CN111768625A - Traffic road event prediction method based on graph embedding - Google Patents

Traffic road event prediction method based on graph embedding Download PDF

Info

Publication number
CN111768625A
CN111768625A CN202010623828.8A CN202010623828A CN111768625A CN 111768625 A CN111768625 A CN 111768625A CN 202010623828 A CN202010623828 A CN 202010623828A CN 111768625 A CN111768625 A CN 111768625A
Authority
CN
China
Prior art keywords
road
data
traffic
graph
embedding
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
CN202010623828.8A
Other languages
Chinese (zh)
Inventor
吴向平
平力俊
徐懂事
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.)
China Jiliang University
Original Assignee
China Jiliang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Jiliang University filed Critical China Jiliang University
Priority to CN202010623828.8A priority Critical patent/CN111768625A/en
Publication of CN111768625A publication Critical patent/CN111768625A/en
Pending legal-status Critical Current

Links

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/901—Indexing; Data structures therefor; Storage structures
    • G06F16/9024—Graphs; Linked lists
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing

Abstract

The invention belongs to the technical field of intelligent traffic, and relates to a traffic road event prediction method based on graph embedding. The invention is beneficial to the management decision of urban managers on urban traffic, maintains the stable traffic operation order, relieves the urban traffic pressure and makes urban residents go out more conveniently.

Description

Traffic road event prediction method based on graph embedding
Technical Field
The invention belongs to the technical field of intelligent traffic, and relates to a traffic road event prediction method based on graph embedding.
Background
In the process of promoting the urbanization process, the urban vehicles rapidly grow, so that the pressure of urban traffic is gradually increased, particularly, the occurrence of a serious road traffic incident has great interference on traffic road conditions, the normal travel of urban residents is influenced, and the serious traffic incident can directly or indirectly bring casualties and economic and property losses to cities. The analysis of the urban traffic incident has the difficulties that firstly, the occurrence of the traffic incident has uncertainty, secondly, the traffic incident can be divided into various types, not only are the reasons for the occurrence complicated and various, but also the influence procedures on the traffic incident are inconsistent, and finally, the data of the traffic incident is sparse compared with the traffic volume information of roads, and for the historical data of the traffic incident, a better prediction effect is difficult to generate through a traditional statistical learning method.
Graph Embedding (Graph Embedding) is a process for mapping Graph data into data vectors, and by adopting a Graph Embedding method, the structural information and potential characteristics of a Graph model can be well reserved while the Graph structural data can be converted into vector data. The current graph embedding algorithm mainly has three methods: (1) a factorization based approach; (2) a walk-based method; (3) a method based on deep learning. GraphSAGE (graph Sample and aggregate) is a framework of inductive learning capable of efficiently generating unknown vertex Embelling by using attribute information of the vertex, and the core idea of the framework is to generate an Embelling vector of a target vertex by learning a function for performing aggregation representation on neighbor vertices. However, in the research on the data embedding method of the traffic road network, most of the data embedding methods are based on the road network structure of roads, and are not analyzed and searched in combination with traffic events and other data attribute information.
Kernel Density Estimation (KDE) is one of the important methods in non-parametric statistical tests, and is commonly used to estimate unknown Density functions.
Disclosure of Invention
In order to solve the technical problems in the prior art, the invention provides a traffic road event prediction method based on graph embedding, which has the following specific technical scheme:
a traffic road event prediction method based on graph embedding comprises the following steps:
step 1, acquiring road speed data in 2 hours before a road with a traffic incident in a historical traffic data set and counting basic attribute characteristic data of the road, and performing normalization processing to obtain the road speed data and the basic attribute characteristic data as input data of a graph embedding model;
step 2, mapping an actual road network into data of a graph structure, training a graph embedding model by adopting a graph embedding method based on GraphSAGE, and embedding each road with a traffic event into a vector by utilizing attribute characteristic data and a space structure of the road;
step 3, using the embedded vector of the road with the traffic incident obtained in the step 2 to perform dimension reduction processing, obtaining a data result reduced to two dimensions as a sample point, and fitting the input sample point by using a Gaussian kernel function as a kernel function of kernel density estimation to obtain a well-fitted probability density model;
and 4, counting attribute characteristic data of roads within 2 hours before the road where the traffic event occurs, using a trained graph embedding model, embedding to obtain a data vector, reducing dimensions, inputting the data vector into a fitted probability density model to obtain the occurrence probability of the traffic event on the roads, and drawing a kernel density estimation graph for visual display.
Further, the content of step 1 is as follows:
step 1.1, speed data in the first 2 hours of traffic incidents of all roads in the historical traffic data set are obtained, average speed in each time period is counted by taking 30 minutes as time granularity, and the average speed in each time period is normalized by using a min-max normalization method, wherein the calculation formula is as follows:
Figure BDA0002564706620000021
wherein, VvalueRepresenting the actual speed, V, of the target road during the corresponding time periodmaxRepresenting the maximum value, V, of all statistical road speeds in the corresponding time periodminRepresents the minimum of all the statistical link speeds within the corresponding time period,
Figure BDA0002564706620000022
expressed as a result of normalization of the target road speed over the corresponding time period;
step 1.2, the road length of the road where each event occurs and the number of the roads connected with each road are counted, and a min-max standardization method is also used for carrying out normalization processing; secondly, counting time characteristics when the traffic incident occurs, wherein the time characteristics are judged whether the incident time is in the morning and evening peak time periods of 06:00 to 09:00 and 16:00 to 19:00, if the incident time is in the morning and evening peak time periods, the data value of the time characteristics is set to be 1, and if not, the data value of the time characteristics is 0;
and 1.3, splicing all data items in each road to serve as attribute characteristic data of each road.
Further, the content of step 2 is as follows:
step 2.1, taking road sections in an actual road network as nodes V in a graph network structure G (V, E), taking intersection points of roads in the actual road network as edges E of the graph network, and mapping the actual road network into a graph network structure G;
2.2, sampling 2-order neighbor nodes of the target node by using a random resampling method;
and 2.3, for the collected set of adjacent nodes of the target node, aggregating the characteristics of the adjacent nodes to the target node in an average aggregation mode, namely, firstly, taking an average value for each dimension of the characteristics of the adjacent nodes, then splicing the average value with the characteristics of the target node, and performing nonlinear transformation after splicing, wherein the calculation formula is as follows:
Figure BDA0002564706620000031
Figure BDA0002564706620000032
where v is the arriving neighbor node that the target node u randomly samples,
Figure BDA0002564706620000033
representing the current characteristics of the vertices, mean is expressed as an averaging operation,
Figure BDA0002564706620000034
features representing k-th layer neighbor aggregation, sigma represents an activation function Sigmoid function, WkExpressing optimization parameters, and expressing CONCAT as characteristic splicing operation;
step 2.4, after the expression vectors of the local domain information characteristics of the nodes are obtained, training is carried out in an unsupervised mode, the learning rate is initialized to be 0.00001, the degree of the maximum node is set to be 7, the dimension of the embedded vector is set to be 256 dimensions, the number of samples selected in each training is 20, and the loss function of the unsupervised function is expressed as:
Figure BDA0002564706620000035
wherein Z isuRepresenting the embedding vector generated by graph embedding algorithm of GraphSAGE for a target node sigma, sigma representing the Sigmoid function of the activation function, v being the arriving neighbor node randomly sampled by the target node u, Pn(v) Is the probability distribution of the negative samples, vn～Pn(v) Denoted as node vnNegative sample distribution P from target node unSampled, Q is the number of samples sampled; obtaining an embedded vector Z of a target node u through forward propagationuThen, a back propagation optimization parameter W is carried out through a gradient descent algorithmkAnd an aggregation function internal parameter.
Further, the content of step 3 is as follows:
using the vector obtained by the road with the traffic incident through the map embedding algorithm obtained in the step 2, reducing the dimensionality to two dimensions through a t-SNE method, using two-dimensional data as sample points, fitting the input sample points through a kernel density estimation method by adopting a smooth kernel function to obtain a fitted probability density model, and finally simulating the real probability distribution through the fitted probability density model, wherein the calculation formula is as follows:
Figure BDA0002564706620000041
where f is the probability density function, x1，x2，x3，…xiIs n independent and equally distributed sample points, kernelfunctionRepresenting a kernel function, using a gaussian kernel density estimate as a kernel function for the density estimate; h is a smoothing coefficient that controls the degree of smoothing of the fitted curve.
Further, the content of step 4 is as follows: and (3) counting attribute characteristic data of roads within 2 hours before the road with the probability of the occurrence of the traffic incident needing to be predicted, embedding the attribute characteristic data into a trained graph embedding model to obtain data vectors of the roads, reducing the dimensionality to two dimensions, inputting the data vectors into a fitted probability density model to obtain the probability value of the occurrence of the traffic incident of the road points, and carrying out visual display on the result in a mode of a kernel density estimation graph.
Furthermore, a "+" scale label on the abscissa of the nuclear density estimation diagram represents the projection of each sample road on the abscissa axis, a circle point on the nuclear density estimation diagram represents a corresponding road, and the height of the corresponding ordinate of the circle point represents the probability value of the occurrence of the traffic accident on the corresponding road.
The invention can provide reference for relevant management departments of cities to assist in improving the decision efficiency of the departments on traffic management, and has important practical significance for the construction and development of urban traffic ecological environment.
Drawings
FIG. 1 is a schematic flow diagram of a graph-based embedded traffic road event prediction method of the present invention;
FIG. 2 is a schematic diagram of a probability density visualization result of the graph embedding-based traffic road event prediction method.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings.
As shown in fig. 1 and 2, a traffic road event prediction method based on map embedding includes the following steps:
step 1, acquiring road speed data in 2 hours before a road with a traffic incident in a historical traffic data set and counting basic attribute characteristic data of the road, and performing normalization processing to obtain the road speed data and the basic attribute characteristic data as input data of a graph embedding model;
step 2, mapping an actual road network into data of a graph structure, training a graph embedding model by adopting a graph embedding method based on GraphSAGE, and embedding each road with a traffic event into a vector by utilizing attribute characteristic data and a space structure of the road;
step 3, using the embedded vector of the road with the traffic incident obtained in the step 2 to perform dimension reduction processing, obtaining a data result reduced to two dimensions as a sample point, and fitting the input sample point by using a Gaussian kernel function as a kernel function of kernel density estimation to obtain a well-fitted probability density model;
and 4, counting attribute characteristic data of roads within 2 hours before the road where the traffic event occurs, using a trained graph embedding model, embedding to obtain a data vector, reducing dimensions, inputting the data vector into a fitted probability density model to obtain the occurrence probability of the traffic event on the roads, and drawing a kernel density estimation graph for visual display.
The contents of the step 1 are as follows:
step 1.1, speed data in the first 2 hours of traffic incidents of all roads in the historical traffic data set are obtained, average speed in each time period is counted by taking 30 minutes as time granularity, and the average speed in each time period is normalized by using a min-max normalization method, wherein the calculation formula is as follows:
Figure BDA0002564706620000051
wherein, VvalueRepresenting the actual speed, V, of the target road during the corresponding time periodmaxRepresenting the maximum value, V, of all statistical road speeds in the corresponding time periodminRepresents the minimum of all the statistical link speeds within the corresponding time period,
Figure BDA0002564706620000052
expressed as a result of normalization of the target road speed over the corresponding time period;
step 1.2, the road length of the road where each event occurs and the number of the roads connected with each road are counted, and a min-max standardization method is also used for carrying out normalization processing; secondly, counting time characteristics when the traffic incident occurs, wherein the time characteristics are judged whether the incident time is in the morning and evening peak time periods of 06:00 to 09:00 and 16:00 to 19:00, if the incident time is in the morning and evening peak time periods, the data value of the time characteristics is set to be 1, and if not, the data value of the time characteristics is 0;
and 1.3, splicing all data items in each road to serve as attribute characteristic data of each road.
The contents of the step 2 are as follows:
step 2.1, taking road sections in an actual road network as nodes V in a graph network structure G (V, E), taking intersection points of roads in the actual road network as edges E of the graph network, and mapping the actual road network into a graph network structure G;
2.2, sampling 2-order neighbor nodes of the target node by using a random resampling method;
and 2.3, for the collected set of adjacent nodes of the target node, aggregating the characteristics of the adjacent nodes to the target node in an average aggregation mode, namely, firstly, taking an average value for each dimension of the characteristics of the adjacent nodes, then splicing the average value with the characteristics of the target node, and performing nonlinear transformation after splicing, wherein the calculation formula is as follows:
Figure BDA0002564706620000061
Figure BDA0002564706620000062
where v is the arriving neighbor node that the target node u randomly samples,
Figure BDA0002564706620000063
representing the current characteristics of the vertices, mean is expressed as an averaging operation,
Figure BDA0002564706620000064
features representing k-th layer neighbor aggregation, sigma represents an activation function Sigmoid function, WkExpressing optimization parameters, and expressing CONCAT as characteristic splicing operation;
step 2.4, after the expression vectors of the local domain information characteristics of the nodes are obtained, training is carried out in an unsupervised mode, the learning rate is initialized to be 0.00001, the degree of the maximum node is set to be 7, the dimension of the embedded vector is set to be 256 dimensions, the number of samples selected in each training is 20, and the loss function of the unsupervised function is expressed as:
Figure BDA0002564706620000065
wherein Z isuRepresenting an embedding vector generated by a graph embedding algorithm of GraphSAGE for a target node u, sigma representing an activation function Sigmoid function, v being an arriving neighbor node randomly sampled by the target node u, Pn(v) Is the probability distribution of the negative samples, vn～Pn(v) Denoted as node vnNegative sample distribution P from target node unSampled, Q is the number of samples sampled; obtaining an embedded vector Z of a target node u through forward propagationuThen, a back propagation optimization parameter W is carried out through a gradient descent algorithmkAnd an aggregation function internal parameter.
The contents of the step 3 are as follows:
using the vector obtained by the road with the traffic incident through the map embedding algorithm obtained in the step 2, reducing the dimensionality to two dimensions through a t-SNE method, using two-dimensional data as sample points, fitting the input sample points through a kernel density estimation method by adopting a smooth kernel function to obtain a fitted probability density model, and finally simulating the real probability distribution through the fitted probability density model, wherein the calculation formula is as follows:
Figure BDA0002564706620000066
where f is the probability density function, x1，x2，x3，…xiIs n independent and equally distributed sample points, kernelfunctionRepresenting a kernel function, using a gaussian kernel density estimate as a kernel function for the density estimate; h is a smoothing coefficient that controls the degree of smoothing of the fitted curve.
The contents of the step 4 are as follows: and (3) counting attribute characteristic data of roads within 2 hours before the road with the probability of the occurrence of the traffic incident needing to be predicted, embedding the attribute characteristic data into a trained graph embedding model to obtain data vectors of the roads, reducing the dimensionality to two dimensions, inputting the data vectors into a fitted probability density model to obtain the probability value of the occurrence of the traffic incident of the road points, and carrying out visual display on the result in a mode of a kernel density estimation graph. The "+" scale label on the abscissa of the nuclear density estimation graph represents the projection of each sample road on the abscissa axis, the corresponding road is represented by a dot on the nuclear density estimation graph, and the corresponding ordinate height of the dot represents the probability value of the traffic accident on the corresponding road.
In practical application, the map embedding-based traffic road event prediction method disclosed by the invention is applied to a multi-attribute map embedding method of a road traffic event, a probability density model is fitted by combining a kernel density estimation method so as to output the probability value of the occurrence of the traffic event, the probability value is used for predicting the probability of the traffic event occurring on the road, and a reference is provided for relevant management departments of cities so as to assist in improving the decision efficiency of the traffic management, maintain the stable traffic running order, relieve the urban traffic pressure, enable urban residents to travel more conveniently and quickly, and have important practical significance for the construction and development of the urban traffic ecological environment.

Claims (6)

1. A traffic road event prediction method based on graph embedding is characterized by comprising the following steps:
step 1, acquiring road speed data in 2 hours before a road with a traffic incident in a historical traffic data set and counting basic attribute characteristic data of the road, and performing normalization processing to obtain the road speed data and the basic attribute characteristic data as input data of a graph embedding model;
step 2, mapping an actual road network into data of a graph structure, training a graph embedding model by adopting a graph embedding method based on GraphSAGE, and embedding each road with a traffic event into a vector by utilizing attribute characteristic data and a space structure of the road;
step 3, using the embedded vector of the road with the traffic incident obtained in the step 2 to perform dimension reduction processing, obtaining a data result reduced to two dimensions as a sample point, and fitting the input sample point by using a Gaussian kernel function as a kernel function of kernel density estimation to obtain a well-fitted probability density model;
and 4, counting attribute characteristic data of roads within 2 hours before the road where the traffic event occurs, using a trained graph embedding model, embedding to obtain a data vector, reducing dimensions, inputting the data vector into a fitted probability density model to obtain the occurrence probability of the traffic event on the roads, and drawing a kernel density estimation graph for visual display.
2. The method according to claim 1, wherein the step 1 comprises the following steps:
step 1.1, speed data in the first 2 hours of traffic incidents of all roads in the historical traffic data set are obtained, average speed in each time period is counted by taking 30 minutes as time granularity, and the average speed in each time period is normalized by using a min-max normalization method, wherein the calculation formula is as follows:
Figure FDA0002564706610000011
wherein, VvalueRepresenting the actual speed, V, of the target road during the corresponding time periodmaxRepresenting the maximum value, V, of all statistical road speeds in the corresponding time periodminRepresents the minimum of all the statistical link speeds within the corresponding time period,
Figure FDA0002564706610000012
expressed as a result of normalization of the target road speed over the corresponding time period;
step 1.2, the road length of the road where each event occurs and the number of the roads connected with each road are counted, and a min-max standardization method is also used for carrying out normalization processing; secondly, counting time characteristics when the traffic incident occurs, wherein the time characteristics are judged whether the incident time is in the morning and evening peak time periods of 06:00 to 09:00 and 16:00 to 19:00, if the incident time is in the morning and evening peak time periods, the data value of the time characteristics is set to be 1, and if not, the data value of the time characteristics is 0;
and 1.3, splicing all data items in each road to serve as attribute characteristic data of each road.
3. The method according to claim 1, wherein the step 2 comprises the following steps:
step 2.1, taking road sections in an actual road network as nodes V in a graph network structure G (V, E), taking intersection points of roads in the actual road network as edges E of the graph network, and mapping the actual road network into a graph network structure G;
2.2, sampling 2-order neighbor nodes of the target node by using a random resampling method;
and 2.3, for the collected set of adjacent nodes of the target node, aggregating the characteristics of the adjacent nodes to the target node in an average aggregation mode, namely, firstly, taking an average value for each dimension of the characteristics of the adjacent nodes, then splicing the average value with the characteristics of the target node, and performing nonlinear transformation after splicing, wherein the calculation formula is as follows:
Figure FDA0002564706610000021
Figure FDA0002564706610000022
where v is the arriving neighbor node that the target node u randomly samples,
Figure FDA0002564706610000023
representing the current characteristics of the vertices, mean is expressed as an averaging operation,
Figure FDA0002564706610000024
features representing k-th layer neighbor aggregation, sigma represents an activation function Sigmoid function, WkExpressing optimization parameters, and expressing CONCAT as characteristic splicing operation;
step 2.4, after the expression vectors of the local domain information characteristics of the nodes are obtained, training is carried out in an unsupervised mode, the learning rate is initialized to be 0.00001, the degree of the maximum node is set to be 7, the dimension of the embedded vector is set to be 256 dimensions, the number of samples selected in each training is 20, and the loss function of the unsupervised function is expressed as:
Figure FDA0002564706610000025
wherein Z isuRepresenting an embedding vector generated by a graph embedding algorithm of GraphSAGE for a target node u, sigma representing an activation function Sigmoid function, v being an arriving neighbor node randomly sampled by the target node u, Pn(v) Is the probability distribution of the negative samples, vn～Pn(v) Denoted as node vnNegative sample distribution P from target node unSampled, Q is the number of samples sampled; by forward propagationObtaining an embedded vector Z of a target node uuThen, a back propagation optimization parameter W is carried out through a gradient descent algorithmkAnd an aggregation function internal parameter.
4. The method according to claim 1, wherein the step 3 comprises the following steps:
using the vector obtained by the road with the traffic incident through the map embedding algorithm obtained in the step 2, reducing the dimensionality to two dimensions through a t-SNE method, using two-dimensional data as sample points, fitting the input sample points through a kernel density estimation method by adopting a smooth kernel function to obtain a fitted probability density model, and finally simulating the real probability distribution through the fitted probability density model, wherein the calculation formula is as follows:
Figure FDA0002564706610000031
where f is the probability density function, x1，x2，x3，…xiIs n independent and equally distributed sample points, kernelfunctionRepresenting a kernel function, using a gaussian kernel density estimate as a kernel function for the density estimate; h is a smoothing coefficient that controls the degree of smoothing of the fitted curve.
5. The method according to claim 1, wherein the step 4 comprises the following steps: and (3) counting attribute characteristic data of roads within 2 hours before the road with the probability of the occurrence of the traffic incident needing to be predicted, embedding the attribute characteristic data into a trained graph embedding model to obtain data vectors of the roads, reducing the dimensionality to two dimensions, inputting the data vectors into a fitted probability density model to obtain the probability value of the occurrence of the traffic incident of the road points, and carrying out visual display on the result in a mode of a kernel density estimation graph.
6. The graph-embedding-based traffic road event prediction method as claimed in claim 5, wherein a "+" scale label on the abscissa of the kernel density estimation graph represents the projection of each sample road on the abscissa axis, the corresponding road is represented by a dot on the kernel density estimation graph, and the corresponding ordinate height of the dot represents the probability value of the traffic accident occurring on the corresponding road.
CN202010623828.8A 2020-07-01 2020-07-01 Traffic road event prediction method based on graph embedding Pending CN111768625A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010623828.8A CN111768625A (en) 2020-07-01 2020-07-01 Traffic road event prediction method based on graph embedding

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010623828.8A CN111768625A (en) 2020-07-01 2020-07-01 Traffic road event prediction method based on graph embedding

Publications (1)

Publication Number Publication Date
CN111768625A true CN111768625A (en) 2020-10-13

Family

ID=72723249

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010623828.8A Pending CN111768625A (en) 2020-07-01 2020-07-01 Traffic road event prediction method based on graph embedding

Country Status (1)

Country Link
CN (1) CN111768625A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112562312A (en) * 2020-10-21 2021-03-26 浙江工业大学 GraphSAGE traffic network data prediction method based on fusion characteristics
CN113077625A (en) * 2021-03-24 2021-07-06 合肥工业大学 Road traffic accident form prediction method
CN113129585A (en) * 2021-03-05 2021-07-16 浙江工业大学 Road traffic flow prediction method based on graph aggregation mechanism of reconstructed traffic network
CN116665456A (en) * 2023-07-27 2023-08-29 交通运输部科学研究院 Method for evaluating traffic state by combining high-dimensional index dimension reduction processing

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101751782A (en) * 2009-12-30 2010-06-23 北京大学深圳研究生院 Crossroad traffic event automatic detection system based on multi-source information fusion
CN101976504A (en) * 2010-10-13 2011-02-16 北京航空航天大学 Multi-vehicle video tracking method based on color space information
CN107610469A (en) * 2017-10-13 2018-01-19 北京工业大学 A kind of day dimension regional traffic index forecasting method for considering multifactor impact
CN109753589A (en) * 2018-11-28 2019-05-14 中国科学院信息工程研究所 A kind of figure method for visualizing based on figure convolutional network
US20190204100A1 (en) * 2017-12-29 2019-07-04 ANI Technologies Private Limited Method and system for predicting traffic conditions
KR102030583B1 (en) * 2017-11-23 2019-10-11 (주)에이텍티앤 Artificial intelligence based traffic accident prediction system and method
CN110895878A (en) * 2019-10-09 2020-03-20 浙江工业大学 Traffic state virtual detector generation method based on GE-GAN
CN111145535A (en) * 2019-11-28 2020-05-12 银江股份有限公司 Travel time reliability distribution prediction method under complex scene

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101751782A (en) * 2009-12-30 2010-06-23 北京大学深圳研究生院 Crossroad traffic event automatic detection system based on multi-source information fusion
CN101976504A (en) * 2010-10-13 2011-02-16 北京航空航天大学 Multi-vehicle video tracking method based on color space information
CN107610469A (en) * 2017-10-13 2018-01-19 北京工业大学 A kind of day dimension regional traffic index forecasting method for considering multifactor impact
KR102030583B1 (en) * 2017-11-23 2019-10-11 (주)에이텍티앤 Artificial intelligence based traffic accident prediction system and method
US20190204100A1 (en) * 2017-12-29 2019-07-04 ANI Technologies Private Limited Method and system for predicting traffic conditions
CN109753589A (en) * 2018-11-28 2019-05-14 中国科学院信息工程研究所 A kind of figure method for visualizing based on figure convolutional network
CN110895878A (en) * 2019-10-09 2020-03-20 浙江工业大学 Traffic state virtual detector generation method based on GE-GAN
CN111145535A (en) * 2019-11-28 2020-05-12 银江股份有限公司 Travel time reliability distribution prediction method under complex scene

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112562312A (en) * 2020-10-21 2021-03-26 浙江工业大学 GraphSAGE traffic network data prediction method based on fusion characteristics
CN113129585A (en) * 2021-03-05 2021-07-16 浙江工业大学 Road traffic flow prediction method based on graph aggregation mechanism of reconstructed traffic network
CN113129585B (en) * 2021-03-05 2022-03-01 浙江工业大学 Road traffic flow prediction method based on graph aggregation mechanism of reconstructed traffic network
CN113077625A (en) * 2021-03-24 2021-07-06 合肥工业大学 Road traffic accident form prediction method
CN113077625B (en) * 2021-03-24 2022-03-15 合肥工业大学 Road traffic accident form prediction method
CN116665456A (en) * 2023-07-27 2023-08-29 交通运输部科学研究院 Method for evaluating traffic state by combining high-dimensional index dimension reduction processing
CN116665456B (en) * 2023-07-27 2023-12-19 交通运输部科学研究院 Method for evaluating traffic state by combining high-dimensional index dimension reduction processing

Similar Documents

Publication Publication Date Title
CN111768625A (en) Traffic road event prediction method based on graph embedding
WO2022241802A1 (en) Short-term traffic flow prediction method under complex road network, storage medium, and system
CN112561146B (en) Large-scale real-time traffic flow prediction method based on fuzzy logic and depth LSTM
Sun et al. A machine learning method for predicting driving range of battery electric vehicles
CN113487066B (en) Long-time-sequence freight volume prediction method based on multi-attribute enhanced graph convolution-Informer model
CN114944053B (en) Traffic flow prediction method based on space-time hypergraph neural network
Yang et al. Application of data science technologies in intelligent prediction of traffic congestion
Tian et al. A network traffic prediction method based on IFS algorithm optimised LSSVM
US20240143999A1 (en) Multi-modal data prediction method based on causal markov model
CN115204477A (en) Bicycle flow prediction method of context awareness graph recursive network
Zhai et al. Optimization of traffic congestion management in smart cities under bidirectional long and short-term memory model
CN112562312B (en) GraphSAGE traffic network data prediction method based on fusion features
Li et al. Short term passenger flow forecast of metro based on inbound passenger plow and deep learning
Miao et al. A queue hybrid neural network with weather weighted factor for traffic flow prediction
CN115985132A (en) Roadside parking availability prediction method based on graph neural network
SHI et al. Simulation and prediction of streamflow based on phase space reconstruction and deep learning algorithm
Pu et al. Spatial-temporal traffic congestion prediction based on attention mechanism
CN111653087A (en) Construction method of urban traffic anomaly detection model
CN113469504A (en) Dynamic testing method, medium and system for water traffic risk
Chen et al. Hybrid spatio-temporal graph convolution network for short-term traffic forecasting
CN112183597A (en) Unbalanced traffic accident severity analysis method based on spectral clustering
Zhang et al. Traffic Speed Prediction Method Based on Spatiotemporal Sampling and LSTM
Wang et al. Social networks and railway passenger capacity: An empirical study based on text mining and deep learning
Zhang et al. Kalman Filter-Based CNN-BiLSTM-ATT Model for Traffic Flow Prediction.
CN115909746A (en) Traffic flow prediction method, system and medium based on federal learning

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20201013

RJ01 Rejection of invention patent application after publication