CN102637363A - SVM (Support Vector Machine)-based road vehicle running speed prediction method - Google Patents
SVM (Support Vector Machine)-based road vehicle running speed prediction method Download PDFInfo
- Publication number
- CN102637363A CN102637363A CN2012101049181A CN201210104918A CN102637363A CN 102637363 A CN102637363 A CN 102637363A CN 2012101049181 A CN2012101049181 A CN 2012101049181A CN 201210104918 A CN201210104918 A CN 201210104918A CN 102637363 A CN102637363 A CN 102637363A
- Authority
- CN
- China
- Prior art keywords
- prediction
- svm
- data
- road
- vehicle running
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention belongs to the intelligent traffic field, and relates to a SVM(Support Vector Machine)-based road vehicle running speed prediction method. The method comprises the following steps of: acquiring actual measured road speed data of a to-be-predicted road; normalizing the acquired actual measured road speed data, and grouping to get a data set; selecting a radial basis function as a kernel function of the SVM, and obtaining an optimal parameter of quadratic programming by a dynamic regulation method; solving a prediction function; predicting the road vehicle running speed according to the prediction function generated in the previous step, comparing the prediction result with the data in a test sample set, and evaluating the prediction error; and if the error is large, predicting after regulating the SVM parameter. With the prediction method of the invention, the generalization ability of a learning machine is improved, the local minimum problem is avoided; and in unsteady short-time traffic flow information prediction, even if the input signals are mixed with noise, the prediction precision also can be very high.
Description
Technical field
The invention belongs to intelligent transportation field, can directly apply to certain section interior Vehicle Speed of a road period of prediction.
Background technology
These years, China's economy continues to keep rapid growth, and living standards of the people significantly improve; The urbanization degree obviously improves; Automobile consumption also grows with each passing day, and it is outstanding more and thorny that the urban traffic blocking problem becomes, and people's work and life have been produced great influence.These influences mainly show the following aspects: the first, and time waste; The second, the wasting of resources; The 3rd, reduce accident processing response speed.Therefore, through being predicted, road vehicle running speed judges that the congestion in road situation just seems very important.Simultaneously; Along with development of ITS; Intelligent transportation control and vehicle guidance system become one of core topic of intelligent transportation system research; But it is the prediction of road vehicle running speed accurately in real time that these systems are able to existing prerequisite and key, and the Vehicle Speed accuracy of predicting will be directly connected to traffic control and vehicle guidance effect, so predicts during Vehicle Speed and more and more come into one's own.
In setting up the traffic information predicting model, neural network relies on advantages such as its ability of approaching any nonlinear function and fault-tolerant, the self study that is had, and by the Chinese scholars widespread use, and has obtained many effectively achievements in research.But neural network is a kind of heuristic technique that depends on experience, and learning process adopts the empiric risk minimization principle, under the small sample situation, occurs the study phenomenon easily and causes generalization ability low; In addition, for the short-term traffic flow information of non-stationary, when input signal was mixed with noise, the ratio of precision of neural network prediction was lower.
Summary of the invention
The above-mentioned deficiency that the objective of the invention is to overcome prior art provides a kind of more accurate and reliable traffic information predicting method.Technical scheme of the present invention is following:
A kind of road vehicle running speed Forecasting Methodology based on SVMs comprises the following steps:
1) the road speeds measured data of collection road to be predicted;
2) the road speeds measured data that is collected is carried out normalization and handle, and divide into groups, obtain data set, wherein a part is used for training pattern, is called training sample set, and another part is used for finally estimating predicting the outcome, and is called the test sample book collection;
3) select the kernel function of RBF (RBF), adopt dynamic adjustment method to obtain the parameters optimization of quadratic programming as SVM;
4) utilize training sample set to try to achieve Lagrange multiplier a
i(i=1,2 ..., l) with biasing b, thereby confirm anticipation function
Wherein K is a RBF, and x is a vector data to be predicted.
5) anticipation function that generates according to a last step predicts road vehicle running speed, and will predict the outcome and compare with the data of test sample book collection, and predicated error is estimated; If error is bigger, then return step 3, readjust the SVM parameter and predict again.
The present invention use based on the theoretical traffic information predicting algorithm of SVMs, remedied the deficiency of the neural network prediction algorithm of current extensive employing, solved and crossed study and cause the low problem of generalization ability; Adopt the method for support vector regression; Both improved the generalization ability of learning machine, and made local minimum problem not exist, and made again and carrying out the short-term traffic flow information prediction of non-stationary; Even input signal is mixed with noise, still can obtain very high precision.
Description of drawings
The key step diagram of Fig. 1 Forecasting Methodology of the present invention.
Fig. 2 experimental result synoptic diagram, series 1 is represented the real road car speed, serial 2 generation system prediction car speed.Fig. 2 ordinate is represented Vehicle Speed, and unit is kilometer/hour, and horizontal ordinate is represented time point, and unit is h.
Embodiment
The present invention is a kind of based on the theoretical road vehicle running speed prediction algorithm of SVMs (SVM), mainly is to use the theory of SVM to come to predict accurately the Vehicle Speed of certain section road in following a period of time.
SVMs (SVM) is a kind of novel machine learning method; It has complete theoretical foundation and outstanding learning performance; Its outstanding feature is to learn according to structural risk minimization (SRM); Can not have local minimization problem from improving the generalization ability of learning machine in essence, and the utilization kernel function has solved problem of dimension dexterously.
Traffic information predicting is core missions of intelligent transportation field; The invention provides a kind of theoretical method of SVMs of using and predict traffic flow; Through historical traffic information is handled; The interior Vehicle Speed through a certain fixedly highway section of prediction certain hour, thereby the jam situation of judgement road are for people's trip and road early warning decision etc. provide a good platform.The concrete realization is such, like Fig. 1:
1) the road speeds measured data of collection road to be predicted,
2) the road speeds measured data that is collected is carried out normalization and handles, and divide into groups, obtain data set:
3) select kernel function, confirm the SVM parameter.
Obtain after the data set, select the kernel function of RBF (RBF), confirm SVM parameters needed C and ε simultaneously as SVM.
Here; Parameters optimization C and ε are introduced the dynamically thought of adjustment; Promptly confirm second parameter with enumeration method with fixing first parameter of priori; And then the fixing parameter of having optimized confirms first parameter, and two parameters after will optimizing are at last verified its optimality in neighborhood separately.
4) input data set, the generation forecast function
The input sample set is tried to achieve Lagrange multiplier a
i(i=1,2 ..., l) with biasing b, thereby confirm anticipation function
Wherein K is a RBF, and x is a vector data to be predicted.
5) predict and miss analysis, like Fig. 2.
Anticipation function according to a last step generates predicts, and will predict the outcome and compare with the data of test sample book collection, and predicated error is estimated; If error is bigger, then return step 3, readjust the SVM parameter and predict again.
The present invention is incorporated into intelligent transportation field with the theory of SVM, carries out the prediction of road vehicle speed, for traffic information predicting provides new thinking.Thought of the present invention is to use Nonlinear Mapping to be feature space to sample vector from former spatial mappings to height, at this high-order feature space structure optimal decision function, utilize the construction minimizes principle, and the utilization kernel function has solved problem of dimension cleverly.Its core has been to utilize the thinking of support vector regression to come predicted traffic information.
Below in conjunction with embodiment the present invention is further specified.When present embodiment invention use is carried out traffic information predicting based on the SVM theory, mainly be divided into data processing, parameter optimization, four steps of model training and prediction:
Data processing
This part mainly comprises the collection of data, packet and data normalization.Data collection section needs to gather certain section whole day data that fixing road is continuous 5 days, and the statistical interval of these data is 15min; Packet is a unit with the sky, and these data of 5 days are divided into 5 groups, and 96 pairs every group, preceding four groups are used to train SVM, and last group data are used for verifying the car speed prediction accuracy; Data normalization, with the data normalization to 0 that obtains between 1.
Parameter optimization
Select suitable kernel function, confirm parameter.Through contrast polynomial kernel function, three kinds of different kernel functions of linear kernel function and RBF are found RBF owing to other kernel functions, and the present invention uses the kernel function of RBF (RBF) as SVM; Adopt dynamic adjustment method to obtain the parameters optimization of quadratic programming, optimizing the back parameter is C=1024, ε=2, and C is the complexity of regression model and the compromise between the sample fitting precision here, ε is for returning the maximum error that allows.
Model training
This part is the core of whole algorithm; The model that obtains has directly determined final prediction accuracy; Input is with the ready training data in front, thus definite final training pattern, and this model is exactly the instrument that will be used for following a certain moment car speed after us.
Prediction
The model that use is trained is out predicted, data and real data that prediction is obtained compare.
Fig. 2 has described in 24 hours a whole day the actual speed of every vehicle ' at a distance from 15 minutes and the comparison of predetermined speed, from contrast, can find that the Forecasting Methodology that the present invention adopted can be predicted the situation of the travel speed of following vehicle basically.
Claims (1)
1. based on the road vehicle running speed Forecasting Methodology of SVMs, comprise the following steps:
1) the road speeds measured data of collection road to be predicted;
2) the road speeds measured data that is collected is carried out normalization and handle, and divide into groups, obtain data set, wherein a part is used for training pattern, is called training sample set, and another part is used for finally estimating predicting the outcome, and is called the test sample book collection;
3) select the kernel function of RBF (RBF), adopt dynamic adjustment method to obtain the parameters optimization of quadratic programming as SVM;
4) utilize training sample set to try to achieve Lagrange multiplier a
i(i=1,2 ..., l) with biasing b, thereby confirm anticipation function
Wherein K is a RBF, and x is a vector data to be predicted.
Anticipation function according to a last step generates predicts road vehicle running speed, and will predict the outcome and compare with the data of test sample book collection, and predicated error is estimated; If error is bigger, then return step 3, readjust the SVM parameter and predict again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101049181A CN102637363A (en) | 2012-04-11 | 2012-04-11 | SVM (Support Vector Machine)-based road vehicle running speed prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101049181A CN102637363A (en) | 2012-04-11 | 2012-04-11 | SVM (Support Vector Machine)-based road vehicle running speed prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102637363A true CN102637363A (en) | 2012-08-15 |
Family
ID=46621743
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012101049181A Pending CN102637363A (en) | 2012-04-11 | 2012-04-11 | SVM (Support Vector Machine)-based road vehicle running speed prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102637363A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646553A (en) * | 2013-11-11 | 2014-03-19 | 北京信息科技大学 | Investigation system for road traffic flow and realization method thereof |
| CN104143260A (en) * | 2013-05-10 | 2014-11-12 | 北京航天长峰科技工业集团有限公司 | Vehicle track predicating method based on data fusion optimization model |
| CN107451599A (en) * | 2017-06-28 | 2017-12-08 | 青岛科技大学 | A kind of traffic behavior Forecasting Methodology of the collective model based on machine learning |
| CN110213827A (en) * | 2019-05-24 | 2019-09-06 | 南京理工大学 | Vehicle data collection frequency dynamic adjusting method based on deeply study |
| FR3097471A1 (en) | 2019-06-19 | 2020-12-25 | Psa Automobiles Sa | EQUIVALENT TEMPERATURE REGULATION PROCESS IN A COCKPIT OF A MEANS OF TRANSPORT EQUIPPED WITH AN AIR CONDITIONING SYSTEM |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101271625A (en) * | 2008-04-03 | 2008-09-24 | 东南大学 | Method for detecting freeway traffic event by integration supporting vector machine |
| WO2009075443A1 (en) * | 2007-12-11 | 2009-06-18 | Thinkware Systems Corporation | Method and apparatus for estimating traffic flow |
-
2012
- 2012-04-11 CN CN2012101049181A patent/CN102637363A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009075443A1 (en) * | 2007-12-11 | 2009-06-18 | Thinkware Systems Corporation | Method and apparatus for estimating traffic flow |
| CN101271625A (en) * | 2008-04-03 | 2008-09-24 | 东南大学 | Method for detecting freeway traffic event by integration supporting vector machine |
Non-Patent Citations (4)
| Title |
|---|
| 姚智胜,邵春福,高永亮: "基于支持向量回归机的交通状态短时预测方法研究", 《北京交通大学学报》, vol. 30, no. 3, 30 June 2006 (2006-06-30), pages 19 - 22 * |
| 徐启华，杨 瑞: "支持向量机在交通流量实时预测中的应用", 《公路交通科技》, vol. 22, no. 12, 15 December 2005 (2005-12-15), pages 131 - 134 * |
| 温惠英，李俊辉: "基于改进支持向量机的交通流量预测算法研究", 《交通与计算机》, vol. 26, no. 2, 15 April 2008 (2008-04-15), pages 5 - 6 * |
| 郑勋烨，黄晶晶: "基于支持向量回归机的交通状态短时预测和北京某区域实例分析", 《数学的实践与认识》, vol. 40, no. 10, 23 May 2010 (2010-05-23), pages 77 - 83 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104143260A (en) * | 2013-05-10 | 2014-11-12 | 北京航天长峰科技工业集团有限公司 | Vehicle track predicating method based on data fusion optimization model |
| CN104143260B (en) * | 2013-05-10 | 2017-10-03 | 北京航天长峰科技工业集团有限公司 | A kind of track of vehicle Forecasting Methodology based on data fusion Optimized model |
| CN103646553A (en) * | 2013-11-11 | 2014-03-19 | 北京信息科技大学 | Investigation system for road traffic flow and realization method thereof |
| CN107451599A (en) * | 2017-06-28 | 2017-12-08 | 青岛科技大学 | A kind of traffic behavior Forecasting Methodology of the collective model based on machine learning |
| CN110213827A (en) * | 2019-05-24 | 2019-09-06 | 南京理工大学 | Vehicle data collection frequency dynamic adjusting method based on deeply study |
| FR3097471A1 (en) | 2019-06-19 | 2020-12-25 | Psa Automobiles Sa | EQUIVALENT TEMPERATURE REGULATION PROCESS IN A COCKPIT OF A MEANS OF TRANSPORT EQUIPPED WITH AN AIR CONDITIONING SYSTEM |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Research on optimized GA-SVM vehicle speed prediction model based on driver-vehicle-road-traffic system | |
| CN103093643B (en) | Public parking lot berth quantity confirming method | |
| CN106205156B (en) | A kind of intersection self-healing combination control method for the mutation of part lane flow | |
| CN106816008B (en) | A kind of congestion in road early warning and congestion form time forecasting methods | |
| CN102637363A (en) | SVM (Support Vector Machine)-based road vehicle running speed prediction method | |
| CN104008644B (en) | A kind of traffic noise on urban roads measuring method based on Gradient Descent | |
| CN104166171B (en) | The Forecasting Methodology of thundercloud area coverage based on lightning location system and system | |
| CN104992244A (en) | Airport freight traffic prediction analysis method based on SARIMA and RBF neural network integration combination model | |
| CN106945530A (en) | A kind of electric automobile and its continual mileage Forecasting Methodology, system | |
| CN102938562B (en) | Prediction method of total wind electricity power in area | |
| CN103496368A (en) | Automobile cooperative type self-adaptive cruise control system and method with learning ability | |
| CN105243428A (en) | Bus arrival time prediction method through optimizing support vector machine based on bat algorithm | |
| CN103854518A (en) | Calculating method of space-time flow of air route network nodes | |
| Li et al. | The study on multi-scale prediction of future driving cycle based on Markov chain | |
| CN104851280A (en) | Vehicle driving control method, device, system and related equipment | |
| CN105551250A (en) | Method for discriminating urban road intersection operation state on the basis of interval clustering | |
| CN103400039A (en) | Wind power climbing prediction model switching method based on gale weather classification | |
| CN108364490A (en) | Municipal highway transit system vehicle runs regulation and control method | |
| CN116187161A (en) | Intelligent energy management method and system for hybrid electric bus in intelligent networking environment | |
| CN108364491A (en) | Driver information based reminding method based on vehicle flowrate prediction | |
| CN109255948A (en) | A kind of divided lane wagon flow scale prediction method based on Kalman filtering | |
| CN117351734A (en) | Intelligent regulation and control method and system for vehicle delay | |
| Li et al. | Traffic flow prediction with big data: A learning approach based on SIS-complex networks | |
| Farhi et al. | The Logit lane assignment model: first results | |
| CN105469157A (en) | Spatial heterogeneity analysis method for sensitivity of roadside parking demand to parking fee |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120815 |