CN113065684A - Expressway travel time prediction method based on VAE and deep learning combined model - Google Patents

Expressway travel time prediction method based on VAE and deep learning combined model Download PDF

Info

Publication number
CN113065684A
CN113065684A CN202110200510.3A CN202110200510A CN113065684A CN 113065684 A CN113065684 A CN 113065684A CN 202110200510 A CN202110200510 A CN 202110200510A CN 113065684 A CN113065684 A CN 113065684A
Authority
CN
China
Prior art keywords
data
travel time
vae
deep learning
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110200510.3A
Other languages
Chinese (zh)
Inventor
于海洋
张浩洋
任毅龙
于海生
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.)
JILIN PROVINCIAL EXPRESSWAY ADMINISTRATION
Beihang University
Original Assignee
JILIN PROVINCIAL EXPRESSWAY ADMINISTRATION
Beihang 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 JILIN PROVINCIAL EXPRESSWAY ADMINISTRATION, Beihang University filed Critical JILIN PROVINCIAL EXPRESSWAY ADMINISTRATION
Priority to CN202110200510.3A priority Critical patent/CN113065684A/en
Publication of CN113065684A publication Critical patent/CN113065684A/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
    • 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
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G07—CHECKING-DEVICES
    • G07B—TICKET-ISSUING APPARATUS; FARE-REGISTERING APPARATUS; FRANKING APPARATUS
    • G07B15/00—Arrangements or apparatus for collecting fares, tolls or entrance fees at one or more control points
    • G07B15/06—Arrangements for road pricing or congestion charging of vehicles or vehicle users, e.g. automatic toll systems

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • General Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Educational Administration (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Quality & Reliability (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Game Theory and Decision Science (AREA)
  • Software Systems (AREA)
  • Operations Research (AREA)
  • Primary Health Care (AREA)
  • Finance (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Traffic Control Systems (AREA)

Abstract

The patent discloses a highway travel time prediction method based on a VAE and deep learning combined model, which comprises the following steps: preprocessing highway toll data, removing missing and abnormal data, and arranging the processed data into travel time sequence data; step two, carrying out stabilization processing on the travel time data; and step three, training the travel time data by utilizing a recurrent neural network (LSTM), and predicting the travel time. The invention utilizes the process of VAE algorithm coding and decoding to extract the data according to the condition distribution of the data, and carries out stabilization processing on the data on the premise of not influencing the time sequence characteristics of the data. And secondly, a circulating neural network for deep learning is combined, so that long time sequence data can be well processed, and the prediction precision is improved.

Description

Expressway travel time prediction method based on VAE and deep learning combined model
Technical Field
The invention relates to travel time prediction, in particular to a highway travel time prediction method based on a VAE and deep learning combined model.
Background
Travel time is an important measure of highway operating conditions and refers to the time from the entry of a toll station to the exit of the toll station. Accurate prediction of travel time may provide great help for traveler's routing and even traffic guidance. For travel time estimation, there are many methods that utilize fused multi-source data, such as fixed coil and floating car data. The data targeted by the invention is highway toll data.
The highway toll data directly records the time of arrival and departure of the vehicle, so that the time series of travel times between toll stations can be obtained after processing, and the prediction of the time series is essentially carried out. The traditional method for predicting the time series is mainly a Box-Jenkins method, which firstly determines the appropriate p, d and q values in an ARIMA model, then estimates the specific parameter values of the model by an estimation method, such as maximum likelihood estimation, and finally checks the appropriateness of the model and improves the model. In recent years, with the development of machine learning and deep learning, the number of prediction methods based on support vector machines is increasing, and in the aspect of deep learning, what is well performed for time series prediction is a long-time memory network (LSTM), which has great advantages for processing long-sequence data and time-sequence data.
However, most data have the characteristic of instability for the time series, the instability of the data has a great influence on the accuracy of prediction, and the method for solving the instability of the data is to perform differential processing on the data, but the effect of the method is general, so that how to smooth the data is also very important when the time series is predicted.
Disclosure of Invention
The invention aims to provide a highway travel time prediction method based on a VAE and deep learning combined model.
The technical solution for realizing the purpose of the invention is as follows: a travel time prediction method based on highway charging data comprises the following steps:
s1: preprocessing highway toll data, removing missing and abnormal data, and arranging the processed data into travel time sequence data;
s2: carrying out stabilization processing on the travel time data;
s3: and training the travel time data by using a recurrent neural network (LSTM), and predicting the travel time.
Further, the data preprocessing in step S1 includes the following steps:
s11: removing missing data in the highway toll data;
s12: for the remaining complete highway toll data, subtracting the station-entering time from the station-exiting time to obtain the travel time data among stations;
s13: calculating mean values for travel time data between sites
Figure RE-GDA0003043550420000021
And standard deviation of
Figure RE-GDA0003043550420000022
And are provided with
Figure RE-GDA0003043550420000023
Exception data is culled for the range.
Further, the step S2 of smoothing the travel time includes the following steps:
s21: fitting function f using interstation travel time data1And f2Mean value ukSum variance
Figure RE-GDA0003043550420000025
A sequence;
s22: sampling random errors from the standard normal distribution, and adding the random errors into the calculation of the hidden variable Z;
s23: sampling in conditional distribution by using hidden variable Z to obtain new sample
Figure RE-GDA0003043550420000026
S24: iterating the whole network by using a gradient descent method to obtain a time sequence after final restoration
Figure RE-GDA0003043550420000027
The formula for the gradient descent is as follows:
Figure RE-GDA0003043550420000024
further, the training of the travel time data by using the recurrent neural network LSTM in step S3 and the predicting of the travel time include the following steps.
S31: dividing a data set into a training set and a testing set according to the proportion of 8: 2;
s32: setting parameters and putting the parameters into an LSTM model for training;
s33: and predicting the travel time according to the trained model.
The invention utilizes the process of VAE algorithm coding and decoding to extract the data according to the condition distribution of the data, and carries out stabilization processing on the data on the premise of not influencing the time sequence characteristics of the data. And secondly, a circulating neural network for deep learning is combined, so that long time sequence data can be well processed, and the prediction precision is improved.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a schematic illustration of highway toll data content;
FIG. 3 is a schematic diagram of the VAE algorithm;
FIG. 4 is a diagram of an LSTM single memory cell.
Detailed Description
The invention is further illustrated with reference to the accompanying drawings:
the method of the invention has a flow diagram as shown in the attached figure 1, and the specific contents are as follows:
the charging data in this example is from 17 highways in Jilin province. 5680799 rows of 157 toll booths in total. The time span is from 2018, 9 and 10 days to 2018, 10 and 4 days, the charging data of common working days, weekends and national festivals are included, and the types are enriched. Mainly comprises the following characteristics: license plate number, vehicle type, vehicle arrival name, vehicle arrival time, vehicle departure name, and vehicle departure time, as shown in fig. 2.
The method comprises the following steps: and (4) preprocessing the charging data.
(1) And missing data is eliminated, wherein the missing data refers to data with partial missing of six characteristic values of the charging information, only the license plate number does not influence the prediction of the subsequent travel time, and the data with missing of other five characteristic values is eliminated.
(2) The travel time data set is sorted, because the content of the invention is travel time prediction, the travel time information needs to be extracted from the charging data in advance, the data with the same name for the vehicle to enter and exit the charging station are classified into one class firstly, the class is the travel time data set between the next two stations, the specific travel time is the departure time minus the arrival time, and in addition, the data can be classified according to the type of the vehicle.
(3) The abnormal data is selected, wherein the abnormal data refers to that the travel time between two stations calculated by the data is too large or too small compared with the data of most vehicles, which may be caused by factors such as errors, midway parking of the vehicles, too fast or too slow speed of individual vehicles and the like when the system records the data, and the small amount of abnormal data is called noise. The noise should be removed by the process.
Firstly, a minimum threshold value of travel time is calculated, the speed limit of the expressway for the small vehicles is 120km/h, therefore, the minimum travel time threshold value is equal to the distance/speed limit, and data with the travel time smaller than the threshold value is removed by taking the minimum travel time threshold value as a standard.
Secondly, calculating the average travel time according to the travel time of the process statistics in the step (2)
Figure RE-GDA0003043550420000031
The specific calculation formula is as follows:
Figure RE-GDA0003043550420000032
wherein i and j denote the numbers of the inbound and outbound toll stations, respectively, tnijRepresenting the time it takes for vehicle n to enter the station from the i toll station and the exit station from the j toll station. N denotes the number of vehicles traveling between toll stations i and j.
Calculating the standard deviation S of the travel time as:
Figure RE-GDA0003043550420000041
to be provided with
Figure RE-GDA0003043550420000042
And screening data for the range, if the data is not in the range, rejecting the data, and recalculating the average value and standard deviation of the data set until the screened data set does not have data outside the range. And dividing the travel time data after being screened by taking 15min as a time period until the data is preprocessed.
Step two: travel time data reconstruction repair
A historical data set about travel time may be obtained from step one, taking data between a pair of toll stations as an example, and defining X ═ X1,X2,…,XnN is the number of data sets. The data set has a time series of features.
The general stability detection is that unit root inspection is carried out on the time sequence by ADF, but most of the time sequences have no good stability, so that certain reconstruction and repair are needed.
Fig. 3 shows a schematic diagram of the VAE algorithm for repairing time-series data, and first, an original sample X is set to { X ═ X1,X2,…,XnCoding, extracting partial characteristics and obtaining an implicit variable Z ═ Z1,Z2,…,ZnAnd decoding the sequence Z to obtain
Figure RE-GDA0003043550420000043
The encoding and decoding are performed by fitting parameters through a neural network, and the encoding process is performed on the data X through the neural networkk(k-1, 2, …, n) average value ukSum variance
Figure RE-GDA0003043550420000044
The sequence, the formula is as follows:
uk＝f1(Xk)
Figure RE-GDA0003043550420000045
wherein the function f1And f2Namely a nonlinear function fitted by the neural network, and sub-sequences which are all in Gaussian distribution are obtained through representation.
The formula for computing the latent variable Z is:
Zk＝μk+σk⊙ε
wherein Z iskTo correspond to XkThe hidden variable of (a) is a hadamard product, and ε is a random error sampled from a normal distribution.
Then the condition distribution can be found through the original sample X and the hidden variable Z
Figure RE-GDA0003043550420000046
Sampling from the conditional distribution, the decoder network generating new samples
Figure RE-GDA0003043550420000051
The decoder will in turn optimize the mean ukSum variance
Figure RE-GDA0003043550420000052
And learning according to the gradient descending direction of the network, and finishing gradient descending by differentiating the hidden variable. The formula is as follows:
Figure RE-GDA0003043550420000053
wherein Z, u, σ generally refers to Z abovek,uk,σk。
The value of epsilon is continuously changed through gradient reduction of the neural network, and the time sequence after final repair is obtained through multiple rounds of iteration
Figure RE-GDA0003043550420000054
When the traditional difference processing method is used for processing data, if the data stability is poor, multiple times of difference processing may be required, and the more the difference times are, the larger the calculation amount is brought to the subsequent prediction process. Compared with the traditional differential processing, the VAE algorithm processes the data, the data processing result is more direct, and the calculation amount of the subsequent prediction process can be reduced.
Step three: training sample data of recurrent neural network and predicting
The long-short term memory network (LSTM) is a circulating neural network with a special structure, can solve the problems of gradient disappearance and gradient explosion which can occur in the training process of a long sequence, and has better performance in processing the long-term sequence compared with the traditional circulating neural network.
Through the first step and the second step, two-dimensional n multiplied by m travel time data can be obtained
Figure RE-GDA0003043550420000055
Where n represents n time slots divided according to 15min as described in step one, and m represents m inbound and outbound pairs present in the data. The data in the nth row and mth column represents the average travel time of the mth inbound outbound pair in the nth time step, and is input into the LSTM network, and the network parameters "input size 5, batch size 10, output size 1" are set, and the data set is divided into training set and test set at a ratio of 8: 2. The LSTM structure is shown in the figure.
The LSTM maintains and updates state through the forget gate, the input gate, and the output gate.
Wherein forget in the door:
ft＝σ(Wf·[ht-1,xt]+bf)
wherein the sigma function is a sigmoid activation function. WfAnd bfTo forget the door parameter, ht-1For data after t-1 neuron processing, xtData input for the t-th neuron.
In the input gate:
it＝σ(Wi·[ht-1,xt]+bi)
Figure RE-GDA0003043550420000061
wherein the tan h function is the tan h activation function, Wi、bi、WCAnd bCRespectively, the corresponding parameters of the input gate. i.e. itAnd
Figure RE-GDA0003043550420000062
the data output from the input gate in the t-th neuron.
In the output gate:
ot＝σ(Wo·[ht-1,xt]+bo)
ht＝ot*tanh(Ct)
in the formula, WoAnd boTo output the gate parameter, otThe data output by the output gate in the t-th neuron.
The LSTM update state formula is:
Figure RE-GDA0003043550420000063
wherein, CtRepresents the state of the t-th neuron, htThat is, the LSTM network is formed by connecting a plurality of LSTM units, and each unit inputs the current time step length and finally outputs htThe result is the predicted result. LSTM can be paired
Figure RE-GDA0003043550420000064
The two-dimensional data structure is directly processed, and travel time predicted values of m inbound and outbound pairs with the same plurality of time steps can be output.
The LSTM algorithm not only solves the problems of gradient disappearance and gradient explosion which may occur in the training process of a long sequence, but also is very suitable for the patent because the highway toll data volume is not large, and compared with other neural network algorithms, the LSTM algorithm has a simple structure, short training time, a long-term memory function and simple realization.

Claims (3)

1. A highway travel time prediction method based on a VAE and deep learning combined model is characterized by comprising the following steps:
preprocessing highway toll data, removing missing and abnormal data, and arranging the processed data into travel time sequence data;
step two, carrying out stabilization processing on the travel time data; s21: fitting function f using interstation travel time data1And f2Mean value ukSum variance
Figure FDA0002948561190000011
A sequence; s22: sampling random errors from the standard normal distribution, and adding the random errors into the calculation of the hidden variable Z; s23: sampling in conditional distribution by using hidden variable Z to obtain new sample
Figure FDA0002948561190000012
S24: iterating the whole network by using a gradient descent method to obtain a time sequence after final restoration
Figure FDA0002948561190000013
The formula for the gradient descent is as follows:
Figure FDA0002948561190000014
wherein z is a hidden variable, u is a mean value of the coding process to the data through the neural network, and sigma is a variance of the mean value of the coding process to the data through the neural network;
and step three, training the travel time data by utilizing a recurrent neural network (LSTM), and predicting the travel time.
2. The method for predicting the travel time of the highway based on the VAE and deep learning combined model as claimed in claim 1, wherein the data preprocessing in the first step comprises the following steps:
s11: removing missing data in the highway toll data;
s12: for the remaining complete highway toll data, subtracting the station-entering time from the station-exiting time to obtain the travel time data among stations;
s13: calculating mean values for travel time data between sites
Figure FDA0002948561190000015
And standard deviation of
Figure FDA0002948561190000016
And are provided with
Figure FDA0002948561190000017
Exception data is culled for the range.
3. The method for predicting the travel time of the highway based on the VAE and the deep learning combined model as claimed in claim 1, wherein the step S3 of training the travel time data by using a recurrent neural network (LSTM) and predicting the travel time comprises the following steps.
S31: dividing a data set into a training set and a testing set according to the proportion of 8: 2;
s32: setting parameters and putting the parameters into an LSTM model for training;
s33: and predicting the travel time according to the trained model.
CN202110200510.3A 2021-02-23 2021-02-23 Expressway travel time prediction method based on VAE and deep learning combined model Pending CN113065684A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110200510.3A CN113065684A (en) 2021-02-23 2021-02-23 Expressway travel time prediction method based on VAE and deep learning combined model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110200510.3A CN113065684A (en) 2021-02-23 2021-02-23 Expressway travel time prediction method based on VAE and deep learning combined model

Publications (1)

Publication Number Publication Date
CN113065684A true CN113065684A (en) 2021-07-02

Family

ID=76558914

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110200510.3A Pending CN113065684A (en) 2021-02-23 2021-02-23 Expressway travel time prediction method based on VAE and deep learning combined model

Country Status (1)

Country Link
CN (1) CN113065684A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114904665A (en) * 2022-04-20 2022-08-16 珠海格力电器股份有限公司 Anti-surge control method and device for magnetic suspension centrifugal machine and storage medium
CN115187266A (en) * 2022-06-29 2022-10-14 山东大学 Credit card fraud detection method and system based on memory variation self-coding model

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106228268A (en) * 2016-07-26 2016-12-14 北京航空航天大学 A kind of Forecasting Methodology of expressway travel time based on charge data
CN106981198A (en) * 2017-05-24 2017-07-25 北京航空航天大学 Deep learning network model and its method for building up for predicting travel time
CN108197736A (en) * 2017-12-29 2018-06-22 北京工业大学 A kind of Air Quality Forecast method based on variation self-encoding encoder and extreme learning machine
US20190088116A1 (en) * 2017-09-16 2019-03-21 Tata Consultancy Services Limited Predicting vehicle travel times by modeling heterogeneous influences between arterial roads
CN111563706A (en) * 2020-03-05 2020-08-21 河海大学 Multivariable logistics freight volume prediction method based on LSTM network

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106228268A (en) * 2016-07-26 2016-12-14 北京航空航天大学 A kind of Forecasting Methodology of expressway travel time based on charge data
CN106981198A (en) * 2017-05-24 2017-07-25 北京航空航天大学 Deep learning network model and its method for building up for predicting travel time
US20190088116A1 (en) * 2017-09-16 2019-03-21 Tata Consultancy Services Limited Predicting vehicle travel times by modeling heterogeneous influences between arterial roads
CN108197736A (en) * 2017-12-29 2018-06-22 北京工业大学 A kind of Air Quality Forecast method based on variation self-encoding encoder and extreme learning machine
CN111563706A (en) * 2020-03-05 2020-08-21 河海大学 Multivariable logistics freight volume prediction method based on LSTM network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
杨英等: "VAE＿LSTM算法在时间序列预测模型中的研究", 《湖南科技大学学报(自然科学版)》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114904665A (en) * 2022-04-20 2022-08-16 珠海格力电器股份有限公司 Anti-surge control method and device for magnetic suspension centrifugal machine and storage medium
CN115187266A (en) * 2022-06-29 2022-10-14 山东大学 Credit card fraud detection method and system based on memory variation self-coding model
CN115187266B (en) * 2022-06-29 2023-08-25 山东大学 Credit card fraud detection method and system based on memory variation self-coding model

Similar Documents

Publication Publication Date Title
CN111709549B (en) SVD-PSO-LSTM-based short-term traffic flow prediction navigation reminding method
CN113065684A (en) Expressway travel time prediction method based on VAE and deep learning combined model
CN109886387B (en) Traffic time sequence prediction method based on gating network and gradient lifting regression
CN111862592B (en) Traffic flow prediction method based on RGCN
CN109637128B (en) Markov-based gray Verhulst short-time traffic flow prediction method and system
CN111301426A (en) Method for predicting energy consumption in future driving process based on GRU network model
CN111027673B (en) Subway passenger flow prediction method
CN111931978A (en) Urban rail transit passenger flow state prediction method based on space-time characteristics
CN111695722A (en) Method for predicting short-term passenger flow in holidays of urban rail transit station
CN111063204B (en) Expressway vehicle speed prediction model training method based on toll station flow
CN112634608B (en) Regional traffic flow prediction method and system
CN109086926B (en) Short-time rail transit passenger flow prediction method based on combined neural network structure
CN109887279B (en) Traffic jam prediction method and system
CN113449905A (en) Traffic jam early warning method based on gated cyclic unit neural network
CN111554118A (en) Dynamic prediction method and system for bus arrival time
CN105303835B (en) A kind of Forecasting Approach for Short-term of road traffic stream mode
CN113406503A (en) Lithium battery SOH online estimation method based on deep neural network
CN115169714A (en) Urban subway station entrance and exit passenger flow prediction method
CN115358491A (en) Subway passenger flow prediction method and system based on CEEMDAN-LSTM
CN107704944A (en) A kind of fluctuation of stock market interval prediction method based on information theory study
CN112634620B (en) Road network dynamic traffic distribution method based on Encoder-Decoder deep neural network
CN115796030A (en) Traffic flow prediction method based on graph convolution
CN112115754A (en) Short-term traffic flow prediction model based on firework differential evolution hybrid algorithm-extreme learning machine
CN113096378B (en) Highway OD prediction method based on depth set empirical mode decomposition
CN114492967A (en) Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model

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: 20210702

RJ01 Rejection of invention patent application after publication