speed and time information of each road section in the previous historical time slot; the external factor characteristics include: meteorological features and real-world features.
On the basis of the foregoing embodiments, the prediction speed sub-sequence prediction unit may be specifically configured to: performing feature splicing on each space fusion speed included in the space fusion speed subsequence and the associated description features under the matched historical time slot to obtain a spliced feature subsequence; inputting the splicing characteristic subsequence into a sequence network trained in advance; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the splicing characteristic subsequence through the sequence-to-sequence network.
On the basis of the above embodiments, the system further comprises a road congestion early warning module, which may be specifically configured to: and after the predicted speed subsequence of the target road section in a plurality of predicted time slots is obtained through prediction, carrying out road congestion early warning on the target road section according to the predicted speed subsequence.
The urban road network traffic state prediction device can execute the urban road network traffic state prediction method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.
Example four
Fig. 4 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention. As shown in fig. 4, the apparatus includes a processor 410, a memory 420, an input device 430, and an output device 440; the number of the processors 410 in the device may be one or more, and one processor 410 is taken as an example in fig. 4; the processor 410, the memory 420, the input device 430 and the output device 440 in the apparatus may be connected by a bus or other means, for example, in fig. 4.
The memory 420 may be used as a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the prediction method of the urban road network traffic state in the embodiment of the present invention (for example, the historical speed sequence generation module 310, the spatial fusion speed sequence generation module 320, the spatial fusion speed subsequence extraction module 330, and the predicted speed subsequence prediction module 340). The processor 410 executes various functional applications and data processing of the device by executing software programs, instructions and modules stored in the memory 420, so as to implement the above-mentioned method for predicting the traffic state of the urban road network, which comprises: according to the road network traffic data, generating a historical speed sequence of a target road network under a plurality of historical time slots, wherein each sequence object of the historical speed sequence comprises: historical traffic speed of each road section in the target road network under a set historical time slot; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
The memory 420 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the terminal, and the like. Further, the memory 420 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some examples, memory 420 may further include memory located remotely from processor 410, which may be connected to devices through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 430 may be used to receive input numeric or character information and generate key signal inputs related to user settings and function control of the apparatus. The output device 440 may include a display device such as a display screen.
EXAMPLE five
An embodiment of the present invention further provides a computer-readable storage medium, which when executed by a computer processor, is configured to perform a method for predicting traffic status of an urban road network, the method including: according to the road network traffic data, generating a historical speed sequence of a target road network under a plurality of historical time slots, wherein each sequence object of the historical speed sequence comprises: historical traffic speed of each road section in the target road network under a set historical time slot; generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network; extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence; and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
Of course, the embodiment of the present invention provides a storage medium containing computer readable instructions, and the computer readable instructions are not limited to the operations of the method described above, and may also perform related operations in the method for predicting traffic status of urban road network provided in any embodiment of the present invention.
From the above description of the embodiments, it is obvious for those skilled in the art that the present invention can be implemented by software and necessary general hardware, and certainly, can also be implemented by hardware, but the former is a better embodiment in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods according to the embodiments of the present invention.
It should be noted that, in the embodiment of the prediction apparatus for traffic state of urban road network, the units and modules included in the prediction apparatus are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
EXAMPLE six
Fig. 5 is a schematic structural diagram of a traffic prediction platform according to a sixth embodiment of the present invention. As shown in fig. 5, the traffic prediction platform may specifically include: a central server 510, and a plurality of traffic information owners 520, wherein:
the traffic information owner 520 is configured to train the locally-held graph convolutional network and the sequence-to-sequence network according to locally-held road network traffic data, and upload training parameters to the central server 510 in a homomorphic encryption manner;
the graph convolution network is used for outputting a matched space fusion speed sequence according to the input road section speed sequence; the sequence-to-sequence network is used for predicting to obtain a predicted speed subsequence of any road section under a plurality of predicted time slots according to a feature subsequence spliced by space fusion speed of the road section under a plurality of historical time slots and associated description features under matched historical time slots;
the central server 510 is configured to perform parameter aggregation on the training parameters sent by each traffic information owner 520, encrypt data of an aggregation result, and return to each traffic information owner for iterative training until a training termination condition is met;
the central server 510 is further configured to execute the method according to any one of claims 1 to 6 based on the graph convolution network and the sequence-to-sequence network jointly trained by the traffic information owners 520.
Illustratively, in order to solve the problem of few traffic data samples and simultaneously achieve privacy protection, a traffic prediction platform architecture based on combination of privacy protection and traffic state online prediction is designed in this embodiment, as shown in fig. 5. The platform combines multiple enterprises which independently own taxi track data, bus track data, road networks, interest point data and meteorological data, respectively correspond to taxi enterprises, bus groups, map navigation enterprises and meteorological bureaus in the figure 5, jointly trains a combined model, designs encryption type parameter transmission in the training process to replace original remote data transmission, can guarantee the data safety and privacy of each data owner, and simultaneously meets the requirements of the issued laws and regulations on data safety.
Specifically, each traffic information owner 520, for example, the taxi enterprise, the public transportation group, the map navigation enterprise, and the weather bureau in fig. 5, performs data preprocessing, such as abnormal value processing, feature coding, and the like, on its own data source. For taxi enterprises and public transportation groups, because the data source is the global positioning system track data, the instantaneous speed is calculated according to the driving direction and track points and is mapped to road sections on the urban bidirectional road network.
Further, the traffic information owner 520 uploads the preprocessed data sets to a traffic prediction platform, and the platform automatically performs sample encryption and alignment on the data sets, that is, finds out intersections of road sections by using an encryption-based sample alignment technology. The central server 510 distributes an initial model, i.e., the traffic prediction deep learning model described above, to each traffic information owner 520. Each traffic information owner 520 trains a local model by using a local data set, and uploads local model training parameters to the central server 510 after being encrypted by a homomorphic encryption technology.
Correspondingly, the central server 510 aggregates all uploaded model gradients, and then transmits the aggregated combined model parameters back to each traffic information owner 520 through an encryption technique, and each traffic information owner 520 updates local model parameters respectively and continues iterative training until the model converges.
After the training is finished, the central server 510 obtains the trained joint traffic prediction model, and can provide real-time traffic prediction services, such as urban road traffic state monitoring, road congestion early warning, and the like.
According to the technical scheme, the traffic prediction platform comprises a central server and a plurality of traffic information owners, wherein the traffic information owners are used for training a locally-held graph convolution network and a sequence-to-sequence network according to locally-held road network traffic data and uploading training parameters to the central server in a homomorphic encryption mode; the central server is used for performing parameter aggregation on the training parameters sent by each traffic information owner, performing data encryption on an aggregation result, and returning to each traffic information owner for iterative training until a training ending condition is met; the central server is further configured to execute the method according to any one of claims 1 to 6 based on the graph convolution network and the sequence-to-sequence network that are obtained by co-training of the traffic information owners. According to the embodiment of the invention, the problem of data isolated island of the road privacy data is solved through the traffic prediction platform, and the traffic data is safely and conveniently utilized. Therefore, the traffic condition prediction is facilitated, and the prediction accuracy of the traffic prediction model on the data domain with scarce data is improved.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims (10)

1. A road network traffic state prediction method is characterized by comprising the following steps:
according to the road network traffic data, generating a historical speed sequence of a target road network under a plurality of historical time slots, wherein each sequence object of the historical speed sequence comprises: historical traffic speed of each road section in the target road network under a set historical time slot;
generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network;
extracting a space fusion speed subsequence of a target road section under the plurality of historical time slots from the space fusion speed sequence;
and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
2. The method of claim 1, wherein generating a spatial fusion speed sequence corresponding to a segment speed sequence according to a network topology corresponding to the target road network comprises: inputting the road section speed sequence into a graph convolution network trained in advance;
and generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the graph convolution network, pre-trained network parameters and a K-order neighborhood matrix determined by the network topology, wherein K is the quantity value of the adjacent road sections.
3. The method of claim 1, wherein predicting a predicted speed subsequence of the target road segment at a plurality of prediction time slots according to the spatial fusion speed subsequence of the target road segment comprises:
and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section and the associated description characteristics of the target road section under the plurality of historical time slots.
4. The method of claim 3, wherein the association description feature comprises at least one of: spatial correlation features, temporal correlation features, and external factor features;
wherein the spatial correlation features comprise: the entrance degree, the exit degree and the distribution characteristics of surrounding buildings of the road section;
the temporal correlation features include: historical traffic speed and time information of each road section in the previous historical time slot;
the external factor characteristics include: meteorological features and fact features.
5. The method according to claim 3 or 4, wherein predicting the predicted speed subsequence of the target road segment at a plurality of prediction time slots according to the spatial fusion speed subsequence of the target road segment and the associated description feature of the target road segment at the plurality of historical time slots comprises:
performing feature splicing on each space fusion speed included in the space fusion speed subsequence and the associated description features under the matched historical time slot to obtain a spliced feature subsequence;
inputting the splicing characteristic subsequence into a sequence network trained in advance;
and predicting to obtain a predicted speed subsequence of the target road section under a plurality of predicted time slots according to the splicing characteristic subsequence through the sequence-to-sequence network.
6. The method of claim 1, further comprising, after predicting the predicted speed subsequence for the target segment at the plurality of predicted time slots:
and performing road congestion early warning on the target road section according to the predicted speed subsequence.
7. A prediction device for a traffic state of a road network, comprising:
the historical speed sequence generation module is used for generating a historical speed sequence of a target road network under a plurality of historical time slots according to road network traffic data, and each sequence object of the historical speed sequence comprises: historical traffic speed of each road section in the target road network under a set historical time slot;
the spatial fusion speed sequence generation module is used for generating a spatial fusion speed sequence corresponding to the road section speed sequence according to the network topology corresponding to the target road network;
the spatial fusion speed subsequence extraction module is used for extracting a spatial fusion speed subsequence of a target road section under the plurality of historical time slots from the spatial fusion speed sequence;
and the predicted speed subsequence prediction module is used for predicting to obtain the predicted speed subsequence of the target road section under a plurality of predicted time slots according to the spatial fusion speed subsequence of the target road section.
8. Computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor when executing the computer program implements a method for prediction of road network traffic status according to any of claims 1-6.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements a method for predicting traffic status of a road network according to any one of claims 1 to 6.
10. A traffic prediction platform, comprising: a central server, and a plurality of traffic information owners, wherein:
the traffic information owner is used for training the locally-held graph convolution network and the sequence-to-sequence network according to locally-held road network traffic data and uploading training parameters to the central server in a homomorphic encryption mode;
the graph convolution network is used for outputting a matched space fusion speed sequence according to the input road section speed sequence; the sequence-to-sequence network is used for predicting to obtain a predicted speed subsequence of any road section under a plurality of predicted time slots according to a feature subsequence spliced by space fusion speed of the road section under a plurality of historical time slots and associated description features under matched historical time slots;
the central server is used for performing parameter aggregation on the training parameters sent by each traffic information owner, performing data encryption on an aggregation result, and returning to each traffic information owner for iterative training until a training ending condition is met;
the central server is further configured to execute the method according to any one of claims 1 to 6 based on the graph convolution network and the sequence-to-sequence network that are obtained by co-training of the traffic information owners.
CN202210177092.5A 2022-02-25 2022-02-25 Method, device, equipment and medium for predicting urban road network traffic state Pending CN114548572A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210177092.5A CN114548572A (en) 2022-02-25 2022-02-25 Method, device, equipment and medium for predicting urban road network traffic state

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210177092.5A CN114548572A (en) 2022-02-25 2022-02-25 Method, device, equipment and medium for predicting urban road network traffic state

Publications (1)

Publication Number Publication Date
CN114548572A true CN114548572A (en) 2022-05-27

Family

ID=81678526

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210177092.5A Pending CN114548572A (en) 2022-02-25 2022-02-25 Method, device, equipment and medium for predicting urban road network traffic state

Country Status (1)

Country Link
CN (1) CN114548572A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117010567A (en) * 2023-08-07 2023-11-07 大连理工大学 Dynamic chemical process prediction method based on static space-time diagram

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117010567A (en) * 2023-08-07 2023-11-07 大连理工大学 Dynamic chemical process prediction method based on static space-time diagram

Similar Documents

Publication Publication Date Title
Xu et al. Real-time prediction of taxi demand using recurrent neural networks
Tang et al. Multi-community passenger demand prediction at region level based on spatio-temporal graph convolutional network
CN109658695B (en) Multi-factor short-term traffic flow prediction method
Chen et al. A graph convolutional stacked bidirectional unidirectional-LSTM neural network for metro ridership prediction
CN114519932B (en) Regional traffic condition integrated prediction method based on space-time relation extraction
US20230334981A1 (en) Traffic flow forecasting method based on multi-mode dynamic residual graph convolution network
Li et al. Deep spatio-temporal adaptive 3d convolutional neural networks for traffic flow prediction
CN113762338B (en) Traffic flow prediction method, equipment and medium based on multiple graph attention mechanism
CN113673769A (en) Graph neural network traffic flow prediction method based on multivariate time sequence interpolation
Luo et al. Stgin: A spatial temporal graph-informer network for long sequence traffic speed forecasting
CN113159403A (en) Method and device for predicting pedestrian track at intersection
CN114757389A (en) Federal learning-based urban traffic flow space-time prediction method
Dai et al. Attention based simplified deep residual network for citywide crowd flows prediction
Chen et al. Prediction of Public Bus Passenger Flow Using Spatial–Temporal Hybrid Model of Deep Learning
CN114548572A (en) Method, device, equipment and medium for predicting urban road network traffic state
CN114202120A (en) Urban traffic travel time prediction method aiming at multi-source heterogeneous data
Zhang et al. Boosted trajectory calibration for traffic state estimation
CN117236492B (en) Traffic demand prediction method based on dynamic multi-scale graph learning
Sheng et al. Deep spatial-temporal travel time prediction model based on trajectory feature
CN116665448A (en) Traffic speed real-time prediction method and system based on graph convolution network
Zhang et al. Spatio-temporal residual graph convolutional network for short-term traffic flow prediction
Liu et al. Co-prediction of multimodal transportation demands with self-learned spatial dependence
Guo et al. Multi‐step traffic speed prediction model with auxiliary features on urban road networks and its understanding
CN116258253A (en) Vehicle OD prediction method based on Bayesian neural network
Xu et al. Road Travel Time Prediction Based on Improved Graph Convolutional Network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination