CN107919016B - Traffic flow parameter missing filling method based on multi-source detector data - Google Patents
Traffic flow parameter missing filling method based on multi-source detector data Download PDFInfo
- Publication number
- CN107919016B CN107919016B CN201711128003.3A CN201711128003A CN107919016B CN 107919016 B CN107919016 B CN 107919016B CN 201711128003 A CN201711128003 A CN 201711128003A CN 107919016 B CN107919016 B CN 107919016B
- Authority
- CN
- China
- Prior art keywords
- traffic flow
- filling
- flow parameters
- source detector
- missing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G08G1/0125—Traffic data processing
- G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
-
- 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
- G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
- G08G1/0116—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from roadside infrastructure, e.g. beacons
Landscapes
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a traffic flow parameter missing filling method based on multi-source detector data, and particularly relates to the technical field of intelligent traffic, wherein the traffic flow parameter missing filling method comprises the following steps: s01, loading historical traffic flow parameters of a road network into the multi-source detector, and calculating and determining an effective road section set; s02, constructing a Feature set of traffic parameters of the multi-source detector aiming at the effective road section set, wherein the Feature set is a traffic parameter Feature matrix Feature which consists of n rows and z columns; s03, training a filling model for representing traffic flow parameters according to the traffic parameter Feature matrix Feature; and S04, calculating the corresponding missing traffic flow parameters on the source detector through the loaded filling model, and obtaining the missing traffic flow parameter filling estimation values. The invention is suitable for various traffic parameter missing estimation of various sources, and can fill up missing data in real time.
Description
Technical Field
The invention relates to the technical field of intelligent traffic, in particular to a traffic flow parameter missing filling method based on multi-source detector data.
Background
In an intelligent traffic system, real-time traffic flow parameters are the basis for traffic management and control, traffic guidance, and traffic information services. At present, fixed detection equipment mainly used for detecting real-time traffic flow parameters comprises a coil detector, an electric police card port, a video detector and the like. The traffic parameter types obtained by different detection sources are different, for example, a coil detector can obtain the lane occupancy, but an electric alarm card cannot obtain the lane occupancy; meanwhile, the same traffic parameters obtained by different detectors are different in definition, for example, the traffic flow speed obtained by the coil detector is the section instantaneous speed, and the traffic flow speed calculated by the data of the electric alarm card port is the road section travel speed. According to the basic information of the road section and the installation position of the detector, certain correlation exists between traffic flow parameters acquired by different detectors.
In the process of detecting real-time traffic flow parameters, data loss of detection equipment caused by construction damage, line faults, processing errors, equipment maintenance or installation coverage rate and the like is a common phenomenon in dynamic traffic data acquisition. The missing value can bring adverse effects to the analysis of traffic flow data and deep data mining, so that the filling quality of the missing data has important significance to the accuracy of traffic information.
In order to fill up missing traffic flow parameter data, technicians develop various estimation methods for filling up missing data, such as a historical averaging method and an interpolation method, and the two methods are simple to operate but cannot adapt to real-time traffic changes. Other methods, such as machine learning, require integrity of the input field during model training or real-time application of the model, and are therefore difficult to engineer. In addition, the existing method is mainly used for filling and recovering missing data of a single traffic parameter based on single-source detection data, and does not consider the connection between different source detection data and the universality of different types of traffic parameters by a filling method of different missing data.
Disclosure of Invention
The invention provides a traffic flow parameter missing filling method based on multi-source detector data, which has the following advantages: the method is suitable for the missing estimation of various traffic parameters of various sources, fills the missing data of various traffic flow parameters of various sources in real time, and provides basic guarantee for analyzing and predicting subsequent traffic flow data, managing and controlling traffic and inducing traffic.
In order to achieve the purpose, the invention provides the following technical scheme:
the traffic flow parameter missing filling method based on the multi-source detector data comprises the following steps:
s01, loading historical traffic flow parameters of a road network into the multi-source detector, and calculating and determining an effective road section set;
s02, constructing a Feature set of multi-source detector traffic parameters aiming at the effective road section set, wherein the Feature set is a traffic parameter Feature matrix Feature which is composed of n rows and z columns;
n is the total time segment number, n is L24 60/T, z can be respectively and independently expressed as a road section attribute parameter, a traffic flow parameter measured by each source detector in the road section, a traffic flow parameter measured by each source detector in the peripheral road section and a historical traffic parameter; l: counting the number of days of historical traffic flow parameters, T: counting intervals of traffic flow parameters;
s03, respectively training a filling model for representing corresponding traffic flow parameters according to the traffic flow parameters of each type of each source in the traffic parameter Feature matrix;
and S04, calculating the corresponding missing traffic flow parameters on the source detector by loading the filling model, and obtaining the missing traffic flow parameter filling estimation value.
Preferably, in step S01, the valid road segment set includes a plurality of valid road segments, where the valid road segments are road segments in which the traffic flow parameters recorded by different source detectors in L days are all larger than C; wherein, C is 24 × 60/T.
Preferably, in step S02, the link attribute parameters include a one-way link number, a link length, a one-way lane number, a road class, and a distance of each source detector from the link exit.
Preferably, in step S02, the traffic flow parameters measured by the source detectors in the road section are arranged into a time sequence according to the occurrence sequence of the traffic flow parameters in the time period T; if data loss caused by equipment failure exists in the historical data, filling the traffic flow parameters in the corresponding z column with-1 to show that data loss exists in the T time period; if the type of source detector is not installed in the corresponding road section, all traffic flow parameters in the corresponding z column are filled by-1, and no data is represented.
Preferably, in step S02, the source detector traffic flow parameters of the peripheral links include upstream source detector traffic flow parameters and downstream source detector traffic flow parameters; and when the upstream road section or the downstream road section is a three-way intersection, filling the corresponding traffic flow parameters in the z row with-1.
Preferably, the historical traffic parameters described in step S02 include the following two types:
the method comprises the following steps of firstly, representing traffic flow parameters corresponding to a t time period of a long-term historical rule;
and secondly, translating 1, 2 and 3T forwards at the study time T for representing the short-term historical rule respectively to correspondingly represent various traffic flow parameters of each source of the current road section or the peripheral road section at the time T-1, T-2 and T-3 corresponding to the studied time period T, and filling a missing part at the tail with-1, wherein T represents the studied time period.
Preferably, in step S03, the traffic parameter of the single source and the traffic class is y, the input feature is x, and the training set of the model isRecording the trained filling model as F (x), setting a micro-loss function L (y, F (x)) and the maximum iteration number M, and training the model according to the following steps:
(1) initialization model, F0(x)=0;
(2) For the mth iteration:
(2.1) fitting the CART regression Tree to fm(x);
(2.2) calculating the loss function value Ω(fm) For the complexity of the CART regression tree,wherein K represents the number of leaf nodes of the regression tree, ω represents the score corresponding to each leaf node, and γ and λ are fixed parameters.
Where L is a loss function set to the squared error, i.e., L (y)i,F(xi))=(yi-F(xi))2,0≤M≤500;
(2.3) minimizing the loss function L by linearly searching K and ω;
(2.4) updating model Fm(x)=Fm-1(x)+fm(x);
(3) When the iteration number M is equal to M, the output training is finished to finish the filling model FM(x) And saving to the local storage space.
More preferably, in step S04, the filling model calculates the input feature x to obtain a corresponding feature parameter value requiring missing filling, fills the corresponding feature parameter value with-1 in the absence of the feature parameter value, and inputs the result into the corresponding filling model to obtain a corresponding filling estimation value.
The invention has the beneficial effects that:
in the method provided by the invention, the input data is allowed to have deficiency when the model is constructed and used, and the method is applicable to various traffic parameter deficiency estimation of various sources. And a model is constructed through historical multi-source traffic parameter information, detector installation position information and a traffic flow parameter space-time change rule, missing data of all source traffic flow parameters are filled in real time, and basic guarantee is provided for analysis and prediction, traffic control and induction based on traffic flow data.
Drawings
Fig. 1 is a flow of constructing a traffic flow parameter missing filling model based on multi-source detector data in this embodiment;
FIG. 2 is a flow chart of real-time filling of traffic flow parameters based on data from a multi-source detector in the present embodiment;
fig. 3 is partial data of the traffic parameter Feature matrix Feature in this embodiment.
Detailed Description
The embodiment provides a technical scheme:
as shown in fig. 1-2, the traffic flow parameter missing filling method based on multi-source detector data includes the following steps:
s01, loading historical traffic flow parameters of a road network into the multi-source detector, and calculating and determining an effective road section set; the mentioned effective road section set comprises a plurality of effective road sections, and the effective road sections are road sections of which all traffic flow parameters recorded by different source detectors in L days are greater than C; and C is 24 × 60/T, T is a traffic flow parameter statistical interval, and L is the number of days for counting the historical traffic flow parameters. In this example, L is taken for 30 days.
S02, constructing a Feature set of the traffic parameters of the multi-source detector aiming at the effective road section set, wherein the Feature set is a traffic parameter Feature matrix Feature which is composed of n rows and z columns. As shown in fig. 3, it is part of the traffic parameter Feature matrix Feature.
Wherein, n is the total time segment number, n is L24 60/T, z can be respectively and independently expressed as a road section attribute parameter, a traffic flow parameter measured by each source detector in the road section, a traffic flow parameter measured by each source detector in the peripheral road section and a historical traffic parameter.
z represents a link attribute parameter: the road section attribute parameters include a unidirectional road section number, a road section length, a unidirectional lane number, a road grade and a distance from each source detector to a road section exit, the unit of the distance from each source detector to the road section exit is meter, the road sections mentioned in the embodiment are all unidirectional road sections, that is, if a road section is an east-west bidirectional four-lane road section, the east-west direction and the west-east direction are two different road section numbers, and the lane number attribute is also the unidirectional lane number.
z represents a traffic flow parameter measured by each source detector in the road segment: arranging the traffic flow parameters measured by the source detectors in the road section into a time sequence according to the occurrence sequence of the traffic flow parameters in the T time period; if data loss caused by equipment failure exists in the historical data, filling the traffic flow parameters in the corresponding z column with-1 to show that data loss exists in the T time period; if the type of source detector is not installed in the corresponding road section, all traffic flow parameters in the corresponding z column are filled by-1, and no data is represented.
z represents the traffic flow parameter measured by each source detector of the peripheral road section: the traffic flow parameters of the source detectors of the peripheral road section comprise the traffic flow parameters of the source detectors at the upstream and the traffic flow parameters of the source detectors at the downstream road section; and when the upstream road section or the downstream road section is a three-way intersection, filling the corresponding traffic flow parameters in the z row with-1.
The historical traffic parameters include the following two categories: the method comprises the following steps of firstly, representing traffic flow parameters corresponding to a t time period of a long-term historical rule; and secondly, translating 1, 2 and 3T forwards at the study time T for representing the short-term historical rule respectively to correspondingly represent various traffic flow parameters of each source of the current road section or the peripheral road section at the time T-1, T-2 and T-3 corresponding to the studied time period T, and filling a missing part at the tail with-1, wherein T represents the studied time period. The historical traffic parameters also include traffic flow parameters obtained by simple statistical analysis of the traffic flow parameters measured by the source detectors in the road segments and the traffic flow parameters measured by the source detectors in the surrounding road segments.
z represents a time attribute parameter: the time attribute parameters are the week number, month, hour number and minute number corresponding to the traffic flow parameter statistical interval T starting time. Wherein the daily numbers of Monday-Sunday are 1-7.
And S03, training a traffic flow parameter filling model based on the XGboost algorithm. Training a filling model for representing traffic flow parameters according to the traffic parameter Feature matrix Feature; for each valid road segment mentioned in step S01, a corresponding missing filling model needs to be constructed and trained, and since the history data of the invalid road segments is too little, the missing filling cannot be performed.
Before the training of the model is started, defining a single-source single-class traffic parameter as y and an input feature as x, and then the training set of the model isRecording the trained filling model as F (x), setting a micro-loss function L (y, F (x)) and the maximum iteration number M, and training the model according to the following steps:
(1) initialization model, F0(x)=0;
(2) For the mth iteration:
(2.1) fitting the CART regression Tree to fm(x);
(2.2) calculating the loss function value Ω(fm) For the complexity of the CART regression tree,wherein K represents the number of leaf nodes of the regression tree, ω represents the score corresponding to each leaf node, and γ and λ are fixed parameters. γ and λ are penalty coefficients for L1 and L2 regularization to boost algorithm speed and prevent overfitting. During training, it is desirable that the model to be fitted should be as simple as possible, with a complexity of Ω (f)m) As low as possible. The larger the gamma and the lambda are, the simpler the model tends to be, and the faster the algorithm converges, but obtaining too large results in under-fitting of the model and poor effect. The value range of gamma and lambda is generally [ 0-5%]The default value is 1, and good effect can be obtained generally without adjustment.
Where L is a loss function set to the squared error, i.e., L (y)i,F(xi))=(yi-F(xi))2,0≤M≤500;
(2.3) minimizing the loss function L by linearly searching K and ω;
(2.4) updating model Fm(x)=Fm-1(x)+fm(x);
(3) When the iteration time M is equal to M, namely the iteration time reaches the preset maximum iteration time M, the output training is finished, and the model F is finally filledM(x) And storing the data in a local storage space for subsequent real-time traffic flow parameter filling. In this embodiment, the update period of the model is defaulted to one month, and the update period may also be dynamically adjusted according to the demand and the operation performance, and since the accuracy is reduced due to the fact that the model is greatly affected by seasonal changes due to an excessively long update period, the update period of the model is optimal between 2 weeks and 1 month. The input data at each updating time is historical traffic parameters and roads of the previous T daysThe segment attributes are used as a Feature matrix Feature formed by z columns.
And S04, calculating the corresponding missing traffic flow parameters on the source detector by loading the filling model, and obtaining the missing traffic flow parameter filling estimation value. And calculating the input characteristic x by the filling model to obtain a corresponding characteristic parameter value needing to be subjected to missing filling, filling the missing characteristic parameter value with-1, and inputting the missing characteristic parameter value into the corresponding filling model to obtain a corresponding filling estimation value.
As shown in fig. 3, for example, a loss filling model of the speed (det1_ speed) of the trained bayonet detector is used, if the source detector in the time segment of 8:00 to 8:05 needs to be filled with the loss during use, firstly, the rest characteristic variables except the speed characteristic parameter are used as input values of the model, the value of the input characteristic is x, if the speed characteristic parameter in the corresponding time segment is lost, the model is filled with-1, that is, up1_ det1_ speed _ t-1 is-1, and the model outputs an estimated value of the lost speed in the time segment.
Claims (9)
1. A traffic flow parameter missing filling method based on multi-source detector data is characterized by comprising the following steps:
s01, loading historical traffic flow parameters of a road network into the multi-source detector, and calculating and determining an effective road section set;
s02, constructing a Feature set of multi-source detector traffic parameters aiming at the effective road section set, wherein the Feature set is a traffic parameter Feature matrix Feature which is composed of n rows and z columns;
wherein, n is the total time segment number, n is L24 60/T, z can include road section attribute parameter, traffic flow parameter measured by each source detector in the road section, traffic flow parameter measured by each source detector in the peripheral road section, historical traffic parameter and time attribute parameter; l: counting the number of days of historical traffic flow parameters, T: counting intervals of traffic flow parameters;
s03, respectively training a filling model for representing corresponding traffic flow parameters according to the traffic flow parameters of each type of each source in the traffic parameter Feature matrix;
and S04, calculating the corresponding missing traffic flow parameters on the source detector by loading the filling model, and obtaining the missing traffic flow parameter filling estimation value.
2. The method for filling up missing traffic flow parameters based on multi-source detector data according to claim 1, wherein in step S01, the valid road segment set includes a plurality of valid road segments, and the valid road segments are road segments in which all traffic flow parameters recorded by different source detectors in L days are greater than C.
3. The method for filling up missing traffic flow parameters based on multi-source detector data of claim 1, wherein in step S02, the link attribute parameters include one-way link number, link length, one-way lane number, road grade and distance of each source detector from the link exit.
4. The method for filling up the missing traffic flow parameters based on the multi-source detector data according to claim 1, characterized in that in step S02, the traffic flow parameters measured by each source detector in the road section are arranged into a time sequence according to the sequence of occurrence in the time period T;
if data loss caused by equipment failure exists in the historical data, filling the traffic flow parameters in the corresponding z column with-1 to show that data loss exists in the T time period; if the type of source detector is not installed in the corresponding road section, all traffic flow parameters in the corresponding z column are filled by-1, and no data is represented.
5. The method for filling up missing source detector data based on traffic flow parameters of claim 4, wherein in step S02, the source detector traffic flow parameters of the peripheral road section include upstream source detector traffic flow parameters and downstream source detector traffic flow parameters;
and when the upstream road section or the downstream road section is a three-way intersection, filling the corresponding traffic flow parameters in the z row with-1.
6. The method for filling missing traffic flow parameters based on multi-source detector data according to claim 1, wherein the historical traffic parameters in step S02 include the following two categories:
the method comprises the following steps of firstly, representing traffic flow parameters corresponding to a t time period of a long-term historical rule;
and secondly, translating 1, 2 and 3T forwards at the study time T for representing the short-term historical rule respectively to correspondingly represent various traffic flow parameters of each source of the current road section or the peripheral road section at the time T-1, T-2 and T-3 corresponding to the studied time period T, and filling a missing part at the tail with-1, wherein T represents the studied time period.
7. The method for filling up missing traffic flow parameters based on multi-source detector data according to claim 1, wherein in step S02, the time attribute parameters are the week number, month, hour number, and minute number corresponding to the start time of the traffic flow parameter statistical interval T.
8. The method for filling in missing traffic flow parameters based on multi-source detector data of claim 1, wherein in step S03, the single-source single-type traffic parameter is y, the input feature is x, and the training set of the model isRecording the trained filling model as F (x), setting a micro-loss function L (y, F (x)) and the maximum iteration number M, and training the model according to the following steps:
(1) initialization model, F0(x)=0;
(2) For the mth iteration:
(2.1) fitting the CART regression Tree to fm(x);
(2.2) calculating the loss function value Ω(fm) For the complexity of the CART regression tree,wherein K represents the number of leaf nodes of the regression tree, omega represents the corresponding fraction of each leaf node, and gamma and lambda are fixed parameters;
where L is a loss function set to the squared error, i.e., L (y)i,F(xi))=(yi-F(xi))2,0≤M≤500;
(2.3) minimizing the loss function L by linearly searching K and ω;
(2.4) updating model Fm(x)=Fm-1(x)+fm(x);
(3) When the iteration number M is equal to M, the output training is finished to finish the filling model FM(x) And saving to the local storage space.
9. The method for filling missing traffic flow parameters based on multi-source detector data of claim 8, wherein in step S04, the filling model calculates input features x to obtain corresponding feature parameter values to be filled with missing data, and the missing feature parameter values are filled with-1 and input into the corresponding filling model to obtain corresponding filling estimation values.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711128003.3A CN107919016B (en) | 2017-11-15 | 2017-11-15 | Traffic flow parameter missing filling method based on multi-source detector data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711128003.3A CN107919016B (en) | 2017-11-15 | 2017-11-15 | Traffic flow parameter missing filling method based on multi-source detector data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107919016A CN107919016A (en) | 2018-04-17 |
CN107919016B true CN107919016B (en) | 2020-02-18 |
Family
ID=61896367
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711128003.3A Active CN107919016B (en) | 2017-11-15 | 2017-11-15 | Traffic flow parameter missing filling method based on multi-source detector data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107919016B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109166309B (en) * | 2018-08-06 | 2021-03-19 | 重庆邮电大学 | Missing traffic data recovery method for complex urban traffic network |
CN110555989B (en) * | 2019-08-16 | 2021-10-26 | 华南理工大学 | Xgboost algorithm-based traffic prediction method |
CN110993100B (en) * | 2019-11-06 | 2023-01-03 | 北京理工大学 | Missing value filling method of juvenile and child myopia prediction system and system using same |
CN111008661B (en) * | 2019-12-04 | 2021-03-09 | 哈尔滨工业大学 | Croston-XGboost prediction method for reserve demand of aircraft engine |
CN111785014B (en) * | 2020-05-26 | 2021-10-29 | 浙江工业大学 | Road network traffic data restoration method based on DTW-RGCN |
CN116206443B (en) * | 2023-02-03 | 2023-12-15 | 重庆邮电大学 | Traffic flow data interpolation method based on time-space road network pixelized representation |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102881162A (en) * | 2012-09-29 | 2013-01-16 | 北京市交通信息中心 | Data processing and fusion method for large-scale traffic information |
CN103093621A (en) * | 2013-01-07 | 2013-05-08 | 北京世纪高通科技有限公司 | Processing method and device of multisource traffic information fusion |
CN103247177B (en) * | 2013-05-21 | 2016-01-20 | 清华大学 | Large-scale road network traffic flow real-time dynamic prediction system |
CN106408184A (en) * | 2016-09-12 | 2017-02-15 | 中山大学 | User credit evaluation model based on multi-source heterogeneous data |
CN106781457A (en) * | 2016-11-29 | 2017-05-31 | 东南大学 | A kind of freeway traffic flow parameter correction method based on multi-source fusion data |
-
2017
- 2017-11-15 CN CN201711128003.3A patent/CN107919016B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102881162A (en) * | 2012-09-29 | 2013-01-16 | 北京市交通信息中心 | Data processing and fusion method for large-scale traffic information |
CN103093621A (en) * | 2013-01-07 | 2013-05-08 | 北京世纪高通科技有限公司 | Processing method and device of multisource traffic information fusion |
CN103247177B (en) * | 2013-05-21 | 2016-01-20 | 清华大学 | Large-scale road network traffic flow real-time dynamic prediction system |
CN106408184A (en) * | 2016-09-12 | 2017-02-15 | 中山大学 | User credit evaluation model based on multi-source heterogeneous data |
CN106781457A (en) * | 2016-11-29 | 2017-05-31 | 东南大学 | A kind of freeway traffic flow parameter correction method based on multi-source fusion data |
Non-Patent Citations (2)
Title |
---|
An Improved XGBoost Based on Weighted Column Subsampling for Object Classification;Xin Gao,Shaohua Fan,Xinpeng Li et al.;《2017 4th international conference on systems and informaics(ICSAI)》;20171113;全文 * |
Fusing Incomplete Multisensor Heterogeneous Data to Estimate Urban Traffic;Zhenyu Shan,Yingjie Xia,Peipei Hou and Jifeng He;《IEEE MultiMedia》;20160525;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN107919016A (en) | 2018-04-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107919016B (en) | Traffic flow parameter missing filling method based on multi-source detector data | |
CN112241814B (en) | Traffic prediction method based on reinforced space-time diagram neural network | |
CN108288096B (en) | Method and device for estimating travel time and training model | |
CN108197739B (en) | Urban rail transit passenger flow prediction method | |
CN111223301B (en) | Traffic flow prediction method based on graph attention convolution network | |
US10636293B2 (en) | Uncertainty modeling in traffic demand prediction | |
CN113591380B (en) | Traffic flow prediction method, medium and equipment based on graph Gaussian process | |
CN114299723B (en) | Traffic flow prediction method | |
CN104021674A (en) | Method for rapidly and accurately forecasting travel time of vehicles for passing through road sections | |
CN111047078B (en) | Traffic characteristic prediction method, system and storage medium | |
CN112242060A (en) | Traffic flow prediction method and apparatus, computer device, and readable storage medium | |
CN115410372B (en) | Reliable prediction method for highway traffic flow based on Bayesian LSTM | |
CN113223293B (en) | Road network simulation model construction method and device and electronic equipment | |
CN116386336B (en) | Road network traffic flow robust calculation method and system based on bayonet license plate data | |
CN110543978A (en) | Traffic flow data prediction method and device based on wavelet neural network | |
CN111785093A (en) | Air traffic flow short-term prediction method based on fractal interpolation | |
CN111612274A (en) | Tidal water level forecasting method based on space-time correlation | |
CN116824851A (en) | Path-based urban expressway corridor traffic jam tracing method | |
CN111160594A (en) | Method and device for estimating arrival time and storage medium | |
CN112797994A (en) | Method for determining estimated arrival time of route, and related device and server | |
CN116304969A (en) | Vehicle track multi-mode prediction method considering road information based on LSTM-GNN | |
CN114692738A (en) | Lightweight real-time series anomaly detection method | |
CN114399901A (en) | Method and equipment for controlling traffic system | |
CN113515890A (en) | Renewable energy day-ahead scene generation method based on federal learning | |
CN113807556A (en) | Tourism index prediction method, device, equipment and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20200115 Address after: Room 701-1, 7 / F, building 1, No. 1326, Wenyi West Road, Cangqian street, Yuhang District, Hangzhou City, Zhejiang Province Applicant after: HANGZHOU YUANTIAO TECHNOLOGY CO., LTD. Address before: Room 73, No. 34, No. 2000, Tianmu Mountain Road, Xihu District, Hangzhou, Zhejiang Applicant before: Xia Yingjie |
|
GR01 | Patent grant | ||
GR01 | Patent grant |