CN108615360B - Traffic demand day-to-day evolution prediction method based on neural network - Google Patents

Traffic demand day-to-day evolution prediction method based on neural network Download PDF

Info

Publication number
CN108615360B
CN108615360B CN201810431068.3A CN201810431068A CN108615360B CN 108615360 B CN108615360 B CN 108615360B CN 201810431068 A CN201810431068 A CN 201810431068A CN 108615360 B CN108615360 B CN 108615360B
Authority
CN
China
Prior art keywords
neural network
network model
data
level fusion
feature level
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
Application number
CN201810431068.3A
Other languages
Chinese (zh)
Other versions
CN108615360A (en
Inventor
刘志远
程启秀
刘洋
魏薇
俞俊
李喆康
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.)
Southeast University
Original Assignee
Southeast 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 Southeast University filed Critical Southeast University
Priority to CN201810431068.3A priority Critical patent/CN108615360B/en
Publication of CN108615360A publication Critical patent/CN108615360A/en
Application granted granted Critical
Publication of CN108615360B publication Critical patent/CN108615360B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
    • G08G1/0145—Measuring and analyzing of parameters relative to traffic conditions for specific applications for active traffic flow control
    • 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

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Traffic Control Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a traffic demand day-to-day evolution prediction method based on a neural network. The invention provides a cyclic neural network model and a feature level fusion neural network model, and has high flexibility and accuracy for short-term dynamic traffic demand prediction.

Description

Traffic demand day-to-day evolution prediction method based on neural network
Technical Field
The invention relates to the technical field of urban traffic, in particular to a traffic demand day-to-day evolution prediction method based on a neural network.
Background
The traffic demand prediction is to build a model according to the traffic conditions and characteristics of the past and the present situation and predict the future traffic flow change. The key to solve the urban traffic problem lies in realizing the balance of traffic supply and demand, and the accurate analysis of traffic demand is the basis for solving the contradiction between supply and demand.
Travel generation prediction is the basis of traffic demand prediction in urban traffic planning, and important support is provided for an effective and accurate dynamic traffic distribution model. However, the existing traffic demand prediction model is mainly based on the traditional manual survey mode and a large amount of relevant data such as economy and population, so that huge manpower and material resources are consumed, the result of the model is lack of accuracy and timeliness, and the dynamic and real-time management requirements on a large-scale traffic network in an intelligent traffic system cannot be met.
In summary, the existing traffic demand prediction method has many defects; the traditional traffic generation prediction research based on the four-stage traffic demand prediction method ignores the space-time internal relevance of a traffic network, and further cannot realize the dynamic real-time analysis of a large-range traffic network in a short time.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a traffic demand day-by-day evolution prediction method based on a neural network, and data mining is performed by using a mobile phone signaling resource, so that the defects of the traditional traffic demand prediction method in the field of traffic planning in the aspects of precision and efficiency are overcome, the road network performance of each day in the traffic network day-by-day dynamic evolution process is fully considered, and the method has strong applicability.
In order to solve the technical problem, the invention provides a traffic demand day-to-day evolution prediction method based on a neural network, which comprises the following steps:
(1) dividing specific partitions of the urban traffic network and numbering;
(2) determining a cell set during neural network model training and a cell set during model verification based on the traffic network partition established in the step (1);
(3) grouping the obtained data sets according to the travel generation cells and the small time segments, and performing data preprocessing;
(4) establishing a cyclic neural network model and a feature level fusion neural network model for predicting traffic demands, wherein the cyclic neural network model and the feature level fusion neural network model are used for function estimation and weight updating of each cell under different time states;
(5) inputting the travel generated quantity of each cell as a training sample, and completing weight determination of the recurrent neural network model and the feature level fusion neural network model; when prediction is carried out, tested community traffic demand data is input into the trained cyclic neural network model and the trained feature level fusion neural network model for calculation, and finally the predicted traffic demand of the region is obtained.
Preferably, in the step (3), the data preprocessing specifically includes the following steps:
(31) filling missing data, replacing the missing data by using the average value of data of two adjacent time periods when single data is missing, and replacing the missing data by using data of the same time period in the next day when a plurality of data are missing;
(32) extracting features by using a sliding time window with the size of r and the step value of 1;
(33) the data of the requirements is standardized and the data of the requirements is,
x＝(x-mean)/std
where std is the standard deviation of the data.
Preferably, in the step (4), the establishing of the recurrent neural network model and the feature level fusion neural network model for predicting the traffic demand specifically includes the following steps:
(41) constructing a cyclic neural network model and a feature level fusion neural network model, and determining the forward propagation process of an RNN (neural network) model, wherein the RNN model consists of an input unit, an output unit and a cyclic hidden layer, and xtIs an input value at time t, stIs the hidden layer state at time t
st＝f(Uxt+Wst-1)
Wherein, the function f is a nonlinear function such as tanh or ReLU, and U and W are corresponding weights respectively;
(42) determining the input value in the neural network model and the calculation mode of a hidden layer, and for a hidden unit:
Figure BDA0001653437690000021
Figure BDA0001653437690000022
Figure BDA0001653437690000023
the input sequence x with the length of T is provided with I input units, H hidden units and K output units;
Figure BDA0001653437690000024
representing the value of the input quantity i at time t,
Figure BDA0001653437690000025
and
Figure BDA0001653437690000026
the input quantity and the excitation value omega of the whole traffic network to the hidden unit h at the time tihAnd ωh′hTo adjust the weight, θhIs composed of
Figure BDA0001653437690000031
To
Figure BDA0001653437690000032
The excitation function of (a);
(43) for the feature level fusion neural network model, taking OD matrix data, the average temperature and weather condition of the region and 5 corresponding variables of date and time as input quantities to train;
(44) the weights in the model are derived using the BPTT algorithm.
Preferably, in step (44), the weight in the model obtained by using the BPTT algorithm is specifically:
(a) in the interval
Figure BDA0001653437690000033
Inner random initialization weights (U, V, W), where n is the previous level incomings connections;
(b) let epoch be 1;
(c) running a forward propagation process in the neural network model;
(d) calculating a loss function L and a state gradient
Figure BDA0001653437690000034
And
Figure BDA0001653437690000035
Figure BDA0001653437690000036
Figure BDA0001653437690000037
Figure BDA0001653437690000038
(e) training an RNN model by using a random gradient descent and backward propagation process, and adding 1 to an epoch count value;
(f) if epoch is equal to M, the loop is stopped, at which point (U, V, W) is the weight value sought, otherwise (c) is returned.
The invention has the beneficial effects that: the method can dynamically predict the travel demand of a large-scale traffic network in real time, provides a recurrent neural network model and a feature level fusion neural network model, and has high flexibility and accuracy for predicting the short-term dynamic traffic demand.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Fig. 2 is a schematic diagram of traffic cell division of the present invention.
Fig. 3(a) is a schematic diagram of the prediction result of the deep neural network model according to the present invention.
FIG. 3(b) is a schematic diagram of the prediction result of the recurrent neural network model of the present invention.
Fig. 3(c) is a schematic diagram of the prediction result of the feature level fusion neural network model of the present invention.
Detailed Description
As shown in fig. 1, a traffic demand day-to-day evolution prediction method based on a neural network constructs a generation attraction OD matrix of traffic travel in a certain time period according to mobile phone signaling data between cells in a demand prediction region, and establishes a recurrent neural network model and a feature level fusion neural network model by using a deep learning method in machine learning, thereby predicting a dynamic traffic demand of the region day-to-day evolution, the method comprising the following steps:
the method comprises the following steps: and dividing specific partitions of the urban traffic network and numbering.
The analysis area is divided into 1261 inner areas (1-3999) and 6 outer areas (4000-;
step two: determining a cell set during neural network model training based on the traffic network partition established in the first step, wherein the cell set is OD (optical density) quantities generated by an external cell with the number 4000-4004 and an internal cell 1261; the cell set during model verification is the OD quantity generated by the external cell with the number 4005 and the internal cell 1261;
step three: in order to solve the problems of data loss, too short time period and the like of the mobile phone signaling data, the obtained data sets are grouped according to the travel generation cell and the small time period, and data preprocessing is carried out, and the method specifically comprises the following steps:
3.1 filling missing data, when single data is missing, replacing by using the data average value of two time slices adjacent to the data, and when a plurality of data are missing, replacing by using the data of the same time slice on the next day;
3.2, extracting features by using a sliding time window with the size of r and the step value of 1;
3.3 the demand data is standardized,
x＝(x-mean)/std
wherein std is the standard deviation of the data;
step four: establishing a cyclic neural network model and a feature level fusion neural network model for predicting traffic demands, and using the cyclic neural network model and the feature level fusion neural network model for function estimation and weight updating of each cell under different time states, wherein the specific steps are as follows:
4.1 constructing a recurrent neural network model and a feature level fusion neural network model, and determining the forward propagation process of the RNN model, wherein the RNN model consists of an input unit, an output unit and a recurrent hidden layer, and xtAnd otRespectively input and output values at time t, stIs the hidden layer state at time t
st＝f(Uxt+Wst-1)
Wherein the function f is a non-linear function such as tanh or ReLU;
4.2 determining the input value in the neural network model and the calculation mode of the hidden layer, and for the hidden unit:
Figure BDA0001653437690000051
Figure BDA0001653437690000052
Figure BDA0001653437690000053
the input sequence x with length T has I input units, H hidden units and K output units.
Figure BDA0001653437690000054
Representing the value of the input quantity i at time t,
Figure BDA0001653437690000055
and
Figure BDA0001653437690000056
the input quantity and the excitation value of the whole traffic network to the hidden unit h at the time t;
4.3 for the feature level fusion neural network model, taking OD matrix data and 5 variables such as the average temperature, the weather condition, the corresponding date and the corresponding time of the region as input quantities to train;
4.4 use BackPropassivation through Time (BPTT) algorithm to get the weights in the model.
Further, in the inventive method, the BPTT algorithm flow in step 4.4 is:
a) in the interval
Figure BDA0001653437690000057
Internal random initialization weights (U, V, W), where n is the previous level of incoming connections
b) Let epo equal to 1
c) Running forward propagation processes in neural network models
d) Calculating loss value
Figure BDA0001653437690000058
Figure BDA0001653437690000059
Figure BDA00016534376900000510
e) Training an RNN model by using a random gradient descent and backward propagation process, and adding 1 to an epoch count value;
f) if the epoch is equal to M, stopping the circulation, wherein (U, V, W) is the weight value, otherwise, returning to c);
step five: inputting the travel generated quantity of each cell as a training sample, and completing weight determination of the recurrent neural network model and the feature level fusion neural network model; secondly, when prediction is carried out, the tested community traffic demand data is input into the trained cyclic neural network model and the trained feature level fusion neural network model for calculation, and finally the predicted traffic demand of the region is obtained.
Fig. 3(a), 3(b) and 3(c) show the output results of three models for the day-by-day dynamic traffic demand prediction problem. The experimental results prove that the cyclic neural network (RNN) is superior to the Deep Neural Network (DNN) in daily traffic demand prediction, and the characteristic-level fusion neural network model performs best in the three models, and the average relative error is lower than 18%.

Claims (2)

1. A traffic demand day-to-day evolution prediction method based on a neural network is characterized by comprising the following steps:
(1) dividing the specific partition of the urban traffic network into an internal area and an external area and numbering;
(2) determining a cell set during training and inspection of a neural network model based on the traffic network partition established in the step (1), wherein data in the set is OD (origin-destination) quantity generated between an external area and an internal area, and the OD quantity is mobile phone signaling data;
(3) grouping the obtained data sets according to the travel generation cells and the small time segments, and performing data preprocessing;
(4) establishing a feature level fusion neural network model for predicting traffic demands, wherein the feature level fusion neural network model is used for function estimation and weight updating of each cell under different time states; the method for establishing the feature level fusion neural network model for predicting the traffic demand specifically comprises the following steps:
(41) constructing a cyclic neural network model, determining the forward propagation process of the cyclic neural network model, wherein the cyclic neural network model consists of an input unit, an output unit and a cyclic hidden layer, and xtIs an input value at time t, stIs the hidden layer state at time t
st＝f(Uxt+Wst-1)
Wherein, the function f is a nonlinear function, and U and W are corresponding weights respectively;
(42) determining the input value in the recurrent neural network model and the calculation mode of a hidden layer, and for a hidden unit:
Figure FDA0003348586900000011
Figure FDA0003348586900000012
Figure FDA0003348586900000013
the input sequence x with the length of T is provided with I input units, H hidden units and K output units;
Figure FDA0003348586900000014
representing the value of the input quantity i at time t,
Figure FDA0003348586900000015
and
Figure FDA0003348586900000016
the input quantity and the excitation value omega of the whole traffic network to the hidden unit h at the time tihAnd ωh′hTo adjust the weight, θhIs composed of
Figure FDA0003348586900000017
To
Figure FDA0003348586900000018
The excitation function of (a);
(43) for the cyclic neural network model, taking OD matrix data, the average temperature and weather condition of the region and 5 corresponding variables of date and time as input quantities to train to obtain a characteristic level fusion neural network model;
(44) using a BPTT algorithm to obtain weights in the feature level fusion neural network model; the weight of the feature level fusion neural network model obtained by using the BPTT algorithm is specifically as follows:
(a) in the interval
Figure FDA0003348586900000021
Inner random initialization weights (U, V, W), where n is the previous level incomings connections;
(b) let epoch be 1;
(c) running a forward propagation process in the feature level fusion neural network model;
(d) computingLoss function L and state gradient
Figure FDA0003348586900000022
And
Figure FDA0003348586900000023
Figure FDA0003348586900000024
Figure FDA0003348586900000025
Figure FDA0003348586900000026
(e) training a feature level fusion neural network model by using a random gradient descent and backward propagation process, and adding 1 to an epoch count value;
(f) if the epoch is equal to M, stopping the circulation, wherein (U, V, W) is the weight value, otherwise, returning to the step (c);
(5) inputting the travel generated quantity of each cell as a training sample to complete the weight determination of the feature level fusion neural network model; and when testing, inputting the tested community traffic demand data into the trained feature level fusion neural network model for calculation to obtain the final feature level fusion neural network model.
2. The neural network-based traffic demand day-by-day evolution prediction method according to claim 1, wherein in the step (3), the data preprocessing specifically comprises the following steps:
(31) filling missing data, replacing the missing data by using the average value of data of two adjacent time periods when single data is missing, and replacing the missing data by using data of the same time period in the next day when a plurality of data are missing;
(32) extracting features by using a sliding time window with the size of r and the step value of 1;
(33) the data of the requirements is standardized and the data of the requirements is,
x＝(x-mean)/std
where std is the standard deviation of the data.
CN201810431068.3A 2018-05-08 2018-05-08 Traffic demand day-to-day evolution prediction method based on neural network Active CN108615360B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810431068.3A CN108615360B (en) 2018-05-08 2018-05-08 Traffic demand day-to-day evolution prediction method based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810431068.3A CN108615360B (en) 2018-05-08 2018-05-08 Traffic demand day-to-day evolution prediction method based on neural network

Publications (2)

Publication Number Publication Date
CN108615360A CN108615360A (en) 2018-10-02
CN108615360B true CN108615360B (en) 2022-02-11

Family

ID=63662187

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810431068.3A Active CN108615360B (en) 2018-05-08 2018-05-08 Traffic demand day-to-day evolution prediction method based on neural network

Country Status (1)

Country Link
CN (1) CN108615360B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110322064B (en) * 2019-06-27 2022-04-15 清华大学 Urban trip demand prediction method
CN111833596B (en) * 2019-11-19 2022-04-29 东南大学 Day-by-day road section flow prediction method considering decision inertia of travelers
CN110942626B (en) * 2019-11-21 2021-09-21 华南理工大学 Road network mixed flow rate daily variation prediction method considering unmanned vehicles
CN114664090B (en) * 2022-04-14 2023-07-04 山东大学 Traffic data filling method and system based on cyclic neural network

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103234540A (en) * 2013-03-20 2013-08-07 河北工业大学 No-blind-zone positioning method for fire fighting and rescue dangerous operation workers
CN104615983A (en) * 2015-01-28 2015-05-13 中国科学院自动化研究所 Behavior identification method based on recurrent neural network and human skeleton movement sequences
CN104766146A (en) * 2015-04-24 2015-07-08 陆化普 Traffic demand forecasting method and system
CN105389980A (en) * 2015-11-09 2016-03-09 上海交通大学 Short-time traffic flow prediction method based on long-time and short-time memory recurrent neural network
CN105809962A (en) * 2016-06-13 2016-07-27 中南大学 Traffic trip mode splitting method based on mobile phone data
CN106504535A (en) * 2016-11-30 2017-03-15 东南大学 A kind of combination Gravity Models and the trip distribution modeling method of Fratar models
CN107273782A (en) * 2016-04-08 2017-10-20 微软技术许可有限责任公司 Detected using the online actions of recurrent neural network
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
CN107730087A (en) * 2017-09-20 2018-02-23 平安科技（深圳）有限公司 Forecast model training method, data monitoring method, device, equipment and medium
CN107730003A (en) * 2017-10-23 2018-02-23 华中科技大学 One kind supports more high-precision NILM implementation methods of appliance type
CN107798877A (en) * 2017-11-22 2018-03-13 北京云星宇交通科技股份有限公司 A kind of method and system being predicted based on expressway tol lcollection data to the volume of traffic

Family Cites Families (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4950596B2 (en) * 2006-08-18 2012-06-13 クラリオン株式会社 Predicted traffic information generation method, predicted traffic information generation device, and traffic information display terminal
US8532927B2 (en) * 2008-11-07 2013-09-10 Intellectual Ventures Fund 83 Llc Generating photogenic routes from starting to destination locations
CN101436345B (en) * 2008-12-19 2010-08-18 天津市市政工程设计研究院 System for forecasting harbor district road traffic requirement based on TransCAD macroscopic artificial platform
CN102324128B (en) * 2011-05-24 2013-07-31 北京交通大学 Method for predicting OD (Origin-Destination) passenger flow among bus stations on basis of IC (Integrated Circuit)-card record and device
KR101473103B1 (en) * 2013-05-29 2014-12-15 서울시립대학교 산학협력단 traffic demand forecast system based on agents
US9047767B2 (en) * 2013-09-09 2015-06-02 International Business Machines Corporation Traffic impact prediction for multiple event planning
JP6314656B2 (en) * 2014-05-23 2018-04-25 日産自動車株式会社 Road traffic demand forecasting apparatus and road traffic demand forecasting method
CN104183119B (en) * 2014-08-19 2016-08-24 中山大学 Based on the anti-arithmetic for real-time traffic flow distribution forecasting method pushed away of section OD
CN104484993B (en) * 2014-11-27 2017-01-18 北京交通大学 Processing method of cell phone signaling information for dividing traffic zones
US20160282132A1 (en) * 2015-03-27 2016-09-29 International Business Machines Corporation Predictive navigation
CN104899443B (en) * 2015-06-05 2018-03-06 陆化普 For assessing the method and system of current trip requirements and the following trip requirements of prediction
CN105142106B (en) * 2015-07-29 2019-03-26 西南交通大学 The identification of traveler duty residence and Trip chain depicting method based on mobile phone signaling data
CN106846805B (en) * 2017-03-06 2019-11-08 南京多伦科技股份有限公司 A kind of dynamic road grid traffic needing forecasting method and its system
CN107170233B (en) * 2017-04-20 2020-08-18 同济大学 Typical daily traffic demand OD matrix acquisition method based on matrix decomposition
CN107180530B (en) * 2017-05-22 2019-09-06 北京航空航天大学 A kind of road network trend prediction method based on depth space-time convolution loop network
CN107134142B (en) * 2017-07-10 2018-06-12 中南大学 A kind of urban road method for predicting based on multisource data fusion
CN107230351B (en) * 2017-07-18 2019-08-09 福州大学 A kind of Short-time Traffic Flow Forecasting Methods based on deep learning

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103234540A (en) * 2013-03-20 2013-08-07 河北工业大学 No-blind-zone positioning method for fire fighting and rescue dangerous operation workers
CN104615983A (en) * 2015-01-28 2015-05-13 中国科学院自动化研究所 Behavior identification method based on recurrent neural network and human skeleton movement sequences
CN104766146A (en) * 2015-04-24 2015-07-08 陆化普 Traffic demand forecasting method and system
CN105389980A (en) * 2015-11-09 2016-03-09 上海交通大学 Short-time traffic flow prediction method based on long-time and short-time memory recurrent neural network
CN107273782A (en) * 2016-04-08 2017-10-20 微软技术许可有限责任公司 Detected using the online actions of recurrent neural network
CN105809962A (en) * 2016-06-13 2016-07-27 中南大学 Traffic trip mode splitting method based on mobile phone data
CN106504535A (en) * 2016-11-30 2017-03-15 东南大学 A kind of combination Gravity Models and the trip distribution modeling method of Fratar models
CN107730087A (en) * 2017-09-20 2018-02-23 平安科技（深圳）有限公司 Forecast model training method, data monitoring method, device, equipment and medium
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
CN107730003A (en) * 2017-10-23 2018-02-23 华中科技大学 One kind supports more high-precision NILM implementation methods of appliance type
CN107798877A (en) * 2017-11-22 2018-03-13 北京云星宇交通科技股份有限公司 A kind of method and system being predicted based on expressway tol lcollection data to the volume of traffic

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
"基于RNN的空气污染时空预报模型研究";范竣翔等;《测绘科学》;20170731;第42卷(第7期);第76-83页 *
"基于RNN的网络安全态势预测方法";胡昕;《现代计算机(专业版)》;20170225;全文 *
"基于深度学习的城市地面公交客流集散点刷卡客流预测——以常州市为例";徐特等;《交通工程》;20180430;第18卷(第2期);第13-18页 *

Also Published As

Publication number Publication date
CN108615360A (en) 2018-10-02

Similar Documents

Publication Publication Date Title
CN108615360B (en) Traffic demand day-to-day evolution prediction method based on neural network
CN109285346B (en) Urban road network traffic state prediction method based on key road sections
CN102074124B (en) Dynamic bus arrival time prediction method based on support vector machine (SVM) and H-infinity filtering
CN108009674A (en) Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks
CN110517482B (en) Short-term traffic flow prediction method based on 3D convolutional neural network
CN109194423A (en) Single-frequency point spectrum prediction method based on optimized long-short term memory model
CN106372402A (en) Parallelization method of convolutional neural networks in fuzzy region under big-data environment
CN112651665A (en) Surface water quality index prediction method and device based on graph neural network
CN106570597A (en) Content popularity prediction method based on depth learning under SDN architecture
CN113378939B (en) Structure digital twin modeling and parameter identification method based on physical driving neural network
CN104217258B (en) A kind of electric load sigma-t Forecasting Methodology
CN110164129B (en) Single-intersection multi-lane traffic flow prediction method based on GERNN
CN112307536B (en) Dam seepage parameter inversion method
CN107358325A (en) A kind of Location Selection of Logistics Distribution Center method, apparatus and computer-readable recording medium
CN112419711B (en) Closed parking lot parking demand prediction method based on improved GMDH algorithm
CN109034500A (en) A kind of mid-term electric load forecasting method of multiple timings collaboration
CN114780739A (en) Time sequence knowledge graph completion method and system based on time graph convolution network
CN104656620A (en) Comprehensive evaluation system for remanufacturing of heavy-duty machine tool
Hu et al. Dynamically Optimized Unstructured Grid (DOUG) for Analog Ensemble of numerical weather predictions using evolutionary algorithms
CN109787821A (en) A kind of Large-scale Mobile customer traffic consumption intelligent Forecasting
CN117610208A (en) River water quality and quantity prediction method based on urban water system model interaction
CN111163430B (en) Water quantity prediction method based on mobile phone base station user positioning data
CN108537581A (en) Based on the GMDH energy consumption Time Series Forecasting Methods selectively combined and device
CN106844626A (en) Using microblogging keyword and the method and system of positional information simulated air quality
CN106980675B (en) A kind of efficient bridge structure health early warning system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant