CN109741604A - Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow - Google Patents
Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow Download PDFInfo
- Publication number
- CN109741604A CN109741604A CN201910162274.3A CN201910162274A CN109741604A CN 109741604 A CN109741604 A CN 109741604A CN 201910162274 A CN201910162274 A CN 201910162274A CN 109741604 A CN109741604 A CN 109741604A
- Authority
- CN
- China
- Prior art keywords
- shot
- long term
- tranquilization
- term memory
- memory network
- 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
- 230000007787 long-term memory Effects 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000004907 flux Effects 0.000 claims abstract description 10
- 238000012360 testing method Methods 0.000 claims abstract description 10
- 238000013135 deep learning Methods 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 abstract description 4
- 238000009499 grossing Methods 0.000 abstract 1
- 238000005311 autocorrelation function Methods 0.000 description 10
- 210000004027 cell Anatomy 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 4
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 4
- 230000000644 propagated effect Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000007774 longterm Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 230000015654 memory Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses one kind to be based on tranquilization shot and long term memory network model prediction intersection traffic method of flow, comprising: is successively constructed based on deep learning network using Tensorflow frame and forms tranquilization shot and long term memory network；Input time sequence data pre-processes time series data to tranquilization shot and long term memory network；By pretreated time series data smoothing input shot and long term memory network, the weight and deviation of input time sequence data are calculated, and saves weight and deviation；It based on weight and is biased to calculate the training result for obtaining tranquilization shot and long term memory network, and sets a judgment threshold, whether condition is met based on judgment threshold training of judgement result；The intersection traffic flux prediction model based on tranquilization shot and long term memory network is constructed if judging result meets condition, input test sample to intersection model of traffic flux forecast exports prediction result；The present invention is small to the prediction error of intersection traffic stream, effectively improves predictablity rate.
Description
Technical field
The invention belongs to technical field of transportation, are mainly used in the prediction technique of intersection traffic flow, and in particular to one
Kind is based on tranquilization shot and long term memory network model prediction intersection traffic method of flow.
Background technique
Intersection is traffic organization form most commonly seen in urban road, is had extremely to traffic administration and control
Important meaning.Intersection traffic condition is complicated and changeable, but also shows apparent time, spatial character.How to combine
The space-time characterisation of traffic flow is analyzed intersection traffic situation by existing wagon detector, to ensure intersection
Traffic safety, give full play to the traffic capacity of intersection, be each traffic people concern the most.Groundwork embodies
The magnitude of traffic flow of future time period is predicted by intersection historical traffic flows data.
It in the prior art, is to optimize and revise model and improvement for the main method for improving short-term traffic flow forecast precision
Model, such as: Shen Xiajiong adjusts the weight of weak learner constantly using based on gradient promotion regression tree model to correct model
Residual error, improve the accuracy of prediction model well；Luo Xianglong etc. combines depth conviction network model and support vector to return
Return classifier as prediction model, carry out traffic flow character study with deepness belief network model, connects and support in network top
Vector regression model carries out volume forecasting etc..But in a practical situation, specific Vehicular behavior also can be to traffic flow
The precision of prediction of amount has a huge impact, therefore only from the prediction for adjusting and improving to model to realize the magnitude of traffic flow, can not
It realizes and the essence of traffic flow forecasting precision is promoted.
Summary of the invention
Prediction essence only can not be promoted from real situation from the improvement of model come predicting traffic flow amount for the above-mentioned prior art
The problem of spending, the present invention are a kind of based on tranquilization shot and long term memory network model prediction intersection traffic method of flow in proposing；
The specific technical solution of this method is as follows:
One kind being based on tranquilization shot and long term memory network model prediction intersection traffic method of flow, which comprises
S1, based on deep learning network using Tensorflow frame successively construct formed tranquilization shot and long term remember net
Network；
S2, input time sequence data to the tranquilization shot and long term memory network, and the time series data is done
Pretreatment；
S3, the pretreated time series data inputs to the tranquilization shot and long term memory network, described in calculating
The weight and deviation of input time sequence data, and save the weight and deviation；
S4, the training result for obtaining the tranquilization shot and long term memory network is calculated based on the weight and deviation, and set
A fixed judgment threshold, judges whether the training result meets condition based on the judgment threshold；
S5, the intersection traffic stream based on tranquilization shot and long term memory network is constructed if the judging result meets condition
Prediction model is measured, input test sample to the intersection traffic flux prediction model exports prediction result.
Further, the time series data is that the Continuous Traffic flow at specified intersection specified time interval is constituted
Training sample set.
Further, the test sample collection is the magnitude of traffic flow for specified intersection specified time being spaced in the setting date.
It is of the invention based on tranquilization shot and long term memory network model prediction intersection traffic method of flow, this method is in depth
It spends under the frame of learning network, deep learning network struction tranquilization shot and long term memory network is based on, according to tranquilization shot and long term
Memory network constructs intersection traffic flux prediction model, is realized by intersection traffic flux prediction model to the magnitude of traffic flow
Prediction；Compared with prior art, method of the invention can reduce the error of vehicle flowrate prediction, promote precision of prediction.
Detailed description of the invention
Fig. 1 is that tranquilization shot and long term memory network model prediction intersection traffic flow is based on described in the embodiment of the present invention
The flow chart of method is illustrated；
Fig. 2 is the schematic diagram signal of shot and long term memory network model described in the embodiment of the present invention；
Fig. 3 is the process predicted using shot and long term memory network model intersection traffic stream in the embodiment of the present invention
Diagram meaning；
Fig. 4 is to be predicted using tranquilization shot and long term memory network model intersection traffic stream in the embodiment of the present invention
Flow chart signal；
Fig. 5 is to be illustrated using the statistical chart for the autocorrelation for recording original traffic time series during the method for the present invention；
Fig. 6 is using the method for the present invention by the auto-correlation of original traffic time series traffic sequence after disposable difference
Property statistical chart signal；
Fig. 7 is tranquilization shot and long term memory network model described in the embodiment of the present invention with the increased penalty values of frequency of training
Trend chart signal；
Fig. 8 carries out the comparison knot of intersection traffic volume forecasting and actual traffic flow using shot and long term memory network model
Fruit diagram meaning；
Fig. 9 carries out intersection traffic volume forecasting and actual traffic flow using tranquilization shot and long term memory network model
Comparing result diagram meaning.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described.
Embodiment one
Refering to fig. 1, in embodiments of the present invention, it provides a kind of based on tranquilization shot and long term memory network model prediction friendship
Prong magnitude of traffic flow method, method specifically include that steps are as follows:
S1, based on deep learning network using Tensorflow frame successively construct formed tranquilization shot and long term remember net
Network；S2, input time sequence data to tranquilization shot and long term memory network, and time series data is pre-processed；Specifically,
Pretreatment carries out season difference to original time series according to the time cycle, so that non-stationary present in time series
Sequence stationary；Wherein, time series data is the training of the Continuous Traffic flow composition at specified intersection specified time interval
Sample set；S3, pretreated time series data is inputted into the tranquilization shot and long term memory network, calculates input time sequence
The weight and deviation of column data, and save weight and deviation；S4, acquisition tranquilization shot and long term memory is calculated based on weight and deviation
The training result of network, and a judgment threshold is set, judge whether the training result meets condition based on judgment threshold；S5,
The intersection traffic flux prediction model based on tranquilization shot and long term memory network, input are constructed if judging result meets condition
Test sample exports prediction result to intersection model of traffic flux forecast；Wherein, test sample collection is that specified intersection is specified
The magnitude of traffic flow of the time interval on the setting date.
The one kind of RNN (RecurrentNeural Networks, Recognition with Recurrent Neural Network) as neural network, and can be by one
The output of a little hidden neurons retains in a network, and with the input collective effect of Recognition with Recurrent Neural Network next stage when current
Between neuron, therefore RNN has preferable effect for the prediction of time series, and still, RNN network can not solve gradient disappearance
The problem of with gradient explosion；Therefore, the present invention is improved based on RNN, forms LSTM (Long-Short Term
Memory, shot and long term remember artificial neural network), to enhance shot and long term memory function, information is allowed no longer to decay, to reach gram
Take the purpose of gradient disappearance problem；Fig. 1 specifically is seen, as the improvement to RNN, the main distinction is that in former algorithm LSTM
On the basis of joined a cell, cell can be used for judging whether information useful；The place of its most critical is the shape of cell
State, the transmission of cell state are passed through from entire cell just as a conveyer belt, vector, have only done a small amount of linear operation；
One cell is made of three threshold structures and a state vector transmission line, and thresholding is to forget door, be passed to door, be defeated respectively
It gos out；Wherein state vector transmission line is responsible for long-range memory, and three thresholdings are responsible for the selection of short-term memory；Wherein, input gate is determined
Determine how information is sent in the memory module of hidden layer；Door is forgotten for judging whether the information being conveyed into conforms to
It asks, satisfactory information is left, and undesirable information passes into silence；Out gate then determines that information passes in which way
It passs away.
Wherein, the shot and long term memory network of formation includes propagated forward and back-propagating two ways:
(1) propagated forward
The first step is the update to door is forgotten during the propagated forward of shot and long term memory network, this step passes through
What kind of information the decision of Sigmoid nervous layer allows pass through this cell.By reading input value ht-1And xt, export a numerical value
Vector between 0,1 indicates the specific gravity for allowing each section information to pass through.Wherein, 1 " all information all pass through " is indicated, 0 table
Show " all information are all given up ", can specifically be indicated by following formula: ft=σ (Wf·[ht-1,xt]+bf), wherein WfAnd bf
The weight and deviation for forgeing door are respectively indicated, σ indicates the Sigmoid function for being used as activation primitive.
Second step is the update to the state value of cell, that is, determines have which information needs to be updated, and with tanh layers
A chosen candidate value is generated to replace oldState value, mainly passes through following formula it=(Wi·[ht-1,xt]+bi) and formulaIt realizes, in formula, WiAnd biIt is the weight and deviation of input gate, W respectivelyCAnd bCIt is tanh respectively
The weight and deviation of the new candidate value of layer creation.
And by state Ct-1It is updated to state Ct.I.e. Ct-1The specific gravity f passed through with informationtIt is multiplied, gives up determining house by specific gravity
The information of abandoning, while plus input itWith candidate valueMultiply, can specifically be indicated by the following formula of formula:
Third step is according to state CtThe value for determining output, by formula ot=σ (Wo·[ht-1,xt]+bo) and formula ht=
ot*tanh(Ct)；Wherein, WoAnd boIt is the weight and deviation of out gate respectively, first passes through Sigmoid layers and determine which portion exported
Point information, then will carry out handling by tanh layer of state vector and by the multiplied by weight of itself and Sigmoid layers of output, obtain
Output result to the end.
(2) back-propagating
Back-propagating is also known as retrospectively calculate, is the calculating for the error amount of each neuron.It is opened from current time t
Begin, calculate the error term at each moment and propagates resulting error term upper layer；It is calculated according to corresponding error term every
The gradient of a weight passes through gradient descent method iteration undated parameter.
Wherein to gradient rewrite process specifically:
Pass through formula I (t)=f (h firstt,yt)=| | ht-yt||2Loss function is defined, wherein htAnd ytPoint
Not Wei output sequence and sample label, the minimum in entire time series is done to I (t), formula can be obtainedFormula
In, T represents entire time series.
Then formula can be obtained by chain methodTo calculate gradient, wherein hitIt is i-th of unit
Output, M is the number of unit in shot and long term memory artificial neural network, network t propagated forward at any time, hitChange not shadow
Ring the loss value before t moment, then it is availableGradient can must be rewritten with this are as follows:
In conjunction with Fig. 3 and Fig. 4, diagram is respectively to be based on shot and long term memory network model and tranquilization shot and long term memory network mould
Type carries out the flow chart of traffic flow data analysis and prediction because time series data that traffic flow data is constituted be one with
The time series of machine, therefore often show the feature of non-stationary, that is, it will appear apparent smearing and ACF value obviously surpass
95% confidence interval out；The present invention is for original traffic data observation sequence, it is assumed that it is { Xt, t=1,2,3 ..., n },
ACF estimated value then can get using sample autocorrelation function method, speciallyAccording to
The ACF value that is calculated confirms the period；And it is based on original traffic stream sequence calculating cycle differenceTo be put down
Steadyization difference sequenceThe ACF value that its sample autocorrelation function is calculated according to differentiated sequence, works as ACF
When value does not occur obvious smearing and is in 95% confidence interval, it can determine that difference sequence belongs to tranquilization sequence, it is no
Then, difference sequence is non-stationary series；Finally, sample set is obtained using tranquilization difference sequence as the output field of sample, and
Based on the sample set, using the model prediction of shot and long term memory network as a result, prediction result is reduced to original traffic flow data.
Embodiment two
It will verify that the present invention is based on tranquilization shot and long term memory network model prediction intersections by practical operation below
The validity of magnitude of traffic flow method, detailed process are as follows:
The experimental data used is 21 days to 2018 September 20 of August in 2018 provided by traffic-police detachment, Nantong City
The day clock show Road-collected truthful data of north Hao bridge intersection forward direction radar vehicle detector.Road vehicle detector pair
The each vehicle by the intersection carries out information collection, and the main field that acquires is to be compiled by time, crossing number, section
Number, lane number, type of vehicle and average speed, time of origin, entrance driveway exit ramp etc., and data deposit database is protected
It deposits, it is specific as shown in table 1.
Table 1 crosses car data primary fields
Basic statistical analysis is carried out to record by database, according to lane number by the data of different directions out of phase
It separates, extracted in conjunction with section number and predicts related field.It is united using the time interval of 15min to traffic flow data
Meter, by statistical result with time interval start time, average speed, vehicle be averaged the holding time, vehicle number etc. fields shape
Formula is presented, and static fields are referring to table 2.
2 magnitude of traffic flow static fields of table (15min time interval)
It is found after statistics, there are a small amount of missing datas in data set；In order to reduce influence of the missing data to prediction, in conjunction with
Two adjacent data of missing data, fill up missing data with the means averaged, so that sample data be made to have continuity.
Simultaneously as the quantity of acquisition data is more, it is commonly present invalid data, abnormal data.Such data are for the later period
Analysis result have large effect, it is therefore desirable to according to the restriction speed of urban road passing vehicle, by average row in data
The data record that speed is more than 80km/h is sailed to be rejected as abnormal data.
The statistical data for being 15min according to time interval, by weather, time interval, average speed, average holding time with
And the t-2 days, the t-1 days, the magnitude of traffic flow at t days certain moment is as input field, the magnitude of traffic flow of the t+1 days synchronizations
Data set is constructed as output field, shown in table 3 specific as follows.
Wherein, it is used as training sample within August 24 days to September data on the 19th total 2592, September data on the 20th is amounted to
96 are used as test sample.It is trained using the magnitude of traffic flow of the tranquilization shot and long term memory models to east exit road straight trip direction
And prediction.
3 prediction algorithm of table uses field
Firstly, to original traffic stream observation sequence { Xt, t=1,2,3 ..., n }, it is obtained using sample autocorrelation function method
ACF estimated value confirms that the period is 96 according to ACF, specifically sees Fig. 5.
As shown, the sequence has apparent smearing, i.e., within longer sample lag time sample from
Related coefficient estimated value is more than confidence interval, it was demonstrated that the sequence is non-stationary series.
Based on original traffic stream sequence, its periodical difference is calculatedObtain tranquilization difference sequenceAnd verify its auto-correlation coefficient estimated value.
Refering to Fig. 6, after first difference, the smearing of sample be improved significantly, traffic sequence is approximated as putting down
Steadyization sequence.Then sample set is obtained using tranquilization difference sequence as the output of sample.
According to sample set data, using tranquilization shot and long term memory models prediction result, and result is reduced to original friendship
Through-current capacity data.
Wherein, the training step of tranquilization shot and long term memory models is as follows:
(1) model parameter is chosen
Initial learning rate is 0.006, and hidden layer unit number is 10, and the Batch_size of training dataset is 80, iteration time
Number is 2000 times, time_steps 10.The time_step of test set is 12.
As shown in Figure 7, after iteration is more than 1,000 times, in zero, which works well loss late tendency of changes.
(2) model training
Training set of 29 days before the middle of the month data as model is chosen, the data of last day are instructed for testing model
Practice result.
Two pictures are by shot and long term memory network prediction model and tranquilization shot and long term memory network respectively below
Predict obtained result.
By the prediction model of training set training section, refering to Fig. 8, it is prediction model that wherein coordinate points, which are the line of grid,
Predicting traffic flow amount, coordinate points are that the line of triangle is the actual measurement magnitude of traffic flow；There it can be seen that prediction gained traffic flow magnitude with
The trend of traffic flow magnitude in different time period obtained by actual measurement is identical, and gap is little.
Refering to Fig. 9, by the difference of predicting traffic flow magnitude and actually measured traffic flow magnitude that model prediction obtains, together
When prediction result is reduced to traffic flow data, wherein coordinate points are that the broken line of grid represents predicted value, and coordinate points are triangle
Broken line represents actual value；It can be found that the obtained predicted value of prediction model based on the building of tranquilization shot and long term memory network with
The actual value goodness of fit is higher.
For the more intuitive degree of fitting for showing algorithm, pass through R squares of (R-Square) Lai Hengliang predicted value and reality
The degree of fitting of value, R squares passes through formulaIt calculates；Meanwhile in order to intuitively reflect herein using prediction technique
Performance, introduce root-mean-square error (Root Mean Square Error, RMSE) and average absolute percentage error (Mean
Absolute Percentage Error, MAPE) measure the accuracy of prediction.
Root-mean-square error (RMSE) refers to square of the quadratic sum observation frequency n ratio of observation and true value deviation
Root, can be by formulaIt calculates and obtains；Mean absolute percentage error (MAPE) refers to predicted value and true
The difference of real value accounts for the arithmetic average of actual measurement percentage, specifically can be by formulaCalculating obtains
?；Prediction result is analyzed by poor index, the prediction error result of the model can be obtained.Prediction result and early period are used
Linear regression, Support vector regression model compare, concrete outcome is as shown in table 4 below.
4 prediction result comparative analysis of table
By comparing it is not difficult to find that shot and long term memory network model is in RMSE and R2On be greatly improved, point
Do not reached 17.08 and 92.69%, but its MAPE value compared to other methods still without very big raising.But tranquilization
The prediction result advantage of shot and long term memory network model becomes apparent, and RMSE value has reached 11.94, MAPE value and has also been down to
12.18%, degree of fitting is up to 95.78%.
It is in summary, of the invention based on tranquilization shot and long term memory network model prediction intersection traffic method of flow,
This method is based on deep learning network struction tranquilization shot and long term memory network under the frame of deep learning network, according to flat
Steadyization shot and long term memory network constructs intersection traffic flux prediction model, passes through the realization pair of intersection traffic flux prediction model
The prediction of the magnitude of traffic flow；Compared with prior art, method of the invention can reduce the error of vehicle flowrate prediction, promote prediction essence
Degree.
The foregoing is merely a prefered embodiment of the invention, is not intended to limit the scope of the patents of the invention, although referring to aforementioned reality
Applying example, invention is explained in detail, still can be to aforementioned each tool for coming for those skilled in the art
Technical solution documented by body embodiment is modified, or carries out equivalence replacement to part of technical characteristic.All benefits
The equivalent structure made of description of the invention and accompanying drawing content is directly or indirectly used in other related technical areas,
Similarly within the invention patent protection scope.
Claims (3)
1. being based on tranquilization shot and long term memory network model prediction intersection traffic method of flow, which is characterized in that the method
Include:
S1, it successively constructs using Tensorflow frame based on deep learning network and forms tranquilization shot and long term memory network；
S2, input time sequence data to the tranquilization shot and long term memory network, and pre- place is done to the time series data
Reason；
S3, the pretreated time series data is inputted into the tranquilization shot and long term memory network, calculates the input
The weight and deviation of time series data, and save the weight and deviation；
S4, the training result for obtaining the tranquilization shot and long term memory network is calculated based on the weight and deviation, and set one
Judgment threshold judges whether the training result meets condition based on the judgment threshold；
S5, intersection traffic flow of the building based on tranquilization shot and long term memory network is pre- if the judging result meets condition
Model is surveyed, input test sample to the intersection traffic flux prediction model exports prediction result.
2. it is based on tranquilization shot and long term memory network model prediction intersection traffic method of flow as described in claim 1,
It is characterized in that, the time series data is the training sample that the Continuous Traffic flow at specified intersection specified time interval is constituted
Collection.
3. it is based on tranquilization shot and long term memory network model prediction intersection traffic method of flow as described in claim 1,
It is characterized in that, the test sample collection is the magnitude of traffic flow for specified intersection specified time being spaced in the setting date.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910162274.3A CN109741604A (en) | 2019-03-05 | 2019-03-05 | Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910162274.3A CN109741604A (en) | 2019-03-05 | 2019-03-05 | Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109741604A true CN109741604A (en) | 2019-05-10 |
Family
ID=66369330
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910162274.3A Pending CN109741604A (en) | 2019-03-05 | 2019-03-05 | Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109741604A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335466A (en) * | 2019-07-11 | 2019-10-15 | 青岛海信网络科技股份有限公司 | Traffic flow forecasting method and equipment |
| CN110517484A (en) * | 2019-08-06 | 2019-11-29 | 南通大学 | Diamond interchange area planar crossing sign occlusion prediction model building method |
| CN111723584A (en) * | 2020-06-24 | 2020-09-29 | 天津大学 | Punctuation prediction method based on consideration of domain information |
| CN111833583A (en) * | 2020-07-14 | 2020-10-27 | 南方电网科学研究院有限责任公司 | Training method, device, equipment and medium for power data anomaly detection model |
| CN111970163A (en) * | 2020-06-30 | 2020-11-20 | 网络通信与安全紫金山实验室 | Network flow prediction method of LSTM model based on attention mechanism |
| CN112001482A (en) * | 2020-08-14 | 2020-11-27 | 佳都新太科技股份有限公司 | Vibration prediction and model training method and device, computer equipment and storage medium |
| CN117292551A (en) * | 2023-11-27 | 2023-12-26 | 辽宁邮电规划设计院有限公司 | Urban traffic situation adjustment system and method based on Internet of Things |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0755039A2 (en) * | 1995-07-07 | 1997-01-22 | MANNESMANN Aktiengesellschaft | Method and system for the prognosis of traffic flow |
| CN105788249A (en) * | 2014-12-16 | 2016-07-20 | 高德软件有限公司 | Traffic flow prediction method, prediction model generation method and device |
| KR101742042B1 (en) * | 2016-11-15 | 2017-05-31 | 한국과학기술정보연구원 | Apparatus and method for traffic flow prediction |
| CN107154150A (en) * | 2017-07-25 | 2017-09-12 | 北京航空航天大学 | A kind of traffic flow forecasting method clustered based on road with double-layer double-direction LSTM |
| CN107230351A (en) * | 2017-07-18 | 2017-10-03 | 福州大学 | A kind of Short-time Traffic Flow Forecasting Methods based on deep learning |
| CN108510741A (en) * | 2018-05-24 | 2018-09-07 | 浙江工业大学 | A kind of traffic flow forecasting method based on Conv1D-LSTM neural network structures |
-
2019
- 2019-03-05 CN CN201910162274.3A patent/CN109741604A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0755039A2 (en) * | 1995-07-07 | 1997-01-22 | MANNESMANN Aktiengesellschaft | Method and system for the prognosis of traffic flow |
| CN105788249A (en) * | 2014-12-16 | 2016-07-20 | 高德软件有限公司 | Traffic flow prediction method, prediction model generation method and device |
| CN105788249B (en) * | 2014-12-16 | 2018-09-28 | 高德软件有限公司 | A kind of traffic flow forecasting method, prediction model generation method and device |
| KR101742042B1 (en) * | 2016-11-15 | 2017-05-31 | 한국과학기술정보연구원 | Apparatus and method for traffic flow prediction |
| CN107230351A (en) * | 2017-07-18 | 2017-10-03 | 福州大学 | A kind of Short-time Traffic Flow Forecasting Methods based on deep learning |
| CN107154150A (en) * | 2017-07-25 | 2017-09-12 | 北京航空航天大学 | A kind of traffic flow forecasting method clustered based on road with double-layer double-direction LSTM |
| CN108510741A (en) * | 2018-05-24 | 2018-09-07 | 浙江工业大学 | A kind of traffic flow forecasting method based on Conv1D-LSTM neural network structures |
Non-Patent Citations (4)
| Title |
|---|
| 刘明宇 等: "基于深度学习的交通流量预测", 《系统仿真学报》 * |
| 国家税务总局收入规划核算司: "《税收统计学》", 30 September 2014 * |
| 曹博 等: "基于LSTM 的短时交通流预测研究", 《研究与开发》 * |
| 王祥雪 等: "基于深度学习的短时交通流预测研究", 《交通运输系统工程与信息》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335466A (en) * | 2019-07-11 | 2019-10-15 | 青岛海信网络科技股份有限公司 | Traffic flow forecasting method and equipment |
| CN110517484A (en) * | 2019-08-06 | 2019-11-29 | 南通大学 | Diamond interchange area planar crossing sign occlusion prediction model building method |
| CN111723584A (en) * | 2020-06-24 | 2020-09-29 | 天津大学 | Punctuation prediction method based on consideration of domain information |
| CN111723584B (en) * | 2020-06-24 | 2024-05-07 | 天津大学 | Punctuation prediction method based on consideration field information |
| CN111970163A (en) * | 2020-06-30 | 2020-11-20 | 网络通信与安全紫金山实验室 | Network flow prediction method of LSTM model based on attention mechanism |
| CN111970163B (en) * | 2020-06-30 | 2022-06-21 | 网络通信与安全紫金山实验室 | Network flow prediction method of LSTM model based on attention mechanism |
| CN111833583A (en) * | 2020-07-14 | 2020-10-27 | 南方电网科学研究院有限责任公司 | Training method, device, equipment and medium for power data anomaly detection model |
| CN111833583B (en) * | 2020-07-14 | 2021-09-03 | 南方电网科学研究院有限责任公司 | Training method, device, equipment and medium for power data anomaly detection model |
| CN112001482A (en) * | 2020-08-14 | 2020-11-27 | 佳都新太科技股份有限公司 | Vibration prediction and model training method and device, computer equipment and storage medium |
| CN112001482B (en) * | 2020-08-14 | 2024-05-24 | 佳都科技集团股份有限公司 | Vibration prediction and model training method, device, computer equipment and storage medium |
| CN117292551A (en) * | 2023-11-27 | 2023-12-26 | 辽宁邮电规划设计院有限公司 | Urban traffic situation adjustment system and method based on Internet of Things |
| CN117292551B (en) * | 2023-11-27 | 2024-02-23 | 辽宁邮电规划设计院有限公司 | Urban traffic situation adjustment system and method based on Internet of things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109741604A (en) | Based on tranquilization shot and long term memory network model prediction intersection traffic method of flow | |
| CN106781489B (en) | A kind of road network trend prediction method based on recurrent neural network | |
| CN110164128A (en) | A kind of City-level intelligent transportation analogue system | |
| CN114049765B (en) | Urban road network traffic flow OD estimation method based on automatic vehicle number plate identification data | |
| CN111160622A (en) | Scenic spot passenger flow prediction method and device based on hybrid neural network model | |
| CN106571030A (en) | Queuing length prediction method in multi-source traffic information environment | |
| CN107563566A (en) | A kind of run time interval prediction method between bus station based on SVMs | |
| CN109711617A (en) | A kind of medium-term and long-term Runoff Forecast method based on BLSTM deep learning | |
| Karwa et al. | Predicting pavement marking retroreflectivity using artificial neural networks: Exploratory analysis | |
| CN104050319B (en) | A kind of method of the complicated traffic control algorithm of real-time online checking | |
| CN110310479A (en) | A kind of Forecast of Urban Traffic Flow forecasting system and method | |
| CN114565187B (en) | Traffic network data prediction method based on graph space-time self-coding network | |
| CN109191845A (en) | A kind of public transit vehicle arrival time prediction technique | |
| CN108417032A (en) | A kind of downtown area curb parking demand analysis prediction technique | |
| CN109920248A (en) | A kind of public transport arrival time prediction technique based on GRU neural network | |
| CN108491664A (en) | A kind of Distributed Hydrological forecasting model paradigmatic system | |
| CN108416458A (en) | A kind of tunnel rich water rock mass Synthetic Geological Prediction Ahead of Construction method based on BP neural network | |
| Xue et al. | Expressway project cost estimation with a convolutional neural network model | |
| US20240310799A1 (en) | Digital and intelligent operation and maintenance system for tailings pond | |
| CN106228268A (en) | A kind of Forecasting Methodology of expressway travel time based on charge data | |
| CN108615360A (en) | Transport need based on neural network Evolution Forecast method day by day | |
| CN105355038A (en) | Method for predicting short-term traffic flow through employing PMA modeling | |
| Xu et al. | Vehicle-induced deflection prediction using long short-term memory networks | |
| Tian et al. | Deep learning method for traffic accident prediction security | |
| Li et al. | A data-driven rutting depth short-time prediction model with metaheuristic optimization for asphalt pavements based on RIOHTrack |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190510 |