CN107331164B  A kind of prediction technique of freeway toll station entrance vehicle number  Google Patents
A kind of prediction technique of freeway toll station entrance vehicle number Download PDFInfo
 Publication number
 CN107331164B CN107331164B CN201710613932.7A CN201710613932A CN107331164B CN 107331164 B CN107331164 B CN 107331164B CN 201710613932 A CN201710613932 A CN 201710613932A CN 107331164 B CN107331164 B CN 107331164B
 Authority
 CN
 China
 Prior art keywords
 step
 number
 time window
 set
 data
 Prior art date
Links
Abstract
Description
Technical field
The invention belongs to the technical field of traffic data prediction more particularly to a kind of freeway toll station entrance vehicle numbers Prediction technique.
Background technology
With the quickening of Urbanization in China, contacting between city and city is increasingly close, Intercity Transportation problem day Benefit is prominent.Highway is crowd in the important channel of each intercity passage, feature more close and bayonet vehicle flowrate by when Between sequence be affected.Can in real time, accurately carry out freeway toll station entrance vehicle number prediction vehicle flowrate is dredged, Solve the problems, such as that highspeed transit is most important.However, there are still problems at this stage：
Shortage is compared in correlative study for express highway section flow and highway bayonet flow；Due to a lack of machine The fusion calculation of learning algorithm and highway bayonet data, traditional prediction method precision be not high；Based on perception such as video, coils The cognitive method of device can only perceive realtime bayonet vehicle number, cannot achieve the prediction of entrance vehicle number.
It follows that in order to solve in the prior art can not Accurate Prediction freeway toll station entrance vehicle number ask Topic, it is really necessary to provide a kind of method accurately predicted for freeway toll station entrance vehicle number.
Invention content
For in the prior art can not Accurate Prediction freeway toll station entrance vehicle number the problem of, the present invention provide one The prediction technique of kind freeway toll station entrance vehicle number realizes accurate prediction freeway toll station entrance vehicle number.
The present invention provides a kind of prediction technique of freeway toll station entrance vehicle number, and method includes：
Step 1：It obtains the ETC data of charge station on highway and carries out denoising；
Wherein, the ETC data include that vehicle drives into the time of charge station and the position of the charge station driven into, vehicle The position of charge station for being driven out to the time of charge station and being driven out to；
Step 2：ETC data in step 1 are grouped according to the preset duration of time window and count each charge It stands going out, entering vehicle number in daily each time window；
Step 3：According to step 2 count vehicle number structure target charge station object time window sample set M；
Wherein, all charge stations K time window before object time window drives into vehicle number, outgoing vehicles in any one day Number and the collection of any one day target data are combined into a sample of sample set, K >=2；
Target data is that vehicle number is driven into target charge station in object time window；
The sample set M includes n sample, and n is the number of days of sample training period；
Wherein, M={ (X_{1},y_{1}),(X_{2},y_{2}),…,(X_{n},y_{n})},X_{n}∈R^{d},y_{n}∈R
(X_{n},y_{n}) indicate nth day corresponding sample in sample set M, X_{n}Indicate nth day all charge station in the object time The set for driving into vehicle number, outgoing vehicles number of the preceding K time window of window；y_{n}Indicate that nth day target data, R are expressed as reality Manifold, R^{d}It is expressed as the set of real numbers of d dimensional features；
Step 4：It is selected to the target charge station from d feature in the sample set using recursion elimination algorithm The traffic capacity influences maximum N number of feature in object time window；
Wherein, N is positive integer, 1<N<d；
Within the sample training period when any one in the preceding K time window of object time window of any one charge station Between the collection for driving into vehicle number or outgoing vehicles number of window be combined into a feature of the sample set；
Step 5：According to the N number of feature construction training set m selected in step 4, and the data in the training set m are defeated Enter support vector regression establish target charge station object time window regression model；
Wherein, any one day all drive into vehicle number, outgoing vehicles number and described arbitrary in selected N number of feature The collection of one day target data is combined into the sample of a training set；
The training set m is made of the sample of n training set in the sample training period；
Wherein, m={ (x_{1},y_{1}),(x_{2},y_{2}),…,(x_{n},y_{n})},x_{n}∈R^{N},y_{n}∈R
(x_{n},y_{n}) indicate the sample of nth day corresponding training set in training set m, x_{n}Indicate N number of feature selected by step 4 In nth day set for driving into vehicle number, outgoing vehicles number；y_{n}Indicate nth day target data, R^{N}It is expressed as the reality of Ndimensional feature Manifold；
Step 6：The object time window at the target charge station is identical as N number of feature in step 4 in predicting the same day Feature the matched regression model of data the input phase, obtain prediction the same day described in target charge station sailed in object time window Enter the predicted value of vehicle number.
Preferably, step 4 specifically executes following steps：
Step 4.1：Setting one sets one for the feature ordering collection F of empty set and according to the sample set of step 3 Former characteristic set S=1,2 ..., τ ... d }；
Wherein, 1≤τ≤d；
τ indicates τ in sample set features；
Step 4.2：Judge whether original characteristic set S is empty set, if empty set, then exports feature ordering collection F and executes step Rapid 4.6, if not empty set, then carry out step 4.3；
Step 4.3：All mesh in all features and the sample set in 4.2 Central Plains characteristic set S of extraction step Data are marked, and are fitted using linear regression model (LRM) to calculate the corresponding weight of each feature；
Step 4.4：The feature P of weight minimum is removed from former characteristic set S, and takes feature ordering collection F={ P } ∪ F, And repeat step 4.2；
Wherein, it removes more early feature P and concentrates sequence more rearward in feature ordering；
Step 4.6：The feature ordering for choosing output concentrates N number of feature of front.
Preferably, step 5 specifically executes following steps：
Step 5.1：Optimal solution will be solved in the data input object function of training set
Wherein, object function is as follows：
Wherein, (x_{i},y_{i}) ∈ m, (x_{j},y_{j}) ∈ m, 0≤α_{i},1≤i≤n, 1≤j≤n；
ε is expressed as error；C is penalty coefficient；R^{2n}The set of real numbers being expressed as under 2n dimension datas；α_{i} ^{*}, α_{j} ^{*}It is expressed as Ith of the syndrome vectors and jth of syndrome vectors that two syndrome vectors are concentrated；α_{i}, α_{j}It is expressed as first syndrome vectors collection In ith of syndrome vectors and jth of syndrome vectors；(x_{i},y_{i}) indicate the sample of ith day corresponding training set in training set m； (x_{j},y_{j}) indicate the sample of the corresponding training set in jth day in training set m；
k(x_{i}·x_{j}) it is expressed as related x_{i}And x_{j}Nonlinear mapping function；
Wherein,Φ is kernel function；
σ is expressed as width parameter；x_{i}, x_{j}Driving into ith day in N number of feature selected by expression step 4, in jth day respectively The set of vehicle number, outgoing vehicles number；y_{i}, y_{j}Ith day, the target data in jth day in training set are indicated respectively；
Optimal solutionAs follows：
Wherein,It is expressed as the nth of the optimal solution that first syndrome vectors collection and the second syndrome vectors are concentrated A syndrome vectors；
And
Wherein,Be expressed as first syndrome vectors concentration optimal solution ith of syndrome vectors and jth Syndrome vectors；Be expressed as second syndrome vectors concentration optimal solution ith of syndrome vectors and jth just Subvector；
Step 5.2：According to the optimal solution of step 5.1And parameter calculation formula calculates the parameter of regressive prediction model
Wherein, the parameter calculation formula is as follows：
Step 5.3：According to the calculated parameter of step 5.2Obtain regression model；
Wherein, regression model is：
Wherein, x be target charge station object time window in predicting the same day with N number of feature phase selected in step 4 The data of same feature.
Preferably, the method further includes：
Acquisition predicts the target charge station come and drives into target charge in vehicle number and ETC data in object time window It stands and actually drives into vehicle number in object time window；
The error of the regression model is calculated according to the data of acquisition；
Wherein, the error is excellent including at least average absolute percent error, mean absolute error, rootmeansquare error, fitting Spend a kind of data in coefficient of determination；
Judge whether calculated error meets preset requirement, be unsatisfactory for preset requirement, N values size is simultaneously in regulating step 4 Regression model is rebuild, until the error of regression model meets preset requirement.
Preferably, denoising is carried out to ETC data in step 1 and executes following steps：
It deletes only to drive into record in the ETC data and be not driven out to the data of record or delete and is only driven out to record and does not have There are the data for driving into record；
Data to lacking charge station location in the ETC data carry out looking into benefit.
Advantageous effect：
The present invention provides a kind of prediction technique of freeway toll station entrance vehicle, by acquiring freeway toll station Discrepancy vehicle number historical data, the use of algorithm of support vector machine is basic model, incorporates recursion elimination algorithm and delete and selects feature, The regression model of freeway toll station entrance vehicle number is obtained, and then realizes prediction freeway toll station entrance vehicle number. Wherein, regression model is obtained by a large amount of historical data combination supporting vector machine algorithm, greatly improves prediction result Accuracy, deleting choosing by using recursion elimination algorithm influences maximum feature, further improves the prediction result of regression model Accuracy.
In addition, the present invention makes its prediction result also by the error of calculating regression model come optimizing regulation regression model Accuracy higher.
Furthermore the present invention is to carry out analyzing processing based on ETC data, compared to the perception such as video, coil in traditional approach The cognitive method of device, the present invention is based on the accuracy highers of ETC data, and then improve the accuracy of prediction result of the present invention.
Description of the drawings
Fig. 1 is a kind of the schematic of the prediction technique of freeway toll station entrance vehicle number provided in an embodiment of the present invention Flow chart；
Fig. 2 is the distribution map of charge station provided in an embodiment of the present invention；
Fig. 3 is the comparison diagram of actual value and predicted value provided in an embodiment of the present invention, wherein (a) indicates that target charge station exists On January 29 15:0016:The actual value of 00 time window and the comparison diagram of predicted value (b) indicate target charge station January 30 19:0020:The actual value of 00 time window and the comparison diagram of predicted value.
Specific implementation mode
It is following that the present invention will be specifically described in conjunction with specific embodiments.
As shown in Figure 1, a kind of prediction technique of freeway toll station entrance vehicle number provided by the invention includes：
Step 1：It obtains the ETC data of charge station on highway and carries out denoising.
The distribution map of charge station as shown in Figure 2.Wherein, electronic charging system without parking (Electronic Toll Collection, ETC) data are referred to as ETC data, in the present embodiment, ETC data include vehicle drive into the time of charge station with And the position of charge station that the position for the charge station driven into, vehicle are driven out to the time of charge station and are driven out to.
Preferably, the position of charge station is the latitude and longitude information of charge station.
Wherein, ETC data progress denoising is specifically executed as follows：
It deletes only to drive into record the data for not being driven out to record or delete only to be driven out to record in ETC data and not sail Enter the data of record；
Data to lacking charge station location in ETC data carry out looking into benefit.
Wherein, the position of the charge station of location information, such as supplementary charge station are lacked according to the title of charge station supplement Latitude and longitude information.
For example, being explained with three highway in Guangdong province ETC data instances on April 1 to April 30th, 2016 It is bright, record will be only driven into first and is not driven out to record, or will only be driven out to record and do not driven into the ETC data of record and deleted It removes；And the freeway toll station point to lacking longitude and latitude, it is supplemented after being searched on Google Maps.It is original such as wide nitrogen station Lack latitude value in ETC data, after in Google Maps inquire supplement.
Step 2：ETC data in step 1 are grouped according to the preset duration of time window and count each charge It stands going out, entering vehicle number in daily each time window.
Specifically, the ETC data in step 1 are carried out according to the preset duration of time and time window in ETC data Grouping.
The preset duration of preferred time window is 1 hour in the present embodiment, and in other feasible embodiments, time window is preset Duration can be set according to the ETC data volumes of acquisition, such as the ETC data volumes of acquisition are bigger, corresponding time window it is pre If duration is bigger.
Specifically, selecting data to build sample from step 2 in the present embodiment, the period of selected research is following to be expressed as The sample training period, that is, model acquired sample time zone, the wherein sample training time using number of days as measurement unit, example Such as April 1 to 30 days between April 30.
It should be appreciated that the data for carrying out model foundation are necessarily taken from the number of a time interval in a territorial scope According to having corresponded to sample training period and a certain number of charge stations.
Step 3：According to step 2 count vehicle number structure target charge station object time window sample set.
It should be noted that each object time window at each target charge station corresponds to a sample set, i.e., same target The corresponding sample set of different target time window of charge station is different, the corresponding sample set of same object time window of different charge stations Also different.
Wherein, all charge stations K time window before object time window drives into vehicle number, outgoing vehicles in any one day Number and the collection of any one day target data are combined into a sample of sample set, K >=2.Such as when object time window is T When, the preceding K time window of object time window is respectively T1, T2 ..., TK.Preferred K=3, i.e. object time in the present embodiment Preceding 3 time windows of window.Target data is that vehicle number is driven into target charge station in object time window.
Sample set M includes n sample, and n is the number of days of sample training period, and specific sample set M is as follows：
M={ (X_{1},y_{1}),(X_{2},y_{2}),…,(X_{n},y_{n})},X_{n}∈R^{d},y_{n}∈R；
Wherein, (X_{n},y_{n}) indicate the sample of nth day corresponding sample set in sample set M, X_{n}Indicate all charges in nth day It stands in the set for driving into vehicle number, outgoing vehicles number of the preceding K time window of object time window；y_{n}Indicate nth day number of targets According to R is expressed as set of real numbers, R^{d}It is expressed as the set of real numbers of d dimensional features.
Wherein, any one charge station is arbitrary in the preceding K time window of object time window within the sample training period The collection for driving into vehicle number or outgoing vehicles number of one time window is combined into a feature of sample set.
Specifically, target charge station is d in the number of the feature of object time window, d is the positive integer more than 4.Wherein d's Calculation formula is as follows：
D=charge stations data × K × 2.
K should be appreciated that the data set that the data set that target charge station is constituted in d feature of object time window is n × d, That is the data set of n rows d row, wherein n are the number of days of sample training period, and n is positive integer.
Such as the data set Q that the corresponding data of d feature are constituted^{(n×d)}It is as follows：
It is shown：
Wherein X_{n}Indicate that nth day all charge station drive into vehicle number, are driven out in the preceding K time window of object time window The set of vehicle number；Expression kth of time window in nth day preceding K time window enters the vehicle number of eth of charge station；Expression kth of time window in ith day preceding K time window enters the vehicle number of eth of charge station；1≤k≤K；E is Positive integer more than or equal to 1 and less than or equal to studied charge station's summation.
Such as the sample training period is October, then n=31；And there are three highways in the region studied and one is shared 191 toll stations, the preceding K time window of the object time window at target charge station include the 1st before object time window, the 2nd, 3rd time window, then all data of the object time window at target charge station be 10 months in every day each charge station at first 3 The 1st in time window, the 2nd, the 3rd time window go out, enter vehicle number and target data, wherein be characterized as 191 charge stations In each charge station the 1st, the 2nd, the 3rd time window in preceding 3 time windows go out, enter vehicle number, calculate known to feature Quantity d=1146, therefore d feature of the object time window at target charge station constitute data set be 31 × 1146 matrix, It is i.e. as follows：
Wherein, { 1,2 ..., 31 } i ∈, e ∈ { 1,2 ..., 191 }, k ∈ { 1,2,3 }.
In addition, target data in sample set be in the sample training period target charge station in daily object time window Drive into vehicle number.The set of the target data of sample set is as follows：
A=(y_{1},y_{2},…,y_{n})^{T}, 1≤i≤n；
Wherein, y_{n}Indicate that vehicle number is driven at nth day in sample training period target charge station in object time window, i.e., Nth day target data.
Step 4：It is selected to the target charge station from d feature in the sample set using recursion elimination algorithm The traffic capacity influences maximum N number of feature in object time window；
Wherein, N is positive integer, and N is more than 1 and is less than d.
It should be appreciated that an object time window due to a target charge station corresponds to a sample set, a mesh One object time window of mark charge station has corresponded to one group of selected N number of feature out.
Specifically, step 4 specifically executes following steps：
Step 4.1：Setting one sets former feature set for the feature ordering collection F of empty set and according to the sample set of step 3 Conjunction S=1,2 ..., τ ... d }；
Wherein, 1≤τ≤d；τ indicates τ in sample set features, i.e. data set Q^{(n×d)}In τ row it is corresponding Data；
Step 4.2：Judge whether original characteristic set S is empty set, if empty set, then exports feature ordering collection F and executes step Rapid 4.6, if not empty set, then carry out step 4.3；
Step 4.3：All mesh in all features and the sample set in 4.2 Central Plains characteristic set S of extraction step Data are marked, and are fitted using linear regression model (LRM) to calculate the corresponding weight of each feature；
Wherein, the practical each coefficient ω for linear regression model (LRM) of the corresponding weight of each feature_{t}, t ∈ { 1,2 ..., d }, That is ω_{t}It is expressed as tth of coefficient of linear regression model (LRM).
Step 4.4：The feature P of weight minimum is removed from former characteristic set S, and takes feature ordering collection F={ P } ∪ F, And repeat step 4.2；
Wherein, it removes more early feature P and concentrates sequence more rearward in feature ordering；
Step 4.6：The feature ordering for choosing output concentrates N number of feature of front.
It should be appreciated that feature ordering concentrates top n to be characterized as to target charge station the traffic capacity shadow in object time window Ring maximum feature.
Step 5：According to the N number of feature construction training set m selected in step 4, and the data in the training set m are defeated Enter support vector regression establish target charge station object time window regression model.
Wherein, training set contains the N number of feature selected in step 5 and corresponds to N groups data and sample in sample set The target data of this concentration.
Specifically, any one day all drive into vehicle number, outgoing vehicles number and described in selected N number of feature The collection of one day target data of meaning is combined into the sample of a training set；
The training set includes the sample of n training set；
Training set m is embodied as follows：
M={ (x_{1},y_{1}),(x_{2},y_{2}),…,(x_{n},y_{n})},x_{n}∈R^{N},y_{n}∈R；
Wherein, (x_{n},y_{n}) indicate the sample of nth day corresponding training set in training set m, x_{n}Indicate the N selected by step 4 Nth day set for driving into vehicle number, outgoing vehicles number in a feature；y_{n}Indicate nth day target data, R^{N}It is expressed as Ndimensional spy The set of real numbers of sign.
Wherein, the data set q that the N number of feature gone out selected by training set m in the sample training period is constituted indicates as follows：
q^{(n×N)}=(x_{1},x_{2}…,x_{i},…,x_{n})^{T}, 1≤i≤n；
Wherein, q^{(n×N)}Indicate that data set q is the set of matrices of n rows N row；x_{i}It indicates the in N number of feature selected by step 4 The i days set for driving into vehicle number, outgoing vehicles number, i.e. q^{(n×N)}In the corresponding data of the ith row；N is the day of sample training period Number.
The number that drives into vehicle number of the target charge station in daily object time window in the sample training period in training set m Indicate as follows according to collection a：
A=(y_{1},y_{2},…,y_{n})^{T}, 1≤i≤n；
Wherein, y_{i}The vehicle number that target charge station is driven into object time window when indicating in the sample training period ith day, I.e. ith day target data.
Specifically, the data of training set are inputted support vector regression to establish target charge station in target in step 5 The regression model of time window, it is specific to execute following operation：
Step 5.1：Optimal solution will be solved in the data input object function of training set
Object function is as follows：
Wherein, (x_{i},y_{i}) ∈ m, (x_{j},y_{j}) ∈ m, 0≤α_{i},1≤i≤n, 1≤j≤n；
ε is expressed as error；C is penalty coefficient；R^{2n}The set of real numbers being expressed as under 2n dimension datas；α_{i} ^{*}, α_{j} ^{*}It is expressed as Ith of the syndrome vectors and jth of syndrome vectors that two syndrome vectors are concentrated；α_{i}, α_{j}It is expressed as first syndrome vectors collection In ith of syndrome vectors and jth of syndrome vectors；(x_{i},y_{i}) indicate the sample of ith day corresponding training set in training set m； (x_{j},y_{j}) indicate the sample of the corresponding training set in jth day in training set m；
k(x_{i}·x_{j}) it is expressed as related x_{i}And x_{j}Nonlinear mapping function；
Wherein,Φ is kernel function；
σ is expressed as width parameter；x_{i}, x_{j}Indicate respectively N number of feature in training set selected by step 4 correspond to ith day, The set for driving into vehicle number, outgoing vehicles number in jth day；y_{i}, y_{j}Ith day, the number of targets in jth day in training set are indicated respectively According to；
Optimal solutionAs follows：
Wherein,It is expressed as the nth of the optimal solution that first syndrome vectors collection and the second syndrome vectors are concentrated A syndrome vectors；
And
Wherein,Be expressed as first syndrome vectors concentration optimal solution ith of syndrome vectors and jth Syndrome vectors；Be expressed as second syndrome vectors concentration optimal solution ith of syndrome vectors and jth just Subvector；
Step 5.2：According to the optimal solution of step 5.1And parameter calculation formula calculates the parameter of regressive prediction model
Wherein, the parameter calculation formula is as follows：
Step 5.3：According to the calculated parameter of step 5.2Obtain regression model；
Wherein, regression model is：
Wherein, k (x_{i}X) it is expressed as related x_{i}And x_{j}Nonlinear mapping function；k(x_{i}X)=Φ (x_{i})·Φ(x)；x For the data of object time window feature identical with N number of feature selected in step 4 in predicting the same day at target charge station. X in regression model_{i}N number of feature correspondence to go out selected by step 4 in training set m drives into vehicle number in ith day, is driven out to vehicle The set of number.
For example, the sample training period is 31 days of October, and after N number of feature is selected in step 5, when being predicted, no matter It is when the driving into vehicle of object time window for predicting some day in November or target charge some day in December station, x_{i}Training is corresponded to N number of feature in October in 31 days selected in step 4 is concentrated to correspond to the data at ith day.
Step 6：By the object time window at target charge station in predicting the same day spy identical with N number of feature in step 4 The matched regression model of data the input phase of sign obtains target charge station in the prediction same day and drives into vehicle number in object time window Predicted value.
Preferably, the above method further includes：
Acquisition predicts the target charge station come and drives into target charge in vehicle number and ETC data in object time window It stands and actually drives into vehicle number in object time window；
The error of regression model is calculated according to the data of acquisition；
Wherein, error is sentenced including at least average absolute percent error, mean absolute error, rootmeansquare error, the goodness of fit Determine a kind of data in coefficient；
Judge whether calculated error meets preset requirement, be unsatisfactory for preset requirement, N values size is simultaneously in regulating step 5 Regression model is rebuild, until the error of regression model meets preset requirement.
As shown in figure 3, abscissa T is that vehicle is actually driven into target charge station in object time window in figure (a) and (b) Number, ordinate S is that vehicle number is driven into the target charge station gone out using forecast of regression model in object time window, wherein figure (a) Data correspond to target charge station 29 days 15 January:0016:The data of the data of 00 time window, figure (b) correspond to target Charge station was 30 days 19 January:0020:The data of 00 time window are fitted to obtain figure (a) to the data of figure (a) and figure (b) The fitting function and its goodness of fit coefficient of determination of corresponding S=0.94T+0.16 is 0.990；Scheme (b) and corresponds to S=0.97T+0.18 Fitting function and its goodness of fit coefficient of determination be 0.986, according to fitting coefficient of determination identify regression model quality, in turn Whether model is optimized.
Error as shown in table 1 below to be calculated：
Table 1
It should be appreciated that the foregoing is merely a prefered embodiment of the invention, it is merely illustrative for the purpose of the present invention, rather than limit Property processed.Those skilled in the art understand that many modifications can be carried out to it in the scope of the claims in the present invention, but It falls in protection scope of the present invention.Based on the embodiments of the present invention, those of ordinary skill in the art are not making The every other embodiment obtained under the premise of creative work, belongs to protection scope of the present invention.
Claims (5)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN201710613932.7A CN107331164B (en)  20170725  20170725  A kind of prediction technique of freeway toll station entrance vehicle number 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN201710613932.7A CN107331164B (en)  20170725  20170725  A kind of prediction technique of freeway toll station entrance vehicle number 
Publications (2)
Publication Number  Publication Date 

CN107331164A CN107331164A (en)  20171107 
CN107331164B true CN107331164B (en)  20180828 
Family
ID=60227543
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN201710613932.7A CN107331164B (en)  20170725  20170725  A kind of prediction technique of freeway toll station entrance vehicle number 
Country Status (1)
Country  Link 

CN (1)  CN107331164B (en) 
Citations (2)
Publication number  Priority date  Publication date  Assignee  Title 

CN104240520A (en) *  20140924  20141224  四川省交通科学研究所  GSO (glowworm swarm optimization) based highway traffic prediction method 
CN104269055A (en) *  20140924  20150107  四川省交通科学研究所  Expressway traffic flow forecasting method based on time series 
Family Cites Families (1)
Publication number  Priority date  Publication date  Assignee  Title 

EP3114574A4 (en) *  20140303  20180307  Inrix, Inc.  Traffic obstruction detection 

2017
 20170725 CN CN201710613932.7A patent/CN107331164B/en active IP Right Grant
Patent Citations (2)
Publication number  Priority date  Publication date  Assignee  Title 

CN104240520A (en) *  20140924  20141224  四川省交通科学研究所  GSO (glowworm swarm optimization) based highway traffic prediction method 
CN104269055A (en) *  20140924  20150107  四川省交通科学研究所  Expressway traffic flow forecasting method based on time series 
NonPatent Citations (1)
Title 

高速公路进出流量差异的BP神经网络预测模型;徐明 等;《交通与计算机》;20050630;第23卷(第3期);第3336页 * 
Also Published As
Publication number  Publication date 

CN107331164A (en)  20171107 
Similar Documents
Publication  Publication Date  Title 

Bhat  A heteroscedastic extreme value model of intercity travel mode choice  
Haghani et al.  Data collection of freeway travel time ground truth with bluetooth sensors  
CN101751777B (en)  Dynamic urban road network traffic zone partitioning method based on space cluster analysis  
US9171485B2 (en)  Geodatabase information processing  
CN102637357A (en)  Regional traffic state assessment method  
CN102097002A (en)  Method and system for acquiring bus stop OD based on IC card data  
CN100452109C (en)  Method for obtaining everage speed of city road section traffic flow  
He et al.  A method for identification of driving patterns in hybrid electric vehicles based on a LVQ neural network  
CN102013163A (en)  Method for bus origindestination (OD) investigation by using mobile phone base station data and operating vehicle global position system (GPS) data  
CN104574967B (en)  A kind of city based on Big Dipper large area road grid traffic cognitive method  
Long et al.  Evaluating the effectiveness of urban growth boundaries using human mobility and activity records  
CN105185115B (en)  A kind of traffic forecast method and forecasting system  
CN102157075A (en)  Method for predicting bus arrivals  
Obaidat et al.  Integration of geographic information systems and computer vision systems for pavement distress classification  
ES2613243T3 (en)  System and procedure to prepare a driver profile  
US20160335894A1 (en)  Bus Station Optimization Evaluation Method and System  
CN104064028B (en)  Based on public transport arrival time Forecasting Methodology and the system of multiple information data  
CN106845371B (en)  A kind of city road network automotive emission remote sensing monitoring system  
CN104183119B (en)  Based on the antiarithmetic for realtime traffic flow distribution forecasting method pushed away of section OD  
CN104575050B (en)  A kind of fast road ramp intellectual inducing method and device based on Floating Car  
CN103473620B (en)  Comprehensive passenger transport hub many modes of transportation Forecasting Methodology and system  
CN104881472B (en)  A kind of travelling route sight spot combined recommendation method based on network data collection  
CN106096623A (en)  A kind of crime identifies and Forecasting Methodology  
CN103984994B (en)  Method for predicting urban rail transit passenger flow peak duration  
CN103778784B (en)  The method of express highway section traffic state information is obtained in real time based on data in mobile phone 
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 