CN109598935A - A kind of traffic data prediction technique based on ultra-long time sequence - Google Patents
A kind of traffic data prediction technique based on ultra-long time sequence Download PDFInfo
- Publication number
- CN109598935A CN109598935A CN201811532408.8A CN201811532408A CN109598935A CN 109598935 A CN109598935 A CN 109598935A CN 201811532408 A CN201811532408 A CN 201811532408A CN 109598935 A CN109598935 A CN 109598935A
- Authority
- CN
- China
- Prior art keywords
- traffic data
- data
- ultra
- long time
- traffic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
- G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
- G08G1/0133—Traffic data processing for classifying traffic situation
Abstract
The present invention relates to a kind of traffic data prediction techniques based on ultra-long time sequence, first, position mark processing is carried out to historical traffic data, the method for carrying out position mark processing to historical traffic data is as follows: determining the corresponding relationship for the data space that individual data position and ultra-long time sequence are formed in traffic data ultra-long time sequence;Position mark is made to the traffic data in traffic data ultra-long time sequence according to corresponding relationship;The ultra-long time sequence refers to across multiple time serieses for dividing the period;Finally, choosing the historical traffic data with position mark as input data, traffic data prediction is carried out.Using a kind of method arbitrary in neural net model method, the method for moving average, exponential smoothing, AR modelling when traffic data is predicted.The present invention improves traffic data precision of prediction by the correlation of Data Position in perception traffic data ultra-long time sequence.
Description
Technical field
The present invention relates to traffic data prediction field more particularly to a kind of traffic data predictions based on ultra-long time sequence
Method.
Background technique
The development of real-time dynamic traffic, so that accurately traffic data becomes hot topic in real time.However, traffic data is by the time
The influence of factor is very big.It is broadly divided into active element and passive factor.Active element has: working day and nonworkdays, seasonality
Transformation, time passage etc..Passively because being known as: festivals or holidays, activity of guide car changes in flow rate etc..
Traffic data time series can be according to the Time segments divisions such as minute, hour, day, week, the moon, year, traffic data prediction
The traffic data that the main traffic data using historical time section carries out future time section is predicted.Existing traffic forecast method,
Such as method of moving average, exponential smoothing, AR modelling, to across it is multiple divide the periods ultra-long time sequences perception compared with
It is weak;Therefore a kind of traffic data prediction technique based on ultra-long time sequence is designed to be necessary.
Summary of the invention
The present invention is to overcome above-mentioned shortcoming, and it is an object of the present invention to provide a kind of traffic data based on ultra-long time sequence
It is pre- to improve traffic data by the correlation of Data Position in perception traffic data ultra-long time sequence by prediction technique, the present invention
Survey precision.
The present invention is to reach above-mentioned purpose by the following technical programs: a kind of traffic data based on ultra-long time sequence is pre-
Survey method, includes the following steps:
(1) position mark processing is carried out to historical traffic data, wherein the method for carrying out position mark processing is as follows:
(1.1) data for determining that individual data position and ultra-long time sequence are formed in traffic data ultra-long time sequence are empty
Between corresponding relationship;
(1.2) position mark is made to the traffic data in traffic data ultra-long time sequence according to corresponding relationship;It is described super
Long-term sequence refers to across multiple time serieses for dividing the period;
(2) historical traffic data with position mark is chosen as input data, carries out traffic data prediction.
Preferably, using regression model, user setting method, neural network model side in the step (1.1)
Any one method in method can determine individual data position and ultra-long time sequence shape in traffic data ultra-long time sequence
At data space corresponding relationship;It is preferred that using regression model, formula is as follows:
Q=∑ (y- ∑ (wpxp+bp))=min
Wherein, y is the traffic data on individual data position, and p indicates data space position, asks so that Q function minimization
W, b, the corresponding relationship as individual data position and data space.
Preferably, according to corresponding relationship to the traffic number in traffic data ultra-long time sequence in the step (1.2)
According to position mark is made, wherein obtaining w, b using regression model, according to the size of w, different stage is divided, as position mark.
Preferably, the division period determines according to the periodic phenomena that traffic data is presented, select multiple groups with
Time window be unit traffic data time series, be averaging similarity, when average similarity meet setting threshold value when, this when
Between window can be used as divide the period;Wherein similarity can pass through cosine similarity, Pearson correlation coefficients, Jaccard similarity factor
Method is calculated.
Preferably, using neural net model method, the method for moving average, exponential smoothing, AR mould in the step (2)
Arbitrary a kind of method progress traffic data prediction in type method.
Preferably, the traffic data includes but is not limited to the data of transit equipment acquisition, is acquired according to transit equipment
Data reduction data;The data of the transit equipment acquisition include but is not limited to the magnitude of traffic flow, traffic speed, the basis
The data of the data reduction of transit equipment acquisition, including but not limited to traffic behavior, traffic index.
Preferably, the neural net model method preferably uses LSTM long, neural network carries out traffic data in short-term
Prediction, specific as follows:
(i) traffic flow forecasting is carried out by the driving sequence of setting input, to predicted value and the remote magnitude of traffic flow
Value establishes linear regression model (LRM);
(ii) flow value that position mark is met the requirements is filtered out after training, and the distant place stream that the weight obtained to screening is big
Magnitude position is marked;
(iii) the traffic flow magnitude with weight with position mark is added in LSTM network input layer, carries out LSTM+
Backpropagation training, the output valve of each neuron of forward calculation first, the secondly error term of each neuron of retrospectively calculate
Value calculates the gradient of each weight finally according to corresponding error term, updates weighted value;
(iv) magnitude of traffic flow in the following preset period is predicted with trained LSTM network model.
Preferably, it is described in step (i), the driving sequence of input is set as X=(x1,1,1, x1,1,2..., xkk,
Nn, mm, yKk, nn+1,1, yKk, nn+1,2..., yKk, nn+1, t-1,),
Wherein, y indicates the data on the prediction same day, and volume forecasting question essence is to be established and pre- flow measurement by driving sequence X
Measure yKk, nn+1, tMapping.
Preferably, the step (i) when establishing linear regression model (LRM), chooses the data near daily t time interval,
Regression model is as follows:
Wherein,D is step-length,It is respectively with b
The weight and biasing of regression model.
Preferably, the LSTM network is one kind, special RNN Recognition with Recurrent Neural Network increases " defeated compared with RNN
Introduction ", " forgeing door ", " out gate ", " input gate ", " forgeing door ", " out gate " and LSTM cell state are as follows:
Input gate:
Forget door:
Cell state:
Out gate:
Wherein, I represents the length for inputting current sequence, and C indicates the quantity of LSTM cell, and P indicates labeled high influence
The quantity of wagon flow magnitude, l indicate input state,Indicate the state of forgetting door, o indicates the state of out gate, and f and g are activation
Function,Indicate c-th of t moment cellular state value, tgpIndicate p-th of labeled wagon flow magnitude for having weight, it is last defeated
Out are as follows:
The beneficial effects of the present invention are: the phase that the present invention passes through Data Position in perception traffic data ultra-long time sequence
Guan Xing, flag data location prominence, then the traffic data of these positions is screened, the base as traffic data prediction
Plinth improves traffic data precision of prediction.Preferably, the wagon flow magnitude of critical positions is added in LSTM, makes LSTM model
Not only have shot and long term memory function, but also also have certain memory function for super-long-term, so that traffic flow forecasting
It is more accurate.
Detailed description of the invention
Fig. 1 is the flow diagram of the method for the present invention.
Specific embodiment
The present invention is described further combined with specific embodiments below, but protection scope of the present invention is not limited in
This:
Embodiment 1: as shown in Figure 1, a kind of traffic data prediction technique based on ultra-long time sequence includes the following steps:
(1) position mark processing is carried out to historical traffic data, wherein the method for carrying out position mark processing is as follows:
(1.1) data for determining that individual data position and ultra-long time sequence are formed in traffic data ultra-long time sequence are empty
Between corresponding relationship;
(1.2) position mark is made to the traffic data in traffic data ultra-long time sequence according to corresponding relationship;It is described super
Long-term sequence refers to across multiple time serieses for dividing the period;
Wherein traffic data includes the data of transit equipment acquisition, such as the magnitude of traffic flow, traffic speed, according to transit equipment
The data of the data reduction of acquisition, such as traffic behavior, traffic index.
Ultra-long time sequence refers to that such as: 4 division periods are respectively 1 minute, 1 across multiple time serieses for dividing the period
Hour, 1 day, 1 week, x [k, m, n, l] indicate that the traffic data of all n-th hour the m days l/mins of kth, k, m, n, l are integer, k
Optional range [1-K], the optional range of m [1-M], the optional range of 1 < M < 8, n [1-N], the optional range of 1 < N < 25, l [1-L], 1 < L <
61;Ultra-long time sequence forms data space [1-K, 1-M, 1-N, 1-L] due to the different division periods, and [k, m, n, l] is right
Wherein a certain data space position is answered.Similar, the time series for dividing the period across 3 also forms ultra-long time sequence.
The division period determines according to the periodic phenomena that traffic data is presented, and selects friendship of the multiple groups as unit of time window
Logical data time series, are averaging similarity, when average similarity meets given threshold, when which can be used as division
Section.In the present embodiment, one group of historical traffic data collection: { (x is obtained1,t1),(x2,t2),…(xj,tj),…,(x300,
t300)},t2-t1It is 1 hour, time window TK is 12 hours, takes traffic data time series of the multiple groups as unit of TK, is with 3 groups
Example, xTK1={ x1,x2,…,x12},xTK2={ x25,x26,…,x37},xTK3={ x13,x14,…,x24, it can be by 12 small times
It chooses, or randomly selects every sequence.
Similarity can be calculated using the methods of cosine similarity, Pearson correlation coefficients, Jaccard similarity factor.S<
xTK1,xTK2> it is time series xTK1And xTK2Similarity.It is averaging similarity STK=(S < xTK1,xTK2>+S<xTK1,xTK3>+S<
xTK2,xTK3>)/3;Average similarity meets given threshold, and 12 hours time windows, which can be used as, divides the period.
Similar, time window TK is 24 hours, and average similarity meets given threshold, and time window can be used as division within 24 hours
Period.
Wherein it is possible to determine traffic data ultra-long time using the methods of regression model, user setting, neural network model
The corresponding relationship for the data space that individual data position and ultra-long time sequence are formed in sequence, preferably uses in the present embodiment
Regression model, formula are as follows:
Q=∑ (y- ∑ (wpxp+bp))=min
Wherein, y is the traffic data on individual data position, and p indicates data space position, and asking makes Q function minimization
W, b, the corresponding relationship as individual data position and data space.According to corresponding relationship in traffic data ultra-long time sequence
Traffic data make position mark, it is above-mentioned that w, b are obtained using regression model, according to the size of w, different stage is divided, as position
Tagging.
(2) historical traffic data with position mark is chosen as input data, carries out traffic data prediction.Wherein, may be used
To carry out traffic data prediction using the methods of neural net model method, the method for moving average, exponential smoothing, AR modelling.
In the present embodiment, traffic data prediction is carried out using neural net model method below;Specifically use LSTM
Long neural network in short-term carries out traffic data prediction, specific as follows:
(i) weaker analysis, the remote high influence flow of predicted value of adjusting the distance are perceived to traffic flow data by LSTM network
Value carries out feature enhancing;Traffic flow forecasting is carried out by the driving sequence of setting input, to predicted value and remote traffic
Flow value establishes linear regression model (LRM);The flow value that position mark is met the requirements is filtered out after training, and to the power that screening obtains
Great distant place flow value position is marked;
Long neural network (LSTM) in short-term is widely used in the prediction of time series, it is a kind of special circulation nerve net
Network (RNN), compared with RNN, LSTM increases " input gate ", " out gate ", and " forgeing door ", these doors allow LSTM to be provided with for a long time
The function of memory.In traffic flow data prediction, the traffic flow data the being predicted not only flow with preceding several timestamps
Data are closely related, but also related to the data on flows near working day a few days ago same timestamp, however since LSTM is missed
The presence of difference accumulation, so that LSTM is weaker for the data perception of overlength sequence, this allows LSTM in the magnitude of traffic flow for having one
The prediction data sensing capability a few days ago of fixing sound is weaker, analyzes and predicts so as to cause suboptimum.
Scheme includes two key steps, and the traffic flow magnitude for adjusting the distance first remote and predicted value establish linear regression mould
Type influences height for perceiving distant place flow value to predicted value, then filters out the flow value that position mark is met the requirements, and will
Later the traffic flow magnitude with position mark is added in LSTM input layer for screening, can effectively solve the problem that LSTM by this method
The weak problem of traffic flow data sensing capability to overlength sequence.
It is assumed herein that the driving sequence of input is
X=(x1,11, x1,1,2..., xKk, nn, mm, yKk, nn+1,1, yKk, nn+1,2..., yKk, nn+1, t-1,),
In order to facilitate differentiation, predict that the data on the same day are indicated with y, volume forecasting problem is exactly to pass through driving sequence X to establish
With predicted flow rate yKk, nn+1, tMapping.
Comparatively, the flow value before n days near same time interval influences predicted value higher, and between the distance t time
It every remoter, influences weaker, therefore in order to reduce the redundancy of data volume, when being returned, it is attached to choose daily t time interval
Close data, regression model are as follows:
WhereinD is step-length,It is back respectively with b
Return the weight and biasing of model, after the completion of training, the flow value position big to those weights is marked.
In the present embodiment, weight screening is carried out using returning, predicts the traffic data of k weeks Tuesday of kth 12:00.If step-length d
=1, internal corresponding weight matrix is as follows after the completion of training:
First week
Second week
After the completion of training, each position will have corresponding weight, with<weight, tag>indicate whether to mark,
Weight indicates corresponding weight, and position of the above example by weight greater than 0.13 is marked, these labeled positions
Data represent high correlation position.
(ii) the traffic flow magnitude with weight with position mark is added in LSTM network input layer, carries out LSTM+
Backpropagation training, the output valve of each neuron of forward calculation first, the secondly error term of each neuron of retrospectively calculate
Value calculates the gradient of each weight finally according to corresponding error term, updates weighted value;
When LSTM is trained, these labeled datas on flows with weight are added in LSTM network, LSTM
" input gate " in network, " forgeing door ", " out gate " and LSTM cell state are as follows:
Input gate:
Forget door:
Cell state:
Out gate:
Wherein I represents the length for inputting current sequence, and C indicates the quantity of LSTM cell, and P indicates labeled high influence vehicle
The quantity of flow value, l indicate input state,Indicate the state of forgetting door, o indicates the state of out gate, and f and g are activation letters
Number,Indicate c-th of t moment cellular state value, tgpIt indicates p-th of labeled wagon flow magnitude for having weight, finally exports
Are as follows:
(iii) magnitude of traffic flow in the following preset period is predicted with trained LSTM network model.
Embodiment 2: a kind of traffic data prediction technique based on ultra-long time sequence includes the following steps:
(1) position mark processing is carried out to historical traffic data, wherein the method for carrying out position mark processing is as follows:
(1.1) data for determining that individual data position and ultra-long time sequence are formed in traffic data ultra-long time sequence are empty
Between corresponding relationship;
(1.2) position mark is made to the traffic data in traffic data ultra-long time sequence according to corresponding relationship;It is described super
Long-term sequence refers to across multiple time serieses for dividing the period;
Wherein traffic data includes the data of transit equipment acquisition, such as the magnitude of traffic flow, traffic speed, according to transit equipment
The data of the data reduction of acquisition, such as traffic behavior, traffic index.
Ultra-long time sequence refers to that such as: 4 division periods are respectively 1 minute, 1 across multiple time serieses for dividing the period
Hour, 1 day, 1 week, x [k, m, n, l] indicate that the traffic data of all n-th hour the m days l/mins of kth, k, m, n, l are integer, k
Optional range [1-K], the optional range of m [1-M], the optional range of 1 < M < 8, n [1-N], the optional range of 1 < N < 25, l [1-L], 1 < L <
61;Ultra-long time sequence forms data space [1-K, 1-M, 1-N, 1-L] due to the different division periods, and [k, m, n, l] is right
Wherein a certain data space position is answered.Similar, the time series for dividing the period across 3 also forms ultra-long time sequence.
The division period determines according to the periodic phenomena that traffic data is presented, and selects friendship of the multiple groups as unit of time window
Logical data time series, are averaging similarity, when average similarity meets given threshold, when which can be used as division
Section.In the present embodiment, one group of historical traffic data collection: { (x is obtained1,t1),(x2,t2),…(xj,tj),…,(x300,
t300)},t2-t1It is 1 hour, time window TK is 12 hours, takes traffic data time series of the multiple groups as unit of TK, is with 3 groups
Example, xTK1={ x1,x2,…,x12},xTK2={ x25,x26,…,x37},xTK3={ x13,x14,…,x24, it can be by 12 small times
It chooses, or randomly selects every sequence.
Similarity can be calculated using the methods of cosine similarity, Pearson correlation coefficients, Jaccard similarity factor.S<
xTK1,xTK2> it is time series xTK1And xTK2Similarity.It is averaging similarity STK=(S < xTK1,xTK2>+S<xTK1,xTK3>+S<
xTK2,xTK3>)/3;Average similarity meets given threshold, and 12 hours time windows, which can be used as, divides the period.
Similar, time window TK is 24 hours, and average similarity meets given threshold, and time window can be used as division within 24 hours
Period.
Wherein it is possible to determine traffic data ultra-long time using the methods of regression model, user setting, neural network model
The corresponding relationship for the data space that individual data position and ultra-long time sequence are formed in sequence, preferably uses in the present embodiment
Regression model, formula are as follows:
Q=∑ (y- ∑ (wpxp+bp))=min
Wherein, y is the traffic data on individual data position, and p indicates data space position, and asking makes Q function minimization
W, b, the corresponding relationship as individual data position and data space.According to corresponding relationship in traffic data ultra-long time sequence
Traffic data make position mark, it is above-mentioned that w, b are obtained using regression model, according to the size of w, different stage is divided, as position
Tagging.
(2) historical traffic data with position mark is chosen as input data, carries out traffic data prediction.
The method of moving average is used in the present embodiment, predicts the traffic data of k+1 weeks Monday of kth 8:00, specific as follows:
Historical traffic data ultra-long time sequence:
1st week
2nd week
x1,1,1Traffic data when for the 1st week Monday 1:00, xkk,nn,mmIt is kth k Zhou Congzhou several n-th n days from 1 point together
Whole several mm hours traffic datas.
Historical traffic data with position mark, formation < xkk,nn,mm,tgkk,nn,mm>tgkk,nn,mmFor xkk,nn,mmThe position of data
Tagging.
It predicts the traffic data of k+1 weeks Monday of kth 1:00, selects tgkk,nn,mmMeet the historical traffic data of threshold value
{xkk,7,24,xkk,7,23,xkk,7,22,xkk,7,1,xkk,6,24,xkk,6,23,xkk-1,7,24..., xkk,7,24,xkk,7,23,xkk,7,22For kth k
First 3 hours of+1 all Mondays 1:00, xkk,7,1,xkk,6,24,xkk,6,23For the time near preceding 1 day 1:00, xkk-1,7,24It is first 2 days
Time near 1:00.Embody different time position logarithm it is predicted that importance.
The operation of average value is done to R historical traffic data, and is successively slided, until all data are processed, finally
An average value is obtained, as the predicted value of traffic data, R is less than the historical traffic data length of time series chosen.
It is specific embodiments of the present invention and the technical principle used described in above, if conception under this invention institute
The change of work when the spirit that generated function is still covered without departing from specification and attached drawing, should belong to of the invention
Protection scope.
Claims (10)
1. a kind of traffic data prediction technique based on ultra-long time sequence, it is characterised in that include the following steps:
(1) position mark processing is carried out to historical traffic data, wherein the method for carrying out position mark processing is as follows:
(1.1) data space that individual data position and ultra-long time sequence are formed in traffic data ultra-long time sequence is determined
Corresponding relationship;
(1.2) position mark is made to the traffic data in traffic data ultra-long time sequence according to corresponding relationship;When the overlength
Between sequence refer to across it is multiple divide the periods time serieses;
(2) historical traffic data with position mark is chosen as input data, carries out traffic data prediction.
2. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 1, it is characterised in that: institute
It states equal using any one method in regression model, user setting method, neural net model method in step (1.1)
It can determine the corresponding pass of individual data position and the data space that ultra-long time sequence is formed in traffic data ultra-long time sequence
System;It is preferred that using regression model, formula is as follows:
Q=∑ (y- ∑ (wpxp+bp))=min
Wherein, y be individual data position on traffic data, p indicate data space position, ask so that the w of Q function minimization,
B, the corresponding relationship as individual data position and data space.
3. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 1, it is characterised in that: institute
It states in step (1.2) and position mark is made to the traffic data in traffic data ultra-long time sequence according to corresponding relationship, wherein adopting
W, b are obtained with regression model, according to the size of w, different stage is divided, as position mark.
4. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 1, it is characterised in that: institute
The division period stated determines according to the periodic phenomena that traffic data is presented, and selects traffic number of the multiple groups as unit of time window
According to time series, it is averaging similarity, when average similarity meets the threshold value of setting, which, which can be used as, divides the period;
Wherein similarity can be calculated by cosine similarity, Pearson correlation coefficients, Jaccard similarity factor method.
5. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 1, it is characterised in that: institute
It states in step (2) using a kind of side arbitrary in neural net model method, the method for moving average, exponential smoothing, AR modelling
Method carries out traffic data prediction.
6. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 1, it is characterised in that: institute
State the data that traffic data includes but is not limited to transit equipment acquisition, the data of the data reduction acquired according to transit equipment;Institute
The data for stating transit equipment acquisition include but is not limited to the magnitude of traffic flow, traffic speed, the data acquired according to transit equipment
The data of conversion, including but not limited to traffic behavior, traffic index.
7. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 5, it is characterised in that: institute
The neural net model method stated preferably uses LSTM long, and neural network carries out traffic data prediction in short-term, specific as follows:
(i) traffic flow forecasting is carried out by the driving sequence of setting input, predicted value and remote traffic flow magnitude is built
Vertical linear regression model (LRM);
(ii) flow value that position mark is met the requirements is filtered out after training, and the distant place flow value that the weight obtained to screening is big
Position is marked;
(iii) the traffic flow magnitude with weight with position mark is added in LSTM network input layer, carries out the anti-of LSTM+
It is trained to propagating, first the output valve of each neuron of forward calculation, secondly the error entry value of each neuron of retrospectively calculate, most
Afterwards according to corresponding error term, the gradient of each weight is calculated, updates weighted value;
(iv) magnitude of traffic flow in the following preset period is predicted with trained LSTM network model.
8. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 7, it is characterised in that: institute
It states in step (i), sets the driving sequence of input as X=(x1,1,1, x1,1,2..., xKk, nn, mm, yKk, nn+1,1,
yKk, nn+1,2..., yKk, nn+1, t-1), wherein and y indicates the data on the prediction same day, and volume forecasting question essence is by driving sequence
X is established and predicted flow rate yKk, nn+1, tMapping.
9. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 7, it is characterised in that: institute
Step (i) is stated when establishing linear regression model (LRM), chooses the data near daily t time interval, regression model is as follows:
Wherein,D is step-length,It is to return mould respectively with b
The weight and biasing of type.
10. a kind of traffic data prediction technique based on ultra-long time sequence according to claim 7, it is characterised in that:
The RNN Recognition with Recurrent Neural Network that it is special that the LSTM network is one kind increases " input gate " compared with RNN, " forgeing door ", " defeated
Go out ", " input gate ", " forgeing door ", " out gate " and LSTM cell state are as follows:
Input gate:
Forget door:
Cell state:
Out gate:
Wherein, I represents the length for inputting current sequence, and C indicates the quantity of LSTM cell, and P indicates labeled high influence wagon flow
The quantity of magnitude, l indicate input state,Indicating the state of forgetting door, o indicates the state of out gate, and f and g are activation primitives,Indicate c-th of t moment cellular state value, tgpIt indicates p-th of labeled wagon flow magnitude for having weight, finally exports are as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811532408.8A CN109598935B (en) | 2018-12-14 | 2018-12-14 | Traffic data prediction method based on ultra-long time sequence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811532408.8A CN109598935B (en) | 2018-12-14 | 2018-12-14 | Traffic data prediction method based on ultra-long time sequence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109598935A true CN109598935A (en) | 2019-04-09 |
CN109598935B CN109598935B (en) | 2020-12-15 |
Family
ID=65962562
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811532408.8A Active CN109598935B (en) | 2018-12-14 | 2018-12-14 | Traffic data prediction method based on ultra-long time sequence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109598935B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009429A (en) * | 2019-04-10 | 2019-07-12 | 金瓜子科技发展(北京)有限公司 | A kind of method, apparatus and computer equipment of predicted flow rate data |
CN110837888A (en) * | 2019-11-13 | 2020-02-25 | 大连理工大学 | Traffic missing data completion method based on bidirectional cyclic neural network |
CN112330442A (en) * | 2020-11-17 | 2021-02-05 | 深圳市欢太科技有限公司 | Modeling method and device based on ultra-long behavior sequence, terminal and storage medium |
CN112508305A (en) * | 2019-12-29 | 2021-03-16 | 山西大学 | Public place entrance pedestrian flow prediction method based on LSTM |
CN112766597A (en) * | 2021-01-29 | 2021-05-07 | 中国科学院自动化研究所 | Bus passenger flow prediction method and system |
CN113112795A (en) * | 2021-04-06 | 2021-07-13 | 中移(上海)信息通信科技有限公司 | Road condition prediction method, device and equipment |
CN114283590A (en) * | 2021-09-02 | 2022-04-05 | 青岛海信网络科技股份有限公司 | Traffic flow peak prediction method and device and electronic equipment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101650876A (en) * | 2009-08-26 | 2010-02-17 | 重庆大学 | Method for obtaining average speed of traffic flow of urban road sections |
CN102034350A (en) * | 2009-09-30 | 2011-04-27 | 北京四通智能交通系统集成有限公司 | Short-time prediction method and system of traffic flow data |
CN104134351A (en) * | 2014-08-14 | 2014-11-05 | 中国科学院自动化研究所 | Short-term traffic flow predicting method |
CN105185106A (en) * | 2015-07-13 | 2015-12-23 | 丁宏飞 | Road traffic flow parameter prediction method based on granular computing |
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 |
-
2018
- 2018-12-14 CN CN201811532408.8A patent/CN109598935B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101650876A (en) * | 2009-08-26 | 2010-02-17 | 重庆大学 | Method for obtaining average speed of traffic flow of urban road sections |
CN102034350A (en) * | 2009-09-30 | 2011-04-27 | 北京四通智能交通系统集成有限公司 | Short-time prediction method and system of traffic flow data |
CN104134351A (en) * | 2014-08-14 | 2014-11-05 | 中国科学院自动化研究所 | Short-term traffic flow predicting method |
CN105185106A (en) * | 2015-07-13 | 2015-12-23 | 丁宏飞 | Road traffic flow parameter prediction method based on granular computing |
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 (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009429A (en) * | 2019-04-10 | 2019-07-12 | 金瓜子科技发展(北京)有限公司 | A kind of method, apparatus and computer equipment of predicted flow rate data |
CN110009429B (en) * | 2019-04-10 | 2021-04-16 | 金瓜子科技发展(北京)有限公司 | Method and device for predicting flow data and computer equipment |
CN110837888A (en) * | 2019-11-13 | 2020-02-25 | 大连理工大学 | Traffic missing data completion method based on bidirectional cyclic neural network |
CN112508305A (en) * | 2019-12-29 | 2021-03-16 | 山西大学 | Public place entrance pedestrian flow prediction method based on LSTM |
CN112330442A (en) * | 2020-11-17 | 2021-02-05 | 深圳市欢太科技有限公司 | Modeling method and device based on ultra-long behavior sequence, terminal and storage medium |
CN112766597A (en) * | 2021-01-29 | 2021-05-07 | 中国科学院自动化研究所 | Bus passenger flow prediction method and system |
CN112766597B (en) * | 2021-01-29 | 2023-06-27 | 中国科学院自动化研究所 | Bus passenger flow prediction method and system |
CN113112795A (en) * | 2021-04-06 | 2021-07-13 | 中移(上海)信息通信科技有限公司 | Road condition prediction method, device and equipment |
CN114283590A (en) * | 2021-09-02 | 2022-04-05 | 青岛海信网络科技股份有限公司 | Traffic flow peak prediction method and device and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN109598935B (en) | 2020-12-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109598935A (en) | A kind of traffic data prediction technique based on ultra-long time sequence | |
CN110070713B (en) | Traffic flow prediction method based on bidirectional nested LSTM neural network | |
CN110570651B (en) | Road network traffic situation prediction method and system based on deep learning | |
CN108564790B (en) | Urban short-term traffic flow prediction method based on traffic flow space-time similarity | |
CN107180530B (en) | A kind of road network trend prediction method based on depth space-time convolution loop network | |
CN107610464B (en) | A kind of trajectory predictions method based on Gaussian Mixture time series models | |
CN106781489B (en) | A kind of road network trend prediction method based on recurrent neural network | |
Khosravi et al. | Prediction intervals to account for uncertainties in travel time prediction | |
CN109934337A (en) | A kind of detection method of the spacecraft telemetry exception based on integrated LSTM | |
Treethidtaphat et al. | Bus arrival time prediction at any distance of bus route using deep neural network model | |
CN107748942B (en) | Radar Echo Extrapolation prediction technique and system based on velocity field sensing network | |
CN109840587A (en) | Reservoir reservoir inflow prediction technique based on deep learning | |
CN108600965B (en) | Passenger flow data prediction method based on guest position information | |
CN104091216A (en) | Traffic information predication method based on fruit fly optimization least-squares support vector machine | |
US20110085649A1 (en) | Fluctuation Monitoring Method that Based on the Mid-Layer Data | |
CN109143408B (en) | Dynamic region combined short-time rainfall forecasting method based on MLP | |
CN106371155A (en) | A weather forecast method and system based on big data and analysis fields | |
CN110110243A (en) | A kind of historical track destination prediction technique based on echo state network | |
CN108898533A (en) | Acquisition methods, device and the computer readable storage medium of movement of population data | |
Arjona et al. | Improving parking availability information using deep learning techniques | |
CN106568445A (en) | Indoor track prediction method based on bidirectional circulation neural network | |
CN109215374A (en) | A kind of bus arrival time prediction algorithm | |
CN107945534A (en) | A kind of special bus method for predicting based on GMDH neutral nets | |
CN109670540A (en) | It is resident number variation tendency Forecasting Approach for Short-term in Passenger Transport Hub region based on kNN algorithm | |
CN113449905A (en) | Traffic jam early warning method based on gated cyclic unit neural network |
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 | ||
CP03 | Change of name, title or address |
Address after: 310012 1st floor, building 1, 223 Yile Road, Hangzhou City, Zhejiang Province Patentee after: Yinjiang Technology Co.,Ltd. Address before: 310023 floor 1, building 1, No. 223, Yile Road, Hangzhou, Zhejiang Patentee before: ENJOYOR Co.,Ltd. |
|
CP03 | Change of name, title or address |