CN109979195A - A kind of short-term traffic flow forecast method of the fusion Spatio-temporal factors based on sparse regression - Google Patents
A kind of short-term traffic flow forecast method of the fusion Spatio-temporal factors based on sparse regression Download PDFInfo
- Publication number
- CN109979195A CN109979195A CN201910222787.9A CN201910222787A CN109979195A CN 109979195 A CN109979195 A CN 109979195A CN 201910222787 A CN201910222787 A CN 201910222787A CN 109979195 A CN109979195 A CN 109979195A
- Authority
- CN
- China
- Prior art keywords
- dictionary
- traffic flow
- spatio
- indicate
- formula
- 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.)
- Granted
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
-
- 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
- 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
Abstract
The short-term traffic flow forecast method for merging Spatio-temporal factors based on sparse regression that the present invention relates to a kind of, comprising steps of 1) traffic flow data pre-processes: being handled using min-max standardization traffic flow data prediction, the traffic flow data of each sensing point is normalized into [0,1] range;2) Spatio-temporal factors dictionary constructs: according to formula come build time factor dictionary, space factor dictionary is constructed according to formula;3) solution and prediction of sparse coefficient: sparse coding is carried out according to formula to training data, sparse coefficient α is solved, the magnitude of traffic flow at next moment is predicted according to sparse coefficient α combination Spatio-temporal factors dictionary.The beneficial effects of the present invention are: the sparse regression prediction technique (ST-SR) of fusion Spatio-temporal factors proposed by the present invention will be significantly better than other several prediction techniques, and it is able to achieve the influence of quantitative analysis factor.
Description
Technical field
The present invention relates to a kind of short-term traffic flow forecast methods, more specifically, it is related to one kind based on sparse regression
Fusion Spatio-temporal factors short-term traffic flow forecast method.
Background technique
Currently, conventional time series prediction technique, pre- in the magnitude of traffic flow based on the Time Series Forecasting Methods of machine learning
There is related application in survey problem.Short-term traffic flow forecast model based on SVM Yu adaptive space-time data fusion, Li Qiaoru
Deng, Beijing University of Technology's journal discloses a kind of short-term traffic flow forecast model for considering space factor on April 3rd, 2015,
Model main thought is that use space sequence prediction value is modified Traffic Flow Time Series prediction result.A kind of magnitude of traffic flow
Prediction technique and system, Cai Xiaoyu etc., patent of invention, August in 2018 disclose within 3rd a kind of magnitude of traffic flow prediction technique and
System, method is according to the same day observed volume information architecture mesh at the A moment successively obtained from default situation database to T moment
Sequence is marked, fusion distance matrix is then constructed according to target sequence and matching sequence matrix, according to fusion distance matrix, default system
Several and preset algorithm determines anticipation function.A kind of improvement gravitation search least square method supporting vector machine forecasting traffic flow, Xu Qinshuai
Deng, computer application research, disclose on September 30th, 2018 a kind of new improvement gravitation search algorithm (TCK-AGSA) to its into
Row parameter optimization, the experimental results showed that the model effectively increases the precision of prediction.The magnitude of traffic flow based on space-time analysis and CNN
Forecasting Approach for Short-term, money are big etc., control engineering, disclose a kind of short-time traffic flow forecast built-up pattern on January 20th, 2019, should
Model includes 2 submodels of Gray system and ELM (ExtremeLearningMachine) neural network, experiments prove that,
Mentioned method is better than existing some achievements.In conclusion most research work at present is substantially still by current
The historical traffic flows data of website predict the magnitude of traffic flow at next moment.Part research consider in the magnitude of traffic flow when
Empty factor, but these methods often face modeling complexity, model poor expandability can not carry out quantitative analysis etc. to factor and ask
Topic.Therefore, how Spatio-temporal factors are efficiently merged, the accuracy for improving traffic flow forecasting is still the difficult point of research.
Summary of the invention
The purpose of the present invention is to overcome the above shortcomings and to provide a kind of fusion Spatio-temporal factors based on sparse regression are in short-term
Traffic flow forecasting method.
The short-term traffic flow forecast method of fusion Spatio-temporal factors based on sparse regression, includes the following steps:
Step 1: traffic flow data pre-processes
Traffic flow data prediction is handled using min-max standardization, by the traffic flow data of each sensing point
It is normalized into [0,1] range, as shown in formula (2):
Wherein, y indicates that original traffic flow data, min and max respectively indicate the minimum value and maximum value of y, and y' is indicated
Standardized result;
Step 2: Spatio-temporal factors dictionary constructs
1) according to formula (3) come build time factor dictionary, it is assumed that predicted time point isTime factor dictionary T definition
It is as follows:
WhereinIndicate that n-th day traffic flow data in t moment in the past, n indicate that the history number of days chosen, k indicate instruction
Practice the length of data;
2) space factor dictionary is constructed according to formula (4), it is assumed that predicted time point isSpace factor dictionary S definition
It is as follows:
WhereinIndicate that traffic flow data of n-th of peripheral information point at the t-1 moment, n indicate the neighbour chosen points
Amount, k indicate the length of training data;
3) Spatio-temporal factors dictionary D is completedfBuilding;
The orthogonal DCT-II dictionary according to shown in formula (5), Kronecker Delta function shown in formula (6) and
Time factor dictionary and space factor dictionary construct magnitude of traffic flow Spatio-temporal factors dictionary DfBuilding, as shown in formula (7);
Kj(n)=δ (n-j) j=N, N+1 ..., 2N-1 (6)
Wherein i, j indicate that i-th and the jth column of dictionary, n indicate the size of dictionary column vector, and N indicates the quantity of column vector;
Wherein, preceding 2N column are the dictionaries generated by orthogonal DCT-II and Kronecker Delta function, last 2 column difference
It is time factor and space factor;
Step 3: the solution and prediction of sparse coefficient
1) sparse coding is carried out according to formula (8) to training data, solves sparse coefficient α;
WhereinIndicate historical traffic flows,It is the Spatio-temporal factors dictionary of step 2 building, the time is from t1To tk, α
It is sparse coefficient to be solved
2) magnitude of traffic flow that next moment is predicted according to sparse coefficient α combination Spatio-temporal factors dictionary, such as formula (9) institute
Show;
WhereinIndicate the magnitude of traffic flow of subsequent time,It is the Spatio-temporal factors dictionary of step 2 building, α indicates to solve
Obtained sparse coefficient.
The beneficial effects of the present invention are: the sparse regression prediction technique (ST-SR) of fusion Spatio-temporal factors proposed by the present invention
To be significantly better than other several prediction techniques, and be able to achieve the influence of quantitative analysis factor.From the point of view of RMSE, ST-SR model is 4
The consensus forecast precision of a website has been respectively increased 2.70% on 5 minute predicting interval compared to SVR, LSTM and KNN,
1.87%, 2.11%, it is respectively increased 17.71%, 16.59%, 9.53% 15 minute predicting interval, between prediction in 30 minutes
Every being respectively increased 35.60%, 27.63%, 10.83%.From the point of view of MAPE, ST-SR model on 5 minute predicting interval
Consensus forecast precision has been respectively increased 1.73%, 2.91% and 1.69%, also wants excellent 15 minutes and 30 minutes predicting intervals
In contrast model.
Detailed description of the invention
Fig. 1 is this method overview flow chart;
Fig. 2 is the weight distribution figure of different factors.
Specific embodiment
The present invention is described further below with reference to embodiment.The explanation of following embodiments is merely used to help understand this
Invention.It should be pointed out that for those skilled in the art, without departing from the principle of the present invention, also
Can be with several improvements and modifications are made to the present invention, these improvement and modification also fall into the protection scope of the claims in the present invention
It is interior.
The short-term traffic flow forecast method of fusion Spatio-temporal factors based on sparse regression, includes the following steps:
Step 1: traffic flow data pre-processes
In traffic flow data, there may be larger differences for the magnitude of traffic flow numerical value between different sensing points, but this is simultaneously
It does not mean that there is no correlativity between sensing point, in order to preferably excavate the relationship between detection site, is convenient for simultaneously
The training of following model, reducing factor value difference bring influences, and the present invention is standardized using min-max and carried out to the magnitude of traffic flow
Pretreatment.
Step 2: Spatio-temporal factors dictionary constructs
The present invention constructs basic dictionary according to DCT dictionary and Kronecker Delta function first.Then from the time
The correlation between the spatial analysis magnitude of traffic flow, and time factor dictionary and space factor dictionary are constructed respectively, it is finally completed
Magnitude of traffic flow Spatio-temporal factors dictionary DfBuilding, as shown in formula (1):
Df=[D, T, S] (1)
Wherein D is the basic dictionary that DCT dictionary and Kronecker Delta function are constituted, and T is time factor dictionary, and S is
Space factor dictionary.
Step 3: the solution and prediction of sparse coefficient
Sparse decomposition is carried out to historical traffic flows data using the Spatio-temporal factors dictionary established in step 2, is solved sparse
Coefficient finally predicts the magnitude of traffic flow of subsequent time using the sparse coefficient combination Spatio-temporal factors dictionary acquired.
The short-term traffic flow forecast method of the fusion Spatio-temporal factors based on sparse regression of the present embodiment, overview flow chart
As shown in Figure 1, the specific steps are as follows:
Step 1: traffic flow data pre-processes
The present invention is handled traffic flow data prediction using min-max standardization, by the traffic flow of each sensing point
It measures in data normalization to [0,1] range, as shown in formula (2):
Wherein, y indicates that original traffic flow data, min and max respectively indicate the minimum value and maximum value of y, and y' is indicated
Standardized result.
Step 2: Spatio-temporal factors dictionary constructs
4) according to formula (3) come build time factor dictionary, it is assumed that predicted time point isTime factor dictionary T definition
It is as follows:
WhereinIndicate that n-th day traffic flow data in t moment in the past, n indicate that the history number of days chosen, k indicate instruction
Practice the length of data.
5) space factor dictionary is constructed according to formula (4), it is assumed that predicted time point isSpace factor dictionary S definition
It is as follows:
WhereinIndicate that traffic flow data of n-th of peripheral information point at the t-1 moment, n indicate the neighbour chosen points
Amount, k indicate the length of training data.
6) Spatio-temporal factors dictionary D is completedfBuilding.
The orthogonal DCT-II dictionary according to shown in formula (5), Kronecker Delta function shown in formula (6) and
Time factor dictionary and space factor dictionary construct magnitude of traffic flow Spatio-temporal factors dictionary DfBuilding, as shown in formula (7).
Kj(n)=δ (n-j) j=N, N+1 ..., 2N-1 (6)
Wherein i, j indicate that i-th and the jth column of dictionary, n indicate the size of dictionary column vector, and N indicates the quantity of column vector.
Wherein, preceding 2N column are the dictionaries generated by orthogonal DCT-II and Kronecker Delta function, last 2 column difference
It is time factor and space factor.
Step 3: the solution and prediction of sparse coefficient
3) sparse coding is carried out according to formula (8) to training data, solves sparse coefficient α.The present invention uses the scholars such as Kim
It is published in the article An Interior- of IEEE journal of selected topics in signal processing
Method in Point Method for Large-Scale l1-Regularized Least Squares carries out sparse system
Several solutions.
WhereinIndicate historical traffic flows,It is the Spatio-temporal factors dictionary of step 2 building, the time is from t1To tk, α
It is sparse coefficient to be solved
4) magnitude of traffic flow that next moment is predicted according to sparse coefficient α combination Spatio-temporal factors dictionary, such as formula (9) institute
Show.
WhereinIndicate the magnitude of traffic flow of subsequent time,It is the Spatio-temporal factors dictionary of step 2 building, α indicates to solve
Obtained sparse coefficient.
Experiment and result:
Data set used in experiment comes from Caltrans Performance Measurement System (PeMS) net
It stands, which provides the traffic flow data more than 39000 detection sites, in order to preferably verify our prediction techniques
Performance, the present invention have selected to be located at urban and suburban, add up to the traffic flow data of 4 websites to carry out related experiment, 4 stations
The ID of point is respectively 500010021,1201100,1017510 and 400665.
The purpose of this method is to provide a kind of short-term traffic flow forecast method that can merge Spatio-temporal factors.In order to measure this
The validity of method, we compared on data set SVR, KNN, LSTM and it is proposed by the present invention fusion Spatio-temporal factors it is pre-
Survey method (ST-SR).Experimental data is in January, 2017 in June, 2017, and removes festivals or holidays therein.Testing time is to have
Last 30 days of number of days are imitated, i.e., measure the performance of model using 30 days mean errors.The error criterion that the present invention uses is
Most common RMSE and MAPE is used in traffic flow forecasting, respectively as shown in formula (10) and formula (11).
Wherein N indicates the length of prediction, FtAnd AtRespectively indicate the predicted value and true value of model.
Influence of 1 space factor of table to different traffic environments
It tests a pair of four future positions and has carried out comparative experiments, first group only increases time factor, and second group increases simultaneously
Add time factor and space factor.Experimental result is as shown in table 1.Wherein the ambient enviroment of future position 1 and future position 3 is suburb,
And future position 2 and future position 4 are located at city, and as known from Table 1, after increasing space factor, comprehensive RMSE and MAPE analysis, prediction
The precision of prediction promotion of point 2 and 4 will become apparent compared to 1 and 3, show that the magnitude of traffic flow in city is easier by periphery
The influence of traffic condition.Fig. 2 shows the distribution situation of different factor weights in ST-SR model, generally, the power of time factor
It is apparently higher than space factor again, shows that time factor is bigger compared to influence of the space factor to the magnitude of traffic flow.
Experiment two compared the prediction effect of SVR, KNN, LSTM and ST-SR, in order to further show each model
Performance, we carried out related experiment on 5 minutes, 15 minutes and 30 minute predicting intervals, and experimental result is respectively such as table 2, table
3, shown in table 4 and table 5.From the point of view of RMSE, ST-SR model compares SVR, LSTM and KNN in the consensus forecast precision of 4 websites
It has been respectively increased 2.70%, 1.87%, 2.11% on 5 minute predicting interval, has been respectively increased 15 minute predicting interval
17.71%, 16.59%, 9.53%, 35.60%, 27.63%, 10.83% has been respectively increased 30 minute predicting interval.From
From the point of view of MAPE, 1.73%, 2.91% and has been respectively increased in the consensus forecast precision on 5 minute predicting interval of ST-SR model
1.69%, also it is better than contrast model on other two kinds of predicting intervals.The results show ST-SR is in four different locations
Precision of prediction will be apparently higher than other several models, show that ST-SR model can preferably adapt to different traffic environments.
The different prediction models of table 2 compare in the performance of future position 1
The different prediction models of table 3 compare in the performance of future position 2
The different prediction models of table 4 compare in the performance of future position 3
The different prediction models of table 5 compare in the performance of future position 4
Claims (1)
1. a kind of short-term traffic flow forecast method of the fusion Spatio-temporal factors based on sparse regression, which is characterized in that including such as
Lower step:
Step 1: traffic flow data pre-processes
Traffic flow data prediction is handled using min-max standardization, by the traffic flow data standard of each sensing point
Change into [0,1] range, as shown in formula (2):
Wherein, y indicates that original traffic flow data, min and max respectively indicate the minimum value and maximum value of y, and y' indicates standard
The result of change;
Step 2: Spatio-temporal factors dictionary constructs
1) according to formula (3) come build time factor dictionary, it is assumed that predicted time point isTime factor dictionary T is defined as follows:
WhereinIndicate that n-th day traffic flow data in t moment in the past, n indicate that the history number of days chosen, k indicate training number
According to length;
2) space factor dictionary is constructed according to formula (4), it is assumed that predicted time point isSpace factor dictionary S is defined as follows:
WhereinIndicate that traffic flow data of n-th of peripheral information point at the t-1 moment, n indicate the Neighbor Points quantity chosen, k
Indicate the length of training data;
3) Spatio-temporal factors dictionary D is completedfBuilding;
The orthogonal DCT-II dictionary according to shown in formula (5), Kronecker Delta function and time shown in formula (6)
Factor dictionary and space factor dictionary construct magnitude of traffic flow Spatio-temporal factors dictionary DfBuilding, as shown in formula (7);
Kj(n)=δ (n-j) j=N, N+1 ..., 2N-1 (6)
Wherein i, j indicate that i-th and the jth column of dictionary, n indicate the size of dictionary column vector, and N indicates the quantity of column vector;
Wherein, preceding 2N column are the dictionaries generated by orthogonal DCT-II and Kronecker Delta function, when last 2 column are respectively
Between factor and space factor;
Step 3: the solution and prediction of sparse coefficient
1) sparse coding is carried out according to formula (8) to training data, solves sparse coefficient α;
WhereinIndicate historical traffic flows,It is the Spatio-temporal factors dictionary of step 2 building, the time is from t1To tk, α be to
The sparse coefficient of solution
2) magnitude of traffic flow that next moment is predicted according to sparse coefficient α combination Spatio-temporal factors dictionary, as shown in formula (9);
WhereinIndicate the magnitude of traffic flow of subsequent time,It is the Spatio-temporal factors dictionary of step 2 building, α is indicated to solve and be obtained
Sparse coefficient.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910222787.9A CN109979195B (en) | 2019-03-22 | 2019-03-22 | Sparse regression-based short-term traffic flow prediction method integrating space-time factors |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910222787.9A CN109979195B (en) | 2019-03-22 | 2019-03-22 | Sparse regression-based short-term traffic flow prediction method integrating space-time factors |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109979195A true CN109979195A (en) | 2019-07-05 |
CN109979195B CN109979195B (en) | 2020-07-03 |
Family
ID=67080117
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910222787.9A Active CN109979195B (en) | 2019-03-22 | 2019-03-22 | Sparse regression-based short-term traffic flow prediction method integrating space-time factors |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109979195B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1560756A (en) * | 2004-03-09 | 2005-01-05 | 北京交通大学 | Intelligent traffic data processing method |
US20090279738A1 (en) * | 2008-05-08 | 2009-11-12 | Denso Corporation | Apparatus for image recognition |
CN104064023A (en) * | 2014-06-18 | 2014-09-24 | 银江股份有限公司 | Dynamic traffic flow prediction method based on space-time correlation |
CN104331913A (en) * | 2014-11-19 | 2015-02-04 | 西安电子科技大学 | Polarized SAR polarization method based on sparse K-SVD (Singular Value Decomposition) |
CN104574336A (en) * | 2015-01-19 | 2015-04-29 | 上海交通大学 | Super-resolution image reconstruction system based on self-adaptation submodel dictionary choice |
CN105430416A (en) * | 2015-12-04 | 2016-03-23 | 四川大学 | Fingerprint image compression method based on adaptive sparse domain coding |
CN108898829A (en) * | 2018-06-07 | 2018-11-27 | 重庆邮电大学 | The dynamic short-time traffic flow forecast system with Sparse is divided for the indifference opposite sex |
-
2019
- 2019-03-22 CN CN201910222787.9A patent/CN109979195B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1560756A (en) * | 2004-03-09 | 2005-01-05 | 北京交通大学 | Intelligent traffic data processing method |
US20090279738A1 (en) * | 2008-05-08 | 2009-11-12 | Denso Corporation | Apparatus for image recognition |
CN104064023A (en) * | 2014-06-18 | 2014-09-24 | 银江股份有限公司 | Dynamic traffic flow prediction method based on space-time correlation |
CN104331913A (en) * | 2014-11-19 | 2015-02-04 | 西安电子科技大学 | Polarized SAR polarization method based on sparse K-SVD (Singular Value Decomposition) |
CN104574336A (en) * | 2015-01-19 | 2015-04-29 | 上海交通大学 | Super-resolution image reconstruction system based on self-adaptation submodel dictionary choice |
CN105430416A (en) * | 2015-12-04 | 2016-03-23 | 四川大学 | Fingerprint image compression method based on adaptive sparse domain coding |
CN108898829A (en) * | 2018-06-07 | 2018-11-27 | 重庆邮电大学 | The dynamic short-time traffic flow forecast system with Sparse is divided for the indifference opposite sex |
Also Published As
Publication number | Publication date |
---|---|
CN109979195B (en) | 2020-07-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wu et al. | Using gated recurrent unit network to forecast short-term load considering impact of electricity price | |
CN109711617A (en) | A kind of medium-term and long-term Runoff Forecast method based on BLSTM deep learning | |
CN115809833B (en) | Intelligent supervision method and device for foundation project based on portrait technology | |
Tan et al. | Multi-node load forecasting based on multi-task learning with modal feature extraction | |
Jiang et al. | A holistic feature selection method for enhanced short-term load forecasting of power system | |
CN105160441A (en) | Real-time power load forecasting method based on integrated network of incremental transfinite vector regression machine | |
Yang et al. | An intelligent singular value diagnostic method for concrete dam deformation monitoring | |
Lee et al. | Dynamic BIM component recommendation method based on probabilistic matrix factorization and grey model | |
CN115860177A (en) | Photovoltaic power generation power prediction method based on combined machine learning model and application thereof | |
Yang et al. | Comparative Study on Deformation Prediction Models of Wuqiangxi Concrete Gravity Dam Based onMonitoring Data. | |
Kong et al. | A variational bayesian inference-based En-Decoder framework for traffic flow prediction | |
Li et al. | Fusion attention mechanism bidirectional LSTM for short-term traffic flow prediction | |
Anh et al. | Effect of gradient descent optimizers and dropout technique on deep learning LSTM performance in rainfall-runoff modeling | |
CN113206756A (en) | Network flow prediction method based on combined model | |
CN109979195A (en) | A kind of short-term traffic flow forecast method of the fusion Spatio-temporal factors based on sparse regression | |
He et al. | A Chinese named entity recognition model of maintenance records for power primary equipment based on progressive multitype feature fusion | |
Cai et al. | Short-term load forecasting for city holidays based on genetic support vector machines | |
CN113642785B (en) | Method, system and equipment for long-term prediction of space debris track based on priori information | |
Zhao et al. | A Hybrid Time Series Model based on Dilated Conv1D and LSTM with Applications to PM2. 5 Forecasting. | |
Zhang et al. | Short-term traffic flow prediction model based on deep learning regression algorithm | |
Wang et al. | The Application Research of Deep Learning Based Method for Short-term Wind Speed Forecasting | |
Kang et al. | Analysis of tourist flow forecasting model based on multiple additive regression tree | |
CN113011657A (en) | Method and device for predicting typhoon water level, electronic equipment and storage medium | |
CN111191826A (en) | Load prediction method based on cosine similarity classification | |
Chang et al. | Real-time prediction of ocean observation data based on transformer 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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |