CN107230351B - A kind of Short-time Traffic Flow Forecasting Methods based on deep learning - Google Patents
A kind of Short-time Traffic Flow Forecasting Methods based on deep learning Download PDFInfo
- Publication number
- CN107230351B CN107230351B CN201710585474.0A CN201710585474A CN107230351B CN 107230351 B CN107230351 B CN 107230351B CN 201710585474 A CN201710585474 A CN 201710585474A CN 107230351 B CN107230351 B CN 107230351B
- Authority
- CN
- China
- Prior art keywords
- traffic flow
- time
- short
- feature
- data
- 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
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
-
- 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
- G06N3/045—Combinations of 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Abstract
The present invention discloses a kind of Short-time Traffic Flow Forecasting Methods based on deep learning method.The variation of the future position point of proximity magnitude of traffic flow, the influence of the time response of future position and its cyclophysis to future position traffic flow are considered simultaneously, obtain the predicted value of short-term traffic flow.The present invention combines convolutional neural networks (Convolutional Neural Network) and shot and long term memory (LSTM) recurrent neural network, constructs a kind of Conv-LSTM deep neural network model；And analyzed using the two-way LSTM model traffic flow historical data previous to the point, its periodic feature is extracted, is finally merged the traffic flow trend and periodic feature that are analyzed, to realize the prediction of traffic flow；The present invention overcomes the deficiencies that existing method cannot make full use of space-time characteristic, the periodic feature of traffic flow data have been merged while sufficiently extracting traffic flow time and space characteristics, to effectively improve the accuracy of Forecasting Short-term Traffic.
Description
Technical field
The present invention relates to intelligent transportation field and deep learning field, it is especially a kind of based on deep learning method in short-term
Traffic flow forecasting method.
Background technique
With the continuous improvement of economic continuous development and the level of urbanization, demand of the people to traffic was originally higher, drove
The frequency of vehicle trip is also higher and higher, and following problem is exactly that traffic congestion is increasingly serious, how to grasp traffic information, joins
Examining traffic information and carrying out planning to travel time and traffic path is that a needs currently solve the problems, such as, is provided accurately for user
Real-time traffic flow variation prediction can save the travel time for user, reduce unnecessary waste, while accurate traffic flow
Information has very big commercial value to being also very helpful in maintenance traffic support and traffic administration.
The domestic and international main having time serial method of traffic flow forecasting method proposed already, Kalman filtering, chaology, mind
Through network and support vector machines (SVM) etc..The data information of traffic flow has very strong temporal correlation and periodicity, while
Will receive the influence for closing on lane flow amount, forecasting traffic flow by sensor collection to data carry out analysis and predict to connect
The get off magnitude of traffic flow trend in the short time, sensor mainly includes mobile phone, and Intelligent bracelet, tablet computer etc. has sensor
Movable equipment also includes road fixed equipment, and road fixed equipment constantly improve, and the movable equipments such as mobile phone quantity is not
Disconnected to be promoted, collected data are more and more, and with the explosion of big data, the data that can be used to analyze are also more and more,
The information that we want how is obtained from data to become more and more important, and is existed in forecasting traffic flow model before very much
Disadvantage, including no addition spatial character do not add time response, and feature extraction is insufficient or time and space information is melted
It closes bad, causes the big problem of forecasting traffic flow error.The invention proposes a kind of fusion space-time characterisation and combine traffic flow
The features such as data are periodical predicts Short-Term Traffic Flow.
Summary of the invention
The purpose of the present invention is to provide a kind of Short-time Traffic Flow Forecasting Methods based on deep learning, this method utilizes city
City's road traffic flow space time correlation information is predicted that the space-time that can overcome existing method that cannot make full use of traffic flow data is special
Seek peace periodic feature the shortcomings that, while further traffic flow data different characteristic being merged, to improve traffic in short-term
Flow the accuracy of prediction.
To achieve the above object, the technical scheme is that a kind of short-time traffic flow forecast side based on deep learning
Method includes the following steps,
Step S1: the traffic flow modes between synchronization future position and point of proximity are extracted, the sky of short-term traffic flow is calculated
Between correlative character；
Step S2: processing time series in different moments each point traffic flow modes, calculate traffic flow variation tendency and
Space-time characteristic；
Step S3: synchronization and the traffic flow data before one week when synchronization on the day before future position are extracted, extracts and hands over
Through-flow periodic feature；
Step S4: space-time characteristic and periodic feature are merged；
Step S5: being compared using traffic flow actual value with frame predicted value, is calculated penalty values, is continued to optimize frame.
In an embodiment of the present invention, in the step S1, the spatial coherence feature of short-term traffic flow passes through as follows
Mode obtains:
Step S11: collecting the traffic flow data collection of target detection point and its point of proximity, leads to point in the synchronization region
Traffic flow data be mapped to one-dimensional vector, the traffic flow data of future position is placed in vector by the point on the basis of future position
The heart, it is northern in preceding, Nan Hou before or after setting data point is future position in vector using north and south as measurement standard, and
Vector is filled according to the distance between datum mark distance；
Step S12: carrying out process of convolution to vector, and using convolution kernel having a size of 3, sliding step 1 is extracted mutually in the same time
Region in traffic flow correlation, generate traffic flow data convolution feature vector, convolution can be regarded as local weighted sum,
Formula is
G (i)=f (Aw+B)
Wherein A indicates the input of convolution kernel, and W indicates convolution kernel to the weight of input processing, and B indicates bias term, and i indicates volume
The step number of product sliding, that is, generate i-th of element in convolution feature vector；
Step S13: average pondization processing is carried out to convolution feature vector caused by previous step.
In an embodiment of the present invention, in the step S2, time series is handled using LSTM recurrent neural network
Data analyze the spatial coherence changing features of different time traffic flow, and temporal characteristics and space characteristics are merged, and obtain traffic
Variation tendency and space-time characteristic are flowed, specifically i.e.:
The spatial coherence feature of different moments extracted traffic flow is pooled into a time series, by time series
It is input among two layers of LSTM, generates the space-time characteristic vector of traffic flow.
In an embodiment of the present invention, in the step S3, before the periodic feature of traffic flow is by collecting future position
The traffic flow data of one day synchronization and the traffic flow data of the last week synchronization, data are separately input to two-way
Among LSTM, day periodic characteristic and cycle feature are extracted.
In an embodiment of the present invention, it in the step S4, will be collected into using the fully-connected network in neural network
Traffic flow space-time characteristic and periodic feature merged.
In an embodiment of the present invention, in the step S5, predicted value and traffic flow actual value that frame is exported
It is compared, calculates penalty values, constantly frame is optimized, uses mean square error for loss function, Computational frame output
Then feature between predicted value and actual traffic flow data is carried out constantly excellent using parameter of the back-propagation algorithm to frame
Change, continuous calculating parameter gradient in back-propagation algorithm, and use the continuous autoadapted learning rate of RMSprop, RMSprop being capable of root
Carry out renewal learning rate according to the case where change of gradient before, RMSprop algorithm uses variable MeanSquare (w, t) Lai Baocun the
The average value of the gradient square of each weight for the previous period when t renewal learning rate, according to this variable come adaptive learning
Rate continues to optimize parameter, and structure is made to be optimal solution；Wherein,
Mean square error function:
RMSprop formula:
For variable W t moment gradient value.
Compared to the prior art, the invention has the following advantages: it is proposed by the present invention a kind of based on deep learning
Short-time Traffic Flow Forecasting Methods are predicted using urban road traffic flow space time correlation information and periodical information, can be overcome
Existing method cannot make full use of the deficiency of space-time characteristic and temporal correlation combines insufficient disadvantage, while utilize depth
The step of practising end-to-end feature, having subtracted artificial extraction data characteristics, few reduces manpower consumption, to improve in short-term
The accuracy of forecasting traffic flow increases the portability of system.
Detailed description of the invention
Fig. 1 is that space-time characteristic extracts schematic diagram in the embodiment of the present invention.
Fig. 2 is that periodic feature extracts schematic diagram in the embodiment of the present invention.
Fig. 3 is to merge space-time characterisation and periodic feature in the embodiment of the present invention and obtain prediction result.
Fig. 4 is the target detection point traffic in short-term generated in the embodiment of the present invention using method prediction proposed by the invention
Flow prediction result and actual comparison.
Specific embodiment
With reference to the accompanying drawing, technical solution of the present invention is specifically described.
A kind of Short-time Traffic Flow Forecasting Methods based on deep learning of the invention, are specifically realized in accordance with the following steps,
Step S1: considering the spatial coherence of traffic flow data, obtains the space characteristics of forecasting traffic flow point；
In this example, firstly, collecting the traffic flow data collection of target detection point and its point of proximity, leading to should by synchronization
The traffic flow data put in region is mapped to one-dimensional vector, and the traffic flow data of future position is placed on by the point on the basis of future position
The center of vector, using north and south as measurement standard, before or after setting data point is future position in vector, north is in preceding, south
Vector is filled rear, and according to the distance between datum mark distance；
Further, to vector carry out process of convolution, use convolution kernel having a size of 3 sliding steps be 1, extract phase in the same time
Region in traffic flow correlation, generate traffic flow data convolution feature vector；
Convolution can be regarded as local weighted sum, and formula is
G (i)=f (Aw+B)
Wherein A indicates the input of convolution kernel, and W indicates convolution kernel to the weight of input processing, and B indicates bias term, and i is indicated
The step number of convolution sliding, that is, generate i-th of element in convolution feature vector.
Further, average pondization processing is carried out to convolution feature vector caused by previous step, be filled into it is some need not
The information wanted obtains more abstract spatial character, increases the robustness of frame identification.
Step S2: handling time series data with LSTM recurrent neural network, and analysis different time traffic fluid space is special
Property variation, temporal characteristics and space characteristics are merged, traffic flow data variation tendency is obtained.
In this example, there are many points in survey region, it would be desirable to the space phase between each point in analyzed area
Guan Xing, and generalized time considers, the traffic flow variation tendency between each point and the trend that influences each other, and the flow tendency of vehicle will
Different moments, extracted traffic data spatial character pooled a time series, by time series be input to two layers of LSTM it
In, the extracted space characteristics of convolution are merged with temporal characteristics, generate the space-time characteristic vector of traffic flow data；
Step S3: consider the periodic feature of traffic flow data
In this example, the friendship of the traffic flow data and the last week synchronization of future position the previous day synchronization is collected
Data are separately input among two-way LSTM by through-flow data, extract day periodic characteristic and cycle feature.
Step S4: comprehensively consider the space-time characteristic and periodic feature of traffic flow data
In this example, it needs to merge extracted space-time characteristic and periodic feature, further use is melted
It closes feature to be returned, the fully-connected network in specifically used neural network is by the traffic flow data space-time characterisation being collected into and week
Phase property feature is merged.
Step S5: it is carried out between the traffic flow forecasting result and actual traffic flow value of proposed algorithm according to the present invention
Compare, calculates penalty values, continue to optimize frame.
Use mean square error for loss function in this example, Computational frame output predicted value and actual traffic flow data it
Between feature, then continued to optimize using parameter of the back-propagation algorithm to frame, constantly calculated in back-propagation algorithm
Parameter gradients, and use the continuous autoadapted learning rate of RMSprop, RMSprop can according to the case where change of gradient before come
Renewal learning rate, each weight when RMSprop algorithm is using the t times renewal learning rate of variable MeanSquare (w, t) Lai Baocun
The average value of gradient square for the previous period continues to optimize parameter according to this variable come autoadapted learning rate, reaches structure
To optimal solution.
Mean square error function (MSE):
RMSprop formula:
For variable W t moment gradient value
Compared to the prior art, the invention has the following advantages: it is proposed by the present invention a kind of based on deep learning
Short-time Traffic Flow Forecasting Methods are predicted using urban road traffic flow space time correlation information and periodical information, can be overcome
Existing method cannot make full use of the deficiency of space-time characteristic and temporal correlation combines insufficient disadvantage, while utilize depth
The step of practising end-to-end feature, having subtracted artificial extraction data characteristics, few reduces manpower consumption, to improve in short-term
The accuracy of forecasting traffic flow increases the portability of system.
In order to allow those skilled in the art to further appreciate that a kind of friendship in short-term based on deep learning proposed by the invention
Through-flow prediction technique, elaborates combined with specific embodiments below.The present embodiment carries out based on the technical solution of the present invention
Implement, the detailed implementation method and specific operation process are given.
As shown in Figure 1, extracting schematic diagram for target detection point traffic flow space-time characteristic.
The present embodiment comprises the following specific steps that:
Step 1: the traffic flow modes between synchronization future position and future position point of proximity are collected, and use convolutional layer
The space characteristics at the moment are extracted with pond layer.
Step 2: the space characteristics of different moments are pooled into a time series, time series is input to bilayer
The space-time characteristic of traffic flow data is extracted among LSTM.
As shown in Fig. 2, being that periodic feature extracts schematic diagram.
Future position the previous day traffic flow data of synchronization and the traffic flow data of the last week synchronization are collected, it will
Collected data are input among two-way LSTM and extract spatial character.
As shown in figure 3, for space-time characterisation and periodic feature are merged and obtain prediction result.
The present embodiment comprises the following specific steps that:
Step 1: extracted space-time characterisation and periodic feature are merged.
Step 2: the composite character merged is input among fully-connected network, is carried out last recurrence, is obtained traffic
Flow prediction result.
As shown in figure 4, for the target detection point short-time traffic flow forecast knot generated using method proposed by the invention prediction
Fruit and actual comparison
The present embodiment comprises the following specific steps that:
Step 1: between the predicted value exported using mean square error loss function Computational frame and actual traffic flow data
Gap.
Step 2: backpropagation optimization is carried out to model framework since loss function, and constantly adaptive using RMSprop
Answer learning rate, RMSprop can be according to, come renewal learning rate, RMSprop algorithm uses variable the case where change of gradient before
The average value of the gradient square of each weight for the previous period, root when the t times renewal learning rate of MeanSquare (w, t) Lai Baocun
Carry out autoadapted learning rate according to this variable, continue to optimize parameter, structure is made to be optimal solution.
Mean square error function (MSE):
RMSprop formula:
For variable W t moment gradient value
Step 3: the feasibility and universality of proposed method to illustrate the invention, frame predicted value is further and existing
There are method such as support vector machines (SVM), backpropagation neural network (BPNN), certainly coding neural network (SAE), LSTM recurrence mind
Prediction result through network is compared, and obtains estimated performance evaluation index as shown in Table 1.
The existing algorithm forecasting traffic flow Comparative result of table 1
Wherein, error function is respectively absolute average error (MAE), root-mean-square error (RMSE), average absolute percentage mistake
Poor (MAPE), specific formula is as follows:
A kind of above-mentioned analytic explanation, Short-time Traffic Flow Forecasting Methods based on deep learning proposed by the invention, can obtain
Error precision more lower than existing method is obtained, improves the prediction accuracy of short-term traffic flow well, there is certain reference
Value and real economy benefit.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made
When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.
Claims (6)
1. a kind of Short-time Traffic Flow Forecasting Methods based on deep learning, it is characterised in that: include the following steps,
Step S1: the traffic flow modes between synchronization future position and point of proximity are extracted, the space phase of short-term traffic flow is calculated
Closing property feature；
Step S2: the traffic flow modes of each point of different moments in processing time series calculate traffic flow variation tendency and space-time
Feature；The time series is by converging the spatial coherence feature of different time traffic flow；
Step S3: traffic flow data when future position the previous day synchronization and the last week synchronization is extracted, traffic flow is extracted
Periodic feature；
Step S4: space-time characteristic and periodic feature are merged；
Step S5: being compared using traffic flow actual value with frame predicted value, is calculated penalty values, is continued to optimize frame.
2. a kind of Short-time Traffic Flow Forecasting Methods based on deep learning according to claim 1, it is characterised in that: in institute
It states in step S1, the spatial coherence feature of short-term traffic flow obtains in the following way:
Step S11: the traffic flow data collection of target detection point and its point of proximity, the traffic that will be put in the synchronization region are collected
Flow data is mapped to one-dimensional vector, and the traffic flow data of future position is placed on the center of vector by the point on the basis of future position, on the south
North is measurement standard, and setting data point is before or after be future position in vector, north in preceding, Nan Hou, and according to
The distance between datum mark distance fills vector；
Step S12: carrying out process of convolution to vector, and using convolution kernel having a size of 3, sliding step 1 extracts area mutually in the same time
Traffic flow correlation in domain, generates traffic flow data convolution feature vector, and convolution can be regarded as local weighted sum, formula
For g (i)=f (Aw+B)
Wherein A indicates the input of convolution kernel, and W indicates convolution kernel to the weight of input processing, and B indicates bias term, and i indicates that convolution is sliding
Dynamic step number, that is, generate i-th of element in convolution feature vector；
Step S13: average pondization processing is carried out to convolution feature vector caused by previous step.
3. a kind of Short-time Traffic Flow Forecasting Methods based on deep learning according to claim 1, it is characterised in that: in institute
It states in step S2, time series data is handled using shot and long term memory network (LSTM), analyze the sky of different time traffic flow
Between correlative character change, temporal characteristics and space characteristics are merged, obtain traffic flow variation tendency and space-time characteristic, specifically
That is: the spatial correlation characteristic feature of traffic flow.
4. a kind of Short-time Traffic Flow Forecasting Methods based on deep learning according to claim 2, it is characterised in that: in institute
State in step S3, the periodic feature of traffic flow by collecting future position on the day before the traffic flow data of synchronization and previous
Data are separately input among two-way LSTM by the traffic flow data of all synchronizations, extract day periodic characteristic and cycle is special
Sign.
5. a kind of Short-time Traffic Flow Forecasting Methods based on deep learning according to claim 4, it is characterised in that: in institute
State in step S4, using the fully-connected network in neural network by the space-time characteristic for the traffic flow being collected into and periodic feature into
Row fusion.
6. a kind of Short-time Traffic Flow Forecasting Methods based on deep learning according to claim 1, it is characterised in that: in institute
State in step S5, the predicted value that frame is exported be compared with traffic flow actual value, calculate penalty values, constantly to frame into
Row optimization, uses mean square error for loss function, the feature between the predicted value and actual traffic flow data of Computational frame output,
Then it is continued to optimize using parameter of the back-propagation algorithm to frame, continuous calculating parameter gradient in back-propagation algorithm,
And use the continuous autoadapted learning rate of RMSprop, RMSprop can according to the case where change of gradient before come renewal learning
Rate, when RMSprop algorithm uses the t times renewal learning rate of variable MeanSquare (w, t) Lai Baocun when each weight the last period
Between the average value of gradient square continue to optimize parameter according to this variable come autoadapted learning rate, be optimal structure
Solution: where
Mean square error function:
RMSprop formula:
For variable W t moment gradient value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710585474.0A CN107230351B (en) | 2017-07-18 | 2017-07-18 | A kind of Short-time Traffic Flow Forecasting Methods based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710585474.0A CN107230351B (en) | 2017-07-18 | 2017-07-18 | A kind of Short-time Traffic Flow Forecasting Methods based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107230351A CN107230351A (en) | 2017-10-03 |
| CN107230351B true CN107230351B (en) | 2019-08-09 |
Family
ID=59957078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710585474.0A Active CN107230351B (en) | 2017-07-18 | 2017-07-18 | A kind of Short-time Traffic Flow Forecasting Methods based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107230351B (en) |
Families Citing this family (54)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107703564B (en) * | 2017-10-13 | 2020-04-14 | 中国科学院深圳先进技术研究院 | Rainfall prediction method and system and electronic equipment |
| CN107730887B (en) * | 2017-10-17 | 2020-05-15 | 海信集团有限公司 | Method and device for realizing traffic flow prediction and readable storage medium |
| CN108229724B (en) * | 2017-12-06 | 2020-12-22 | 华南理工大学 | Short-term traffic data flow prediction method based on temporal-spatial information fusion |
| CN108021980B (en) * | 2017-12-15 | 2020-04-17 | 中国科学院地理科学与资源研究所 | Fine-scale urban population quantity prediction method based on mobile phone data |
| CN108198413A (en) * | 2017-12-20 | 2018-06-22 | 河南中裕广恒科技股份有限公司 | Blocking method is delayed in the intelligent transportation of a kind of big data and autonomous deep learning |
| CN108205889B (en) * | 2017-12-29 | 2021-04-27 | 长春理工大学 | Method for predicting highway traffic flow based on convolutional neural network |
| CN110020738A (en) * | 2018-01-10 | 2019-07-16 | 北京京东金融科技控股有限公司 | For the method, apparatus of data processing, electronic equipment and computer-readable medium |
| CN108280441A (en) * | 2018-02-09 | 2018-07-13 | 北京天元创新科技有限公司 | A kind of crowd's method for predicting and system |
| CN108399749A (en) * | 2018-03-14 | 2018-08-14 | 西南交通大学 | A kind of traffic trip needing forecasting method in short-term |
| CN108734958B (en) * | 2018-04-25 | 2021-04-20 | 江苏大学 | Traffic speed prediction method |
| CN108564228A (en) * | 2018-04-26 | 2018-09-21 | 重庆大学 | A method of based on the temporal aspect predicted orbit traffic OD volumes of the flow of passengers |
| CN108615360B (en) * | 2018-05-08 | 2022-02-11 | 东南大学 | Traffic demand day-to-day evolution prediction method based on neural network |
| CN108510741B (en) * | 2018-05-24 | 2020-10-30 | 浙江工业大学 | Conv1D-LSTM neural network structure-based traffic flow prediction method |
| CN108829766B (en) * | 2018-05-29 | 2021-09-21 | 苏州大学 | Interest point recommendation method, system, equipment and computer readable storage medium |
| CN108564790B (en) * | 2018-06-12 | 2020-12-22 | 国交空间信息技术（北京）有限公司 | Urban short-term traffic flow prediction method based on traffic flow space-time similarity |
| CN108876045A (en) * | 2018-06-25 | 2018-11-23 | 上海应用技术大学 | Emergency tender optimal route recommended method based on LSTM model prediction |
| CN108682418B (en) * | 2018-06-26 | 2022-03-04 | 北京理工大学 | Speech recognition method based on pre-training and bidirectional LSTM |
| CN108846446B (en) * | 2018-07-04 | 2021-10-12 | 国家新闻出版广电总局广播科学研究院 | Target detection method based on multi-path dense feature fusion full convolution network |
| CN109299401B (en) * | 2018-07-12 | 2022-02-08 | 中国海洋大学 | Metropolitan area space-time flow prediction method based on mixed deep learning model LSTM-ResNet |
| CN109242140A (en) * | 2018-07-24 | 2019-01-18 | 浙江工业大学 | A kind of traffic flow forecasting method based on LSTM_Attention network |
| CN109146156B (en) * | 2018-08-03 | 2021-12-03 | 大连理工大学 | Method for predicting charging amount of charging pile system |
| CN108898838B (en) * | 2018-08-03 | 2020-08-28 | 首都经济贸易大学 | Method and device for predicting airport traffic jam based on LSTM model |
| CN110929543B (en) * | 2018-08-30 | 2023-08-25 | 阿里巴巴集团控股有限公司 | Meteorological data processing method and device |
| CN109522600B (en) * | 2018-10-16 | 2020-10-16 | 浙江大学 | Complex equipment residual service life prediction method based on combined deep neural network |
| CN109493599A (en) * | 2018-11-16 | 2019-03-19 | 南京航空航天大学 | A kind of Short-time Traffic Flow Forecasting Methods based on production confrontation network |
| CN109886444B (en) * | 2018-12-03 | 2023-07-11 | 深圳市北斗智能科技有限公司 | Short-time traffic passenger flow prediction method, device, equipment and storage medium |
| CN109726903B (en) * | 2018-12-19 | 2020-10-16 | 中国电子科技集团公司信息科学研究院 | Attention mechanism-based distributed multi-agent cooperative decision method |
| CN109658695B (en) * | 2019-01-02 | 2020-09-22 | 华南理工大学 | Multi-factor short-term traffic flow prediction method |
| CN109711640A (en) * | 2019-01-23 | 2019-05-03 | 北京工业大学 | A kind of Short-time Traffic Flow Forecasting Methods based on fuzzy C-mean algorithm magnitude of traffic flow cluster and error feedback convolutional neural networks |
| CN109741604A (en) * | 2019-03-05 | 2019-05-10 | 南通大学 | Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow |
| CN110021165A (en) * | 2019-03-18 | 2019-07-16 | 浙江工业大学 | A kind of traffic flow forecasting method based on Autoencoder-LSTM Fusion Model |
| CN110059878B (en) * | 2019-04-15 | 2021-04-02 | 中国计量大学 | Photovoltaic power generation power prediction model based on CNN LSTM and construction method thereof |
| CN110675621B (en) * | 2019-05-23 | 2021-01-15 | 北京嘀嘀无限科技发展有限公司 | System and method for predicting traffic information |
| CN110223517B (en) * | 2019-06-20 | 2021-05-18 | 青岛科技大学 | Short-term traffic flow prediction method based on space-time correlation |
| CN110310479A (en) * | 2019-06-20 | 2019-10-08 | 云南大学 | A kind of Forecast of Urban Traffic Flow forecasting system and method |
| CN110516833A (en) * | 2019-07-03 | 2019-11-29 | 浙江工业大学 | A method of the Bi-LSTM based on feature extraction predicts road traffic state |
| CN110335466B (en) * | 2019-07-11 | 2021-01-26 | 青岛海信网络科技股份有限公司 | Traffic flow prediction method and apparatus |
| CN110543825A (en) * | 2019-08-01 | 2019-12-06 | 江苏濠汉信息技术有限公司 | Dangerous construction behavior identification method and device based on space-time characteristics |
| CN110517484A (en) * | 2019-08-06 | 2019-11-29 | 南通大学 | Diamond interchange area planar crossing sign occlusion prediction model building method |
| CN110491146B (en) * | 2019-08-21 | 2020-08-21 | 浙江工业大学 | Deep learning-based traffic signal control scheme real-time recommendation method |
| CN110766942B (en) * | 2019-10-18 | 2020-12-22 | 北京大学 | Traffic network congestion prediction method based on convolution long-term and short-term memory network |
| CN110827544B (en) * | 2019-11-11 | 2022-09-02 | 重庆邮电大学 | Short-term traffic flow control method based on graph convolution recurrent neural network |
| CN111210088B (en) * | 2020-01-16 | 2023-06-02 | 上海理工大学 | Traffic state index prediction method based on space-time factors |
| CN111462479A (en) * | 2020-02-26 | 2020-07-28 | 南京新一代人工智能研究院有限公司 | Traffic flow prediction method based on Fourier-recurrent neural network |
| CN111524349B (en) * | 2020-04-14 | 2021-05-25 | 长安大学 | Context feature injected multi-scale traffic flow prediction model establishing method and using method |
| CN111639791A (en) * | 2020-05-11 | 2020-09-08 | 同济大学 | Traffic flow prediction method, system, storage medium and terminal |
| CN111860997B (en) * | 2020-07-17 | 2021-12-28 | 海南大学 | Cross-data, information, knowledge modality and dimension early warning method and component |
| CN112118143B (en) * | 2020-11-18 | 2021-02-19 | 迈普通信技术股份有限公司 | Traffic prediction model training method, traffic prediction method, device, equipment and medium |
| CN112562339B (en) * | 2020-12-09 | 2022-01-04 | 福州大学 | Robust traffic flow prediction method based on multitask graph convolutional network |
| CN112633579B (en) * | 2020-12-24 | 2024-01-12 | 中国科学技术大学 | Traffic flow migration prediction method based on domain countermeasure |
| CN113112789B (en) * | 2021-03-08 | 2023-07-14 | 张晔婷 | Influence prediction and control method for urban expressway emergency |
| CN113240904B (en) * | 2021-05-08 | 2022-06-14 | 福州大学 | Traffic flow prediction method based on feature fusion |
| CN113313303A (en) * | 2021-05-28 | 2021-08-27 | 南京师范大学 | Urban area road network traffic flow prediction method and system based on hybrid deep learning model |
| CN114861091B (en) | 2022-07-11 | 2022-11-01 | 成都秦川物联网科技股份有限公司 | Smart city traffic path determination method, internet of things system, device and medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106652441A (en) * | 2015-11-02 | 2017-05-10 | 杭州师范大学 | Urban road traffic condition prediction method based on spatial-temporal data |
| CN105513350A (en) * | 2015-11-30 | 2016-04-20 | 华南理工大学 | Time-phased multi-parameter short-term traffic flow prediction method based on time-space characteristics |
| CN105702029B (en) * | 2016-02-22 | 2017-11-28 | 北京航空航天大学 | A kind of Expressway Traffic trend prediction method for considering space-time relationship at times |
| CN106355879A (en) * | 2016-09-30 | 2017-01-25 | 西安翔迅科技有限责任公司 | Time-space correlation-based urban traffic flow prediction method |
-
2017
- 2017-07-18 CN CN201710585474.0A patent/CN107230351B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107230351A (en) | 2017-10-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107230351B (en) | A kind of Short-time Traffic Flow Forecasting Methods based on deep learning | |
| Li et al. | Prediction for tourism flow based on LSTM neural network | |
| Luo et al. | ESTNet: embedded spatial-temporal network for modeling traffic flow dynamics | |
| CN107180530B (en) | A kind of road network trend prediction method based on depth space-time convolution loop network | |
| CN106981198B (en) | Deep learning network model for travel time prediction and establishing method thereof | |
| Gu et al. | Short-term prediction of lane-level traffic speeds: A fusion deep learning model | |
| Duan et al. | Improved deep hybrid networks for urban traffic flow prediction using trajectory data | |
| Chen et al. | A graph convolutional stacked bidirectional unidirectional-LSTM neural network for metro ridership prediction | |
| Fang et al. | FTPG: A fine-grained traffic prediction method with graph attention network using big trace data | |
| CN110223517A (en) | Short-term traffic flow forecast method based on temporal correlation | |
| CN110570651A (en) | Road network traffic situation prediction method and system based on deep learning | |
| CN105513359B (en) | A kind of urban expressway traffic method for estimating state based on smart mobile phone movement detection | |
| Chen et al. | A short-term flood prediction based on spatial deep learning network: A case study for Xi County, China | |
| CN108830430A (en) | Convolutional neural networks multiple spot regressive prediction model for traffic flow forecasting | |
| CN111860951A (en) | Rail transit passenger flow prediction method based on dynamic hypergraph convolutional network | |
| Nourani et al. | Integrated artificial neural network for spatiotemporal modeling of rainfall–runoff–sediment processes | |
| CN107748942B (en) | Radar Echo Extrapolation prediction technique and system based on velocity field sensing network | |
| CN111612243A (en) | Traffic speed prediction method, system and storage medium | |
| CN108205889A (en) | Freeway traffic flow Forecasting Methodology based on convolutional neural networks | |
| CN112365708B (en) | Scenic spot traffic volume prediction model establishing and predicting method based on multi-graph convolution network | |
| Shen et al. | Research on traffic speed prediction by temporal clustering analysis and convolutional neural network with deformable kernels (May, 2018) | |
| CN108734958A (en) | A kind of traffic speed prediction technique | |
| CN110110243A (en) | A kind of historical track destination prediction technique based on echo state network | |
| CN109447374A (en) | A kind of new shop quick site selection method and apparatus based on big data analysis | |
| Chen et al. | A multiscale-grid-based stacked bidirectional GRU neural network model for predicting traffic speeds of urban expressways |
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 |