CN106372722A - Subway short-time flow prediction method and apparatus - Google Patents
Subway short-time flow prediction method and apparatus Download PDFInfo
- Publication number
- CN106372722A CN106372722A CN201610830343.XA CN201610830343A CN106372722A CN 106372722 A CN106372722 A CN 106372722A CN 201610830343 A CN201610830343 A CN 201610830343A CN 106372722 A CN106372722 A CN 106372722A
- Authority
- CN
- China
- Prior art keywords
- subway
- short
- term
- dense layer
- time point
- 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
- 238000000034 method Methods 0.000 title abstract description 20
- 238000013135 deep learning Methods 0.000 claims abstract description 35
- 230000015654 memory Effects 0.000 claims abstract description 34
- 238000013277 forecasting method Methods 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 9
- 230000006403 short-term memory Effects 0.000 claims description 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims 1
- 238000005070 sampling Methods 0.000 abstract description 4
- 230000007935 neutral effect Effects 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000001914 filtration Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 230000010354 integration Effects 0.000 description 3
- 230000001537 neural effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 2
- 125000006850 spacer group Chemical group 0.000 description 2
- 238000000714 time series forecasting Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 241000353097 Molva molva Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 241000728173 Sarima Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Physiology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a subway short-time flow prediction method and apparatus. The method comprises the following steps: inputting subway entry flows and subway departure flows of multiple historical sampling time points into a long-short-term memory deep learning network for learning, and determining parameters of the long-short-term memory deep learning network; and according to the learnt long-short-term memory deep learning network and the subway entry flows and the subway departure flows of the historical sampling time points, predicting a subway entry flow and a subway departure flow of a current sampling time. According to the invention, a subway short-time flow can be predicted, and trip decision-making and transport scheduling are facilitated.
Description
Technical field
The present invention relates to data prediction field, more particularly, to a kind of subway short term traffic forecasting method and device.
Background technology
Increasingly flourishing with public transport links, the Green Travel such as subway and public transport becomes the important composition portion of Urban Traffic
Point.According to statistics, 2014, Shanghai Underground had operating line 15,578 kilometers of total line length, 339, station, system-wide net
Average daily 7,740,000 person-times of the volume of the flow of passengers, accounts for city bus trip proportion and reaches 43% about.Analysis and prediction metro passenger flow, Neng Gouwei
The passenger organization at circuit and station provides data supporting with optimizing, and provides more decision-making foundations for resident trip, and subway congestion is asked
Topic is it is also possible to evacuate offer decision-making foundation for some important sports events, activity.
In Passenger flow forecast model in short-term, specifically include that the Forecasting Methodology such as time serieses, card based on lineary system theory
The methods such as Kalman Filtering；And the Forecasting Methodology based on nonlinear system theory, such as neutral net, wavelet analysises, supporting vector
The methods such as machine.
1. time series forecasting
Time Series Forecasting Methods are broadly divided into three kinds of important methods, and they are regression model respectively, integrated model and
From smoothing model.Ahmed et al. analyzes freeway traffic flow amount using box-jenkins Time series analysis method, sends out
Existing autoregression integration moving average model possesses more excellent effect.Williams et al. improved based on arima it is proposed that
The models such as explanatory variable type autoregression smoothing model seasonality autoregression integration moving average model (sarima), improve specific
Traffic flow forecasting precision under scene, but arima model is set up in the hypothesis that time lag, variable linearly associated, institute
The non-linear relation that physical presence can not be reflected with them.
2. Kalman filtering
Kalman filtering is a kind of filtering method for time-varying stochastic signal.Kalman filtering is successfully answered by numerous scholars
For prediction of short-term traffic volume, obtain better effects.Chien utilize Kalman prediction different time in specific starting point-
The hourage of terminating point pair obtains gratifying result.Wang et al. uses EKF and some traffic controls
Algorithm processed, the traffic behavior that can complete highway estimates, short time traffic conditions are predicted, travelling evaluation and prediction etc..
3. wavelet analysises
Small echo is the waveform that a kind of special limited length, meansigma methodss are 0, can portray the office of signal well in time domain
Portion's property, again can be in the locality of frequency domain reaction signal. and wavelet analysises are applied relatively broad in prediction of short-term traffic volume.hong
Chen etc. is directed to the network topology of highway, continuous and discontinuous flow, and track uplink and downlink flow establishes
Wavelet-rbf forecast model, this model is had degree of precision .bidisha etc. and is divided with reference to Bayes's level using wavelet analysises
Analysis method carries out forecasting traffic flow and achieves good effect.
4. neutral net (nn)
Due to possessing the features such as self adaptation and robustness, various neutral net variants are used for predictive study.Tsaiet etc.
People predicts railway flow by building many timeslices neutral net and concurrent integration neutral net, proves both approaches simultaneously
It is superior to traditional multilayer perceptron (mlp).Cui etc. sets up forecast model, study speed using improving backpropagation neural network
Rate is no longer constant so that model has adaptive feature.Corinne utilizes neural network traffic flow forecasting mould
Type, achieves the pretty good .jiang that predicts the outcome of comparison using a kind of dynamic wavelet neural network self similarity containing traffic flow
Property, singularity and fractal property.Genetic algorithm and neutral net are combined and simplify network structure by abdulhai, improve pre-
Survey precision.
5. support vector machine (svm)
Svm is based on structural risk minimization (srm), compared with the traditional neural based on empirical risk minimization distally (erm)
Network possesses the more preferable learning efficiency and performance.Qian utilizes improved adaptive GA-IAGA Optimized Least Square Support Vector
(lssvm) punishment parameter and nuclear parameter are predicted to metro passenger flow, obtain better effects.Leng combines wavelet analysises
And lssvm to Beijing Metro, passenger flow is predicted in short-term, achieve and divide than wavelet neural network wavelet-nn and mode
Solve the more preferable estimated performance of backward neutral net emd-bpn.
But in passenger flow estimation before, mainly for traffic above-ground stream, air transport network, rarer research is entered to subway
Row short-term prediction.It would therefore be highly desirable to propose one kind passenger flow estimation strategy in short-term, the person that can help metro operation is in peak period to subway
Operational plan is adjusted to adapt to change, and subway operation personnel can also be carried to metro passenger flow by passenger flow estimation in short-term simultaneously
Before do specific aim and prepare, passenger then can know subway passenger flow situation in advance, contribute to slowing down subway congestion.
Content of the invention
Present invention seek to address that problem as described above.It is an object of the present invention to provide a kind of subway flow is pre- in short-term
Survey method and device, to realize the prediction to subway flow in short-term, facilitate trip decision-making and transportation dispatching.
According to the first aspect of the invention, a kind of subway short term traffic forecasting method, comprising: by multiple history samples times
The subway of point enters the station flow and the long memory deep learning network in short-term of subway outbound traffic input is learnt, and determines described length
When memory deep learning network parameter；According to described length memory deep learning network and the history samples time in short-term after study
The subway of point enters the station flow and the prediction of subway outbound traffic obtains the subway of present sample time point and enters the station flow and subway outbound
Flow.
According to the second aspect of the invention, a kind of subway short term traffic forecasting device, comprising: data input cell, is used for
The subway inputting multiple history samples time points enters the station flow and subway outbound traffic to long memory deep learning network in short-term；Long
Short term memory deep learning network, for being learnt based on the input data of described data input cell to determine network ginseng
Number, and the subway of the history samples time point according to the input of described data input cell enters the station flow and subway outbound after study
Flow, the subway that prediction obtains present sample time point enters the station flow and subway outbound traffic.
A kind of subway short term traffic forecasting method and device proposed by the present invention, for subway, passenger flow forecast is asked in short-term
Topic, gives full play to the advantage of deep learning, metro passenger flow is predicted, facilitate trip decision-making and transportation dispatching.
Following description for exemplary embodiment is read with reference to the drawings, other property features of the present invention and advantage will
It is apparent from.
Brief description
It is incorporated in description and constitutes the accompanying drawing of a part for description and show embodiments of the invention, and with
Description is used for explaining the principle of the present invention together.In the drawings, similar reference is used for representing similar key element.Under
Accompanying drawing in the description of face is some embodiments of the present invention, rather than whole embodiments.Those of ordinary skill in the art are come
Say, on the premise of not paying creative work, other accompanying drawings can be obtained according to these accompanying drawings.
Fig. 1 schematically illustrates a kind of flow chart of subway short term traffic forecasting method；
Fig. 2 schematically illustrates a kind of frame of long memory deep learning network in short-term in subway short term traffic forecasting method
Frame figure；
Fig. 3 schematically illustrates the functional structure chart of hidden unit in Fig. 2；
Fig. 4 schematically illustrates a kind of structured flowchart of subway short term traffic forecasting device.
Specific embodiment
Purpose, technical scheme and advantage for making the embodiment of the present invention are clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described it is clear that described embodiment is
The a part of embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art
The every other embodiment being obtained under the premise of not making creative work, broadly falls into the scope of protection of the invention.Need
Illustrate, in the case of not conflicting, the embodiment in the application and the feature in embodiment can mutual combination in any.
As shown in figure 1, a kind of mobile terminal subway short term traffic forecasting method, comprising:
Step 101: by the subway of multiple history samples time points enter the station flow and subway outbound traffic input length remember in short-term
Recall deep learning network to be learnt；
Determined according to experiment, the structure of length of the present invention memory deep learning network in short-term can be as shown in Figure 2: includes from defeated
Enter lstm (long-short term memory, the length time memory) layer, the dense layer that side is sequentially connected to outlet side
(dense layer), the 2nd dense layer and the 3rd dense layer；The activation primitive of wherein said lstm layer and a dense layer is
Sigmoid function (s sigmoid growth curve)；The activation primitive of described 2nd dense layer and the 3rd dense layer is that relu (corrects line
Property unit, rectified linear unit) function.Described lstm layer, a dense layer, the 2nd dense layer and the 3rd
In dense layer, the quantity of hidden unit is respectively 100,256,32 and 1.
The functional structure of each hidden unit is as shown in figure 3, lstm unit preserves historical information by memory element.Go through
The renewal of history information and using mainly being controlled by three doors, is input gate respectively, forgets door, out gate.
If h exports for lstm unit, c is the value of long memory deep learning network mnemon in short-term, and x is input data.
The renewal of lstm unit is broadly divided into following step.
1st step: according to candidate's mnemon value of conventional recursive neural computing current timewxc、whcRight respectively
The weights of memory deep learning network unit output when answering input data and a upper moment length.
2nd step: calculate value i of input gatet, input gate is for controlling the shadow to mnemon state value for the present input data
Ring.The calculating of all is except by present input data xtWith upper moment unit output valve ht-1Impact is outer, is also subject to a upper moment
Mnemon value ct-1Impact.
it=σ (wxixt+whiht-1+wcict-1+bi) (3)
3rd step: calculate value f forgeing doort, forget door and be used for controlling the shadow to current mnemon state value for the historical information
Ring.
ft=σ (wxfxt+whfht-1+wcfct-1+bf) (4)
4th step: calculate current time mnemon state value ct
Wherein represent pointwise product, mnemon state updates and depends on oneself state ct- 1And current candidate's note
Recall cell valueAnd respectively this two parts factor is adjusted by input gate and forgetting door.
5th step: calculate out gate ot, for controlling the output of mnemon state value
ot=σ (wxoxt+whoht-1+wcoct-1+bo) (6)
6th step: calculate the last output of long memory deep learning network unit in short-term
ht=ot·tanh(ct) (7)
Wherein σ typically takes logistic regression function, and span is (0,1):
Preferably, with continued reference to such as Fig. 2, the subway of multiple history samples time points is entered the station flow and subway outbound traffic
Long memory deep learning network in short-term is learnt with the input of time type；Described time type is each history samples time point
Classification；Described classification can comprise seven classes: characterize the first kind on first day working day, the Equations of The Second Kind characterizing in work, characterize work
Make the 3rd class of last day day, characterize the 4th class having a holiday or vacation first day, the 5th class characterizing in having a holiday or vacation, characterize and have a holiday or vacation last
It the 6th class and the 7th class characterizing vacation in single day；Described classification and history samples time point are inputted using one-hot mode,
One-hot is a kind of coding, also referred to as one efficient coding.
Specifically: can be entered the station flow and subway outbound traffic and time using continuous six history samples time point subways
The long memory deep learning network in short-term of type input is learnt；Wherein, two neighboring history samples time point be spaced apart 15
Minute, and front 5 history samples time points are as input feature vector, and the 6th history samples time point is as output.
Step 103: determine the parameter of described length memory deep learning network in short-term；
Step 105: according to the subway of described length memory deep learning network and history samples time point in short-term after study
Enter the station flow and the prediction of subway outbound traffic obtains the subway of present sample time point and enters the station flow and subway outbound traffic.
The present embodiment can also include at least one in following optimal way:
1: described 2nd dense layer and the 3rd dense layer are full articulamentum；
2: the described subway by multiple history samples time points enter the station flow and subway outbound traffic input length remember in short-term
Recall and prevent over-fitting using the mode of dropout [24,25] during deep learning network is learnt, dropout is to intend
Close solution, the core concept of this technology is that unit is abandoned at random together with their connection, prevents unit from being formed excessively
Rely on；
3: the described subway by multiple history samples time points enter the station flow and subway outbound traffic input length remember in short-term
Recall during deep learning network is learnt by dropout ratio setting be 0.2；
The present embodiment subway short term traffic forecasting method, for subway passenger flow forecast problem in short-term, gives full play to depth
The advantage of study, is predicted to metro passenger flow, facilitates trip decision-making and transportation dispatching.Simultaneously by real metro passenger flow
The study of data, using standard the method for assessment it was demonstrated that method presented herein compares other subways passenger flow forecasting in short-term,
Obtain more preferable, more sane predicting the outcome.
Fig. 4 show the corresponding device of method shown in Fig. 1, and the explanation of Fig. 1-Fig. 2 goes for the present embodiment, such as
Shown in Fig. 4, a kind of subway short term traffic forecasting device, comprising:
Data input cell 40, the subway for inputting multiple history samples time points enters the station flow and subway outbound traffic
To long memory deep learning network in short-term；
Long memory deep learning network 42 in short-term, for based on the input data of described data input cell learnt with
Determine network parameter, and the subway of the history samples time point according to the input of described data input cell enters the station flow after study
And subway outbound traffic, the subway that prediction obtains present sample time point enters the station flow and subway outbound traffic.
The present embodiment can also include at least one in following optimal way:
1: the described length lstm layer that memory deep learning network includes being sequentially connected from input side to outlet side in short-term, first
Dense layer, the 2nd dense layer and the 3rd dense layer；The activation primitive of wherein said lstm layer and a dense layer is
Sigmoid function；The activation primitive of described 2nd dense layer and the 3rd dense layer is relu function；:
2: described 2nd dense layer and the 3rd dense layer are full articulamentum.
3: described data input cell is additionally operable to receive the time type of the classification characterizing each history samples time point；Described
Classification and history samples time point are inputted using one-hot mode.
4: the quantity difference of hidden unit in described lstm layer, a dense layer, the 2nd dense layer and the 3rd dense layer
For 100,256,32 and 1.
In concrete training experiment, first the volume of the flow of passengers can be included subway outbound with 15 minutes for predicting interval unit
Flow input lstm is trained with entering the station, and because time type is very big on volume of the flow of passengers impact, time type is included time class
Type and sampling time point are trained as feature, and time type is inputted using one-hot mode with time point.Pre- using 5
Survey spacer units as input feature vector, the 6th spacer units as output, are sought into line parameter by many experiments and analyzing
Excellent, each layer is used when lstm trains hidden unit number and activation primitive are as shown in table1.
Table 1 deep neural network parameter
Over-fitting easily occurs, in the training process, by dropout [24,25] mode more than the deep neural network number of plies
Prevent over-fitting, dropout randomly chooses weights and is updated, so that the probability that two nodes occur simultaneously is reduced, it is to avoid one
Individual node depends on another node to cause model generalization ability to die down.Carry out arameter optimization by many experiments, herein will
Dropout ratio setting is 0.2, finds to add two dense layer prediction effects preferably simultaneously, uses in full articulamentum
Relu is as activation primitive.
Relu can represent, negative loop is set to 0 by relu by formula (9), and the way retaining positive portions implies for model
Layer introduces openness, therefore improves the performance of model, relu is a kind of unsaturation activation primitive with respect to saturation activation letter
Number, can accelerate model convergence rate, can be exported by relu real-valued simultaneously in the training process, and make a depth god
Do not need unsupervised training in advance (pre-train) through network, directly carry out the training having supervision.
The subway short term traffic forecasting device of the present embodiment will be used for the special construction lstm of speech recognition in deep learning
Novelty for subway subway passenger flow forecast in short-term, can flow be carried out efficiently in short-term to the subway in the range of certain time
Prediction；Model refinement is done to lstm simultaneously, using different neutral net activation primitives in different neural net layers, lift flow
Precision of prediction, and lstm training method is improved, prevent volume forecasting process using multiple dropout, full articulamentum
The Expired Drugs of middle generation, prediction effect is relatively stable.
Descriptions above can combine individually or in every way enforcement, and these variant all exist
Within protection scope of the present invention.
Finally it is noted that above example, only in order to technical scheme to be described, is not intended to limit.Although
With reference to the foregoing embodiments the present invention is described in detail, it will be understood by those within the art that: it still may be used
To modify to the technical scheme described in foregoing embodiments, or equivalent is carried out to wherein some technical characteristics；
And these modification or replace, do not make appropriate technical solution essence depart from various embodiments of the present invention technical scheme spirit and
Scope.
Claims (11)
1. a kind of subway short term traffic forecasting method is it is characterised in that include:
The subway of multiple history samples time points is entered the station flow and subway outbound traffic input long memory depth in short-term study net
Network is learnt, and determines the parameter of described length memory deep learning network in short-term；
Entered the station flow and ground according to the subway of described length memory deep learning network and history samples time point in short-term after study
The subway that the prediction of ferrum outbound traffic obtains present sample time point enters the station flow and subway outbound traffic.
2. subway short term traffic forecasting method according to claim 1 is it is characterised in that described length memory depth in short-term
Practise lstm layer, a dense layer, the 2nd dense layer and the 3rd dense that network includes being sequentially connected from input side to outlet side
Layer；The activation primitive of wherein said lstm layer and a dense layer is sigmoid function；Described 2nd dense layer and the 3rd
The activation primitive of dense layer is relu function.
3. subway short term traffic forecasting method according to claim 2 is it is characterised in that described 2nd dense layer and
Three dense layers are full articulamentum, and/or, enter the station flow and subway in the described subway by multiple history samples time points
Outbound traffic input length is prevented using the mode of dropout [24,25] during memory deep learning network is learnt in short-term
Only over-fitting.
4. subway short term traffic forecasting method according to claim 3 is it is characterised in that described by multiple history samples
The subway of time point enter the station flow and the input of subway outbound traffic long during memory deep learning network is learnt in short-term
Dropout ratio setting is 0.2.
5. the subway short term traffic forecasting method according to any one of claim 2-4 is it is characterised in that described will be multiple
The subway of history samples time point enters the station flow and the long memory deep learning network in short-term of subway outbound traffic input is learnt
Including:
Enter the station flow and subway outbound traffic of the subway of multiple history samples time points is inputted long short term memory with time type
Deep learning network is learnt；Described time type is the classification of each history samples time point.
6. subway short term traffic forecasting method according to claim 5, described classification comprises seven classes: characterizes working day first
It the first kind, characterize work in Equations of The Second Kind, characterize last day on working day the 3rd class, characterize have a holiday or vacation first day the 4th
Class, the 5th class characterizing in having a holiday or vacation, the 6th class characterizing last day of having a holiday or vacation and the 7th class characterizing vacation in single day；And/or
Person,
Described classification and history samples time point are inputted using one-hot mode.
7. subway short term traffic forecasting method according to claim 5, the described subway by multiple history samples time points
Entering the station, long memory deep learning network in short-term carries out study and includes with the input of time type for flow and subway outbound traffic:
Long with the input of time type in short-term using enter the station flow and subway outbound traffic of continuous six history samples time point subways
Memory deep learning network is learnt；Wherein, being spaced apart 15 minutes of two neighboring history samples time point, and first 5
, as input feature vector, the 6th history samples time point is as output for history samples time point.
8. subway short term traffic forecasting method according to claim 5, described lstm layer, a dense layer, second
In dense layer and the 3rd dense layer, the quantity of hidden unit is respectively 100,256,32 and 1.
9. a kind of subway short term traffic forecasting device is it is characterised in that include:
Data input cell, the subway for inputting multiple history samples time points enters the station flow and subway outbound traffic to length
When memory deep learning network；
Long memory deep learning network in short-term, for being learnt based on the input data of described data input cell to determine net
Network parameter, and the subway of the history samples time point according to the input of described data input cell enters the station flow and subway after study
Outbound traffic, the subway that prediction obtains present sample time point enters the station flow and subway outbound traffic.
10. subway short term traffic forecasting device according to claim 9 is it is characterised in that described length memory depth in short-term
Lstm layer that learning network includes being sequentially connected from input side to outlet side, a dense layer, the 2nd dense layer and the 3rd
Dense layer；The activation primitive of wherein said lstm layer and a dense layer is sigmoid function；Described 2nd dense layer and
The activation primitive of the 3rd dense layer is relu function；And/or,
Described 2nd dense layer and the 3rd dense layer are full articulamentum.
11. subway short term traffic forecasting devices according to claim 10 it is characterised in that described data input cell also
For receiving the time type of the classification characterizing each history samples time point；Described classification and history samples time point adopt one-
Hot mode inputs；And/or,
In described lstm layer, a dense layer, the 2nd dense layer and the 3rd dense layer the quantity of hidden unit be respectively 100,
256th, 32 and 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610830343.XA CN106372722A (en) | 2016-09-18 | 2016-09-18 | Subway short-time flow prediction method and apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610830343.XA CN106372722A (en) | 2016-09-18 | 2016-09-18 | Subway short-time flow prediction method and apparatus |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106372722A true CN106372722A (en) | 2017-02-01 |
Family
ID=57897588
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610830343.XA Pending CN106372722A (en) | 2016-09-18 | 2016-09-18 | Subway short-time flow prediction method and apparatus |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106372722A (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909989A (en) * | 2017-02-24 | 2017-06-30 | 国网河南省电力公司电力科学研究院 | A kind of grid disturbance Forecasting Methodology and device |
| CN107145962A (en) * | 2017-04-01 | 2017-09-08 | 上海诺悦智能科技有限公司 | A kind of sight spot domestic visitors forecasting system |
| CN107316096A (en) * | 2017-05-05 | 2017-11-03 | 北京市交通信息中心 | A kind of track traffic one-ticket pass passenger amount of entering the station Forecasting Methodology |
| CN108111591A (en) * | 2017-12-15 | 2018-06-01 | 北京小米移动软件有限公司 | The method, apparatus and computer readable storage medium of PUSH message |
| CN108260087A (en) * | 2018-01-31 | 2018-07-06 | 福州大学 | A kind of indoor stream of people's distribution forecasting method based on WIIFI long short-term memories |
| CN108289203A (en) * | 2018-02-02 | 2018-07-17 | 常州高清信息技术有限公司 | A kind of video monitoring system for rail traffic |
| CN108446759A (en) * | 2018-02-11 | 2018-08-24 | 陕西师范大学 | Tourist flows amount prediction technique based on seasonal shot and long term memory network model |
| CN108537392A (en) * | 2018-04-26 | 2018-09-14 | 重庆大学 | A method of standee's flow is pointed out based on space-time characteristic predicted orbit traffic station |
| CN108564227A (en) * | 2018-04-26 | 2018-09-21 | 重庆大学 | A kind of track traffic for passenger flow amount prediction technique based on space-time characteristic |
| CN108564789A (en) * | 2018-06-11 | 2018-09-21 | 重庆大学 | Expressway ramp up-downlink direction flow estimation method based on charge station's data |
| CN109243172A (en) * | 2018-07-25 | 2019-01-18 | 华南理工大学 | Traffic flow forecasting method based on genetic algorithm optimization LSTM neural network |
| CN109308543A (en) * | 2018-08-20 | 2019-02-05 | 华南理工大学 | The short-term passenger flow forecasting of subway based on LS-SVM and real-time big data |
| CN109657881A (en) * | 2019-01-14 | 2019-04-19 | 南京国电南自电网自动化有限公司 | A kind of neural network photovoltaic power generation prediction technique and system suitable for small sample |
| CN110020738A (en) * | 2018-01-10 | 2019-07-16 | 北京京东金融科技控股有限公司 | For the method, apparatus of data processing, electronic equipment and computer-readable medium |
| CN110381524A (en) * | 2019-07-15 | 2019-10-25 | 安徽理工大学 | The mobile flow on-line prediction method of large scene based on Bi-LSTM, system and storage medium |
| CN111311905A (en) * | 2020-01-21 | 2020-06-19 | 北京工业大学 | Particle swarm optimization wavelet neural network-based expressway travel time prediction method |
| CN111597971A (en) * | 2020-05-14 | 2020-08-28 | 北京交通大学 | Method for predicting short-term arrival passenger flow of urban rail transit |
| CN109492814B (en) * | 2018-11-15 | 2021-04-20 | 中国科学院深圳先进技术研究院 | Urban traffic flow prediction method, system and electronic equipment |
| CN113159385A (en) * | 2021-03-22 | 2021-07-23 | 中铁信(北京)网络技术研究院有限公司 | Self-feedback passenger flow control method and device based on deep reinforcement learning |
| CN114091579A (en) * | 2021-11-03 | 2022-02-25 | 深圳技术大学 | Urban rail transit passenger flow early warning system and method |
| CN115146843A (en) * | 2022-06-27 | 2022-10-04 | 北京交通大学 | Meta learning-based method for predicting short-term passenger flow of new subway station |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992223A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Dense population estimation method based on deep learning |
| CN105389980A (en) * | 2015-11-09 | 2016-03-09 | 上海交通大学 | Short-time traffic flow prediction method based on long-time and short-time memory recurrent neural network |
-
2016
- 2016-09-18 CN CN201610830343.XA patent/CN106372722A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992223A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Dense population estimation method based on deep learning |
| CN105389980A (en) * | 2015-11-09 | 2016-03-09 | 上海交通大学 | Short-time traffic flow prediction method based on long-time and short-time memory recurrent neural network |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909989A (en) * | 2017-02-24 | 2017-06-30 | 国网河南省电力公司电力科学研究院 | A kind of grid disturbance Forecasting Methodology and device |
| CN107145962A (en) * | 2017-04-01 | 2017-09-08 | 上海诺悦智能科技有限公司 | A kind of sight spot domestic visitors forecasting system |
| CN107316096A (en) * | 2017-05-05 | 2017-11-03 | 北京市交通信息中心 | A kind of track traffic one-ticket pass passenger amount of entering the station Forecasting Methodology |
| CN108111591A (en) * | 2017-12-15 | 2018-06-01 | 北京小米移动软件有限公司 | The method, apparatus and computer readable storage medium of PUSH message |
| CN110020738A (en) * | 2018-01-10 | 2019-07-16 | 北京京东金融科技控股有限公司 | For the method, apparatus of data processing, electronic equipment and computer-readable medium |
| CN108260087A (en) * | 2018-01-31 | 2018-07-06 | 福州大学 | A kind of indoor stream of people's distribution forecasting method based on WIIFI long short-term memories |
| CN108289203A (en) * | 2018-02-02 | 2018-07-17 | 常州高清信息技术有限公司 | A kind of video monitoring system for rail traffic |
| CN108446759A (en) * | 2018-02-11 | 2018-08-24 | 陕西师范大学 | Tourist flows amount prediction technique based on seasonal shot and long term memory network model |
| CN108564227B (en) * | 2018-04-26 | 2021-08-06 | 重庆大学 | Rail transit passenger flow volume prediction method based on space-time characteristics |
| CN108537392A (en) * | 2018-04-26 | 2018-09-14 | 重庆大学 | A method of standee's flow is pointed out based on space-time characteristic predicted orbit traffic station |
| CN108564227A (en) * | 2018-04-26 | 2018-09-21 | 重庆大学 | A kind of track traffic for passenger flow amount prediction technique based on space-time characteristic |
| CN108537392B (en) * | 2018-04-26 | 2021-08-17 | 重庆大学 | Method for predicting outbound passenger flow of rail transit station based on space-time characteristics |
| CN108564789A (en) * | 2018-06-11 | 2018-09-21 | 重庆大学 | Expressway ramp up-downlink direction flow estimation method based on charge station's data |
| CN108564789B (en) * | 2018-06-11 | 2020-11-17 | 重庆大学 | Expressway ramp up-down direction flow estimation method based on toll station data |
| CN109243172A (en) * | 2018-07-25 | 2019-01-18 | 华南理工大学 | Traffic flow forecasting method based on genetic algorithm optimization LSTM neural network |
| CN109243172B (en) * | 2018-07-25 | 2021-06-08 | 华南理工大学 | Traffic flow prediction method for optimizing LSTM neural network based on genetic algorithm |
| CN109308543A (en) * | 2018-08-20 | 2019-02-05 | 华南理工大学 | The short-term passenger flow forecasting of subway based on LS-SVM and real-time big data |
| CN109308543B (en) * | 2018-08-20 | 2021-10-22 | 华南理工大学 | Subway short-term passenger flow prediction method based on LS-SVM and real-time big data |
| CN109492814B (en) * | 2018-11-15 | 2021-04-20 | 中国科学院深圳先进技术研究院 | Urban traffic flow prediction method, system and electronic equipment |
| CN109657881A (en) * | 2019-01-14 | 2019-04-19 | 南京国电南自电网自动化有限公司 | A kind of neural network photovoltaic power generation prediction technique and system suitable for small sample |
| CN110381524A (en) * | 2019-07-15 | 2019-10-25 | 安徽理工大学 | The mobile flow on-line prediction method of large scene based on Bi-LSTM, system and storage medium |
| CN110381524B (en) * | 2019-07-15 | 2022-12-20 | 安徽理工大学 | Bi-LSTM-based large scene mobile flow online prediction method, system and storage medium |
| CN111311905A (en) * | 2020-01-21 | 2020-06-19 | 北京工业大学 | Particle swarm optimization wavelet neural network-based expressway travel time prediction method |
| CN111597971A (en) * | 2020-05-14 | 2020-08-28 | 北京交通大学 | Method for predicting short-term arrival passenger flow of urban rail transit |
| CN111597971B (en) * | 2020-05-14 | 2024-05-21 | 北京交通大学 | Urban rail transit short-term arrival passenger flow prediction method |
| CN113159385A (en) * | 2021-03-22 | 2021-07-23 | 中铁信(北京)网络技术研究院有限公司 | Self-feedback passenger flow control method and device based on deep reinforcement learning |
| CN114091579A (en) * | 2021-11-03 | 2022-02-25 | 深圳技术大学 | Urban rail transit passenger flow early warning system and method |
| CN115146843A (en) * | 2022-06-27 | 2022-10-04 | 北京交通大学 | Meta learning-based method for predicting short-term passenger flow of new subway station |
| CN115146843B (en) * | 2022-06-27 | 2023-08-15 | 北京交通大学 | Meta learning-based short-term passenger flow prediction method for new subway driving station |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106372722A (en) | Subway short-time flow prediction method and apparatus | |
| CN105654729B (en) | A kind of short-term traffic flow forecast method based on convolutional neural networks | |
| CN109034566B (en) | A kind of intelligent dispatching method and device based on passenger flow above and below bus station | |
| CN108831153A (en) | A kind of traffic flow forecasting method and device using spatial and temporal distributions characteristic | |
| CN109558985A (en) | A kind of bus passenger flow amount prediction technique based on BP neural network | |
| CN106652434A (en) | Bus dispatching method based on rail transit coordination | |
| CN107958316A (en) | The train dwelling optimization method and system of a kind of rail line | |
| CN106205156A (en) | A kind of crossing self-healing control method for the sudden change of part lane flow | |
| CN107480807A (en) | Shared bicycle destination Forecasting Methodology and device based on space-time layered perception neural networks | |
| CN109345832B (en) | Urban road overtaking prediction method based on deep recurrent neural network | |
| CN113537626B (en) | Method for predicting neural network combined time sequence by aggregating information difference | |
| CN106021902B (en) | A kind of stage division for urban track traffic for passenger flow congestion index | |
| CN109002625A (en) | A kind of feeder bus sytem network design method based on robust optimization | |
| CN106934229A (en) | Consider the track traffic station platform aggregation number computational methods of stayer | |
| CN105447116A (en) | Mamdani algorithm based parking guidance decision-making method | |
| CN110210664A (en) | The deep learning method of multiple individual vehicle usage behavior short-term predictions | |
| Rashidi et al. | Using automatic vehicle location data to model and identify determinants of bus bunching | |
| Zou et al. | Short-term traffic flow prediction based on PCC-BiLSTM | |
| CN114418606B (en) | Network vehicle order demand prediction method based on space-time convolution network | |
| Hlaing et al. | Big traffic data analytics for smart urban intelligent traffic system using machine learning techniques | |
| Al-Najada et al. | Real-time incident clearance time prediction using traffic data from internet of mobility sensors | |
| Zhang et al. | Simulation of fire emergency evacuation in metro station based on cellular automata | |
| Ayman et al. | Neural architecture and feature search for predicting the ridership of public transportation routes | |
| Xue et al. | Expressway traffic flow prediction model based on Bi-LSTM neural networks | |
| Osman et al. | Application of long short term memory networks for long-and short-term bus travel time prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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: 20170201 |
|
| RJ01 | Rejection of invention patent application after publication |