CN103413443A - Short-term traffic flow forecasting method based on hidden Markov model - Google Patents
Short-term traffic flow forecasting method based on hidden Markov model Download PDFInfo
- Publication number
- CN103413443A CN103413443A CN2013102765817A CN201310276581A CN103413443A CN 103413443 A CN103413443 A CN 103413443A CN 2013102765817 A CN2013102765817 A CN 2013102765817A CN 201310276581 A CN201310276581 A CN 201310276581A CN 103413443 A CN103413443 A CN 103413443A
- Authority
- CN
- China
- Prior art keywords
- traffic flow
- hidden markov
- markov model
- hmm
- hidden
- 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
- 238000013277 forecasting method Methods 0.000 title abstract 3
- 238000000034 method Methods 0.000 claims abstract description 33
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000007704 transition Effects 0.000 claims description 16
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 claims description 12
- 238000012546 transfer Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 5
- 230000003068 static effect Effects 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000005309 stochastic process Methods 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000009440 infrastructure construction Methods 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 230000021715 photosynthesis, light harvesting Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
Images
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention relates to the field of the intelligent transportation system, and especially relates to short-term traffic flow forecasting with use of a parameter value sequence of a road segment. A short-term traffic flow forecasting method based on a hidden Markov model comprises the following steps: collected data are processed and counted; a prediction window is set; a starting time measured value of the prediction window, and an average parameter value and a sequence contrast ratio in the prediction window are discretized, so as to form a hidden state and an observation state set of the hidden Markov model; and a Baum Welch algorithm combined with training data are used for learning the model parameter. Finally, for a certain prediction window, a Viterbi algorithm is used to obtain an optimal hidden state sequence based on a known observation state sequence, and a last state of the optimal hidden state sequence is a prediction state. The short-term traffic flow forecasting method based on the hidden Markov model can be used to predict the short-term traffic state in the future, and is an effective method of prediction of the short-term traffic state.
Description
Technical field
The present invention relates to the intelligent transportation system field, be specifically related to utilize the traffic flow parameter value sequence in highway section to predict the short-term traffic flow state, particularly a kind of trend prediction method of short-term traffic flow based on Hidden Markov Model (HMM).
Background technology
Along with the in-depth of the development of the national level of urbanization, and the lasting raising of living standards of the people, automobile has entered in everyone life already, and everyone work, life, study is produced to deep effect.What follow with it is the hysteresis of traffic infrastructure and the poor efficiency of level of service, and the traffic congestion, environmental pollution, the energy dissipation that more seriously therefore cause have caused great economic loss.Therefore, intelligent transportation system (Intelligent Transportation System, ITS) arise at the historic moment, it is on the basis of existing traffic infrastructure and delivery vehicle, apply to whole traffic management system by advanced infotech, mechanics of communication, sensing technology, control technology and computer technology etc. are effectively integrated, thus set up a kind of on a large scale, comprehensive, in real time, multi-transportation and management system accurately and efficiently.ITS is focus and the forward position of TRANSPOWORLD transportation development, is the important symbol of Modern Traffic forwarding business.And, along with deepening continuously of ITS research, become the important component part of national development strategy.
And, along with the raising that people require for transport information, often before travel, just wish to obtain following traffic related information, in order to select suitable trip mode, select optimum traffic path.The short-term traffic flow status predication be exactly for certain highway section future the traffic behavior in short-term predict, the short-term traffic flow status predication is different from traffic forecast when long, the former prediction duration is generally the situation that 5min, 10min, 15min etc. are no more than 1h, main services, to liking the traffic human pilot, completes in advanced traffic information system (ATIS).The latter's prediction duration is the traffic forecast of the macroscopic views such as day, month, year, main services is to liking vehicle supervision department, so that it makes infrastructure construction, bus routes setting, development plan etc., it mainly completes in advanced traffic control system (ATMS).
The prediction of short-term traffic flow accurately in real time is the key that realizes that intelligent transportation is controlled and induced, short time traffic conditions is predicted the suitable trip mode of personnel selection that is conducive to go on a journey, plan rational traffic path, and then reach and shorten running time, reduce the purpose of polluting, relieve traffic congestion, improve the municipal services level, so become the hot subject of intelligent transportation research field.And Chinese scholars has proposed a series of forecast model and method, for the prediction to traffic behavior.
Short-term traffic flow forecasting model and method mainly are divided three classes at present, one class is to take the mathematical model of the traditional mathematics methods such as mathematical statistics and infinitesimal analysis as base growth, for example: the historical method of average (History Average), linear regression model (LRM), Time Series Method (Time-series Model), the comprehensive moving average algorithm of autoregression (ARIMA, Auto-Regression Integrated Moving Average), Kalman filtering method (Kalman filtering), Markov prediction.
Another kind of is the forecast model that utilizes the modern scientific methods such as neural network, fuzzy control to propose for basis, be characterized in the match prediction to traffic flow, but the parameter transplantability is poor.
The 3rd class is the compound Forecasting Methodology of comprehensive distinct methods relative merits, becomes gradually the focus of Recent study.Forecasting Methodology combined with wavelet theory such as the forecast model based on wavelet decomposition theory and Kalman filtering algorithm, time series etc.
Hidden Markov Model (HMM) is a kind of probability model of the stochastic process of the description based on Parametric Representation, wherein comprises Markov chain and stochastic process two parts.Markov chain is described the transfer of state, describes with transition probability; Relation between general random process prescription state and observation sequence, describe with probability of happening.Hidden Markov Model (HMM) can mean with five-tuple, i.e. λ=(N, M, Π, A, B), and parameter declaration is as follows:
(1) N: the number of hidden state in model, establishing state set is S={s
1, s
2..., s
N, when t moment Markov chain state is X
t, X
t∈ S.
(2) M: the number of observer state in model, establish the observation set and be O={o
1, o
2..., o
M, when t moment observation state is O
t, O
t∈ O.
(3) Π: the initial probability distribution of each state in Hidden Markov Model (HMM); Be designated as Π={ π
i, (1≤i≤N) here, π
i=P (X
1=s
i), (1≤i≤N); And 0≤π
i≤ 1,
π
iMean the original state s zero hour
iSelected probability.
(4) A: the state transition probability matrix of Hidden Markov Model (HMM), A=(a
Ij)
N * N, (1≤i, j≤N), and wherein, a
Ij=P{X
T+1=s
j| X
t=s
i, (s
i, s
j∈ S), a
IjExpression state s
iTo s
jThe probability shifted.
(5) B: the generation matrix in Hidden Markov Model (HMM), description be the relation of hidden state and corresponding observation state; B=(b
Ij)
N * M, b wherein
Ij=P{O
t=o
j| X
t=s
i, (1≤i≤N, 1≤j≤M).
Summary of the invention
The object of the present invention is to provide the method for a certain road section traffic volume stream mode of a kind of short-term prediction, solved the problem of prediction short-term traffic flow state, a kind of trend prediction method of short-term traffic flow based on Hidden Markov Model (HMM) is provided.
The present invention adopts following technical scheme to realize:
A kind of trend prediction method of short-term traffic flow based on Hidden Markov Model (HMM), comprise the steps:
(1), determine hidden state set and the observation state set of Hidden Markov Model (HMM), specific as follows:
I, with collection period δ, a certain traffic flow modes parameter by certain highway section xsect is gathered, obtain in the period, take the data sequence of collection period δ as interval corresponding to this parameter detecting;
II, set when fixing segment length as prediction window Φ, i.e. short-term prediction duration, described prediction window Φ is the integral multiple of collection period δ, so prediction window Φ contains the data sequence that Φ/δ a certain traffic flow modes parameter value forms;
Set the transition window Δ, mean that prediction window Φ take the transition window Δ and as unit, slide backward successively transfer on time shaft, described transition window Δ is the integral multiple of collection period δ, and scope is δ≤Δ≤Φ; Determine accordingly in the quantity that detects prediction window Φ in the period;
Utilize gray scale united symbiosis Matrix C and contrast, determine the contrast C ON of data sequence in each prediction window Φ; As follows: the element c in gray scale united symbiosis Matrix C
IjThe intensity level (traffic flow parameter value) that means data point is the frequency that the such data combination of j occurs for the intensity level of i and adjacent data point thereof, namely
CON=∑
i,j| j-i| (j-i) c
Ij
Initial time parameter value θ in each prediction window Φ
tAs observed value, all observed values form observed value sequence O={O
1, O
2..., O
T.
The variation range of III, all observed values of statistics, according to statistics, carry out discrete M the interval that turn to by the variation range of observed value, obtains simultaneously corresponding to interval grade, is about to the discrete M of the turning to level of all observed values, sets grade and be observation state o
i(i=1,2 ..., M), thereby obtain observation state set O={o
1, o
2..., o
M;
In like manner, the mean parameter in statistical forecast window Φ
With the variation range of contrast C ON, according to statistics, by mean parameter
With contrast C ON, carry out respectively discrete turn to m and n interval, obtain simultaneously corresponding to interval grade, be about to mean parameter
The discrete m of turning to level, the discrete n level that turns to of contrast C ON, the recycling mean parameter
With the gradational two-dimentional total divisor of contrast C ON, combining the hidden state of description is exactly m * n, thereby obtains hidden state set S={s
1, s
2..., s
N, N=m * n.
(2), after the hidden state set of having determined Hidden Markov Model (HMM) and observation state set, Hidden Markov Model (HMM) is trained, obtain being suitable for the Hidden Markov Model (HMM) of traffic flow
Specific as follows:
At first utilize random assignment to carry out initialization to the Hidden Markov Model (HMM) parameter, obtain Hidden Markov initialization model λ
Initial=(Π, A, B), according to λ
Initial=(Π, A, B) and known observed value sequence O={O
1, O
2..., O
T, utilize Hidden Markov revaluation formula iteration to obtain new Hidden Markov Model (HMM)
Can prove
The revaluation process is continued to iteration until
Convergence, now
Be the required Hidden Markov Model (HMM) that is suitable for traffic flow
(3), in the given Hidden Markov Model (HMM) that is suitable for traffic flow modes
On the basis of observed value sequence, utilize the Viterbi algorithm to try to achieve the hidden status switch of the optimum corresponding with the observed value sequence, the final state in optimum hidden status switch is the traffic flow modes of predicting detected after the period; And pass and carry out the short-term traffic flow status predication successively.Because observed value is corresponding with the initial time measured value of each prediction window, hidden state is corresponding with the traffic flow modes in prediction window, and predicted time length is exactly the length of prediction window.
During work, the present invention take prediction highway section future in short-term the traffic behavior in (5min or 10min) be purpose, according to the traffic flow parameter data sequence gathered from the highway section data station, be basis, by it is analyzed, further understand the Variation Features of traffic flow on time shaft, as can be known, change, nonlinear stochastic process when traffic flow is a kind of.Yet the Classical forecast model carries out quantitative deterministic prediction to traffic behavior often, only the traffic flow static information is studied, and has ignored the variation tendency of traffic flow.Based on this, utilize Hidden Markov (HMM) statistical model to predict traffic behavior.
At first, the Hidden Markov Model (HMM) parameter is constructed, the data sequence in prediction window is carried out to statistical study, obtain its mean value, and data sequence contrast in prediction window, thereby the hidden status switch of model obtained.And interval corresponding to the measured value of setting prediction window zero hour is as observation state (value) sequence.Thereby hidden state set and observation state set have been constructed.
Secondly, utilize the EM algorithm in HMM, utilize the training data gathered to train the parameter of model, obtain model parameter, comprise state-transition matrix, matrix, original state probability distribution occur, and in conjunction with the characteristics of traffic flow, parameter is analyzed.
Then, on the model basis obtained, utilize new Hidden Markov Model (HMM) to predict short time traffic conditions.
The present invention processes and adds up the data that gather, by setting prediction window, to prediction window initial time measured value and prediction window intrinsic parameter mean value and sequence contrast discretize, form hidden state and the observation state set of Hidden Markov Model (HMM), then utilize Baum-Welch algorithm combined training data to learn model parameter.Finally, for certain prediction window, on the basis of known observation state sequence, utilize the Viterbi algorithm to try to achieve optimum hidden status switch, the last state of optimum hidden status switch is predicted state.
The present invention is reasonable in design, the traffic behavior prediction for city expressway future in short-term, and help the to go on a journey rational trip mode of personnel selection or optimal path, be a kind of effective short time traffic conditions Forecasting Methodology.
The accompanying drawing explanation
Fig. 1 is the prediction process flow diagram of short time traffic conditions.
Fig. 2 is prediction window Φ and transition window Δ schematic diagram.
Fig. 3 is gray scale united symbiosis matrix schematic diagram.
Fig. 4 is Viterbi prediction process flow diagram.
Embodiment
Below in conjunction with accompanying drawing, specific embodiments of the invention are elaborated.
In the present embodiment, the Hidden Markov Model (HMM) of determining based on traffic flow modes is called " traffic Hidden Markov Model (HMM) ", notes by abridging as " THMM ".Setup parameter collection period δ is 30s, in 24h, contains 2880 groups of data in one day, usings and amounts to the data of 1001 groups of morning peak data sequences as the present embodiment from 500 to 1500 of every day.For the ease of the structure to THMM, define three time window: prediction window Φ, transition window Δ, acquisition window (cycle) δ.
Definition 1: prediction window Φ, purpose is predict future traffic behavior in short-term, sets the time span of 5min as prediction window, namely be exactly 5min the future " in short-term " of prediction.
Definition 2: the transition window Δ, refer to that by the moment t transfer has occurred traffic behavior in the time window of t+ (1* Δ), for example set 3 kinds of transition window, i.e. 30s/2min/5min.
Definition 3: acquisition window δ: because data sequence take 30s and be sampling interval, so system take 30s and rolls successively as 1 step, and setting 30s is acquisition window δ, fixes in the present embodiment.
Utilize Hidden Markov Model (HMM) (HMM) to study traffic flow, at first need hidden state and the observation state of model are determined.For the prediction window Φ of 5min, set state that the parameter measured value of prediction window Φ zero hour the is corresponding observation state (observed value) as THMM; Because the hidden state of model will mean Static and dynamic information two aspects of traffic flow simultaneously, so that hidden state is combined by two factors is definite, namely utilize the mean parameter of the argument sequence in time window
ON combines statement with contrast C.Be illustrated in figure 2 the schematic diagram of THMM model hidden state and observed value on time shaft.
Because sampling interval is 30s, the 500-1500 that gets the parameter value sequence morning peak period in the present embodiment amounts to 1001 groups of data and studies.Prediction window Φ length is 5min, wherein contains the data point of 10 30s, and owing to setting 30 seconds as the transition window Δ, therefore, morning peak has contained 992 prediction window in the period, between window, have lap.Wherein, S
t(1≤t≤992) mean the hidden state of traffic flow in each 5min prediction window Φ.In order to utilize model to predict the state in prediction window, set the parameter value θ of the initial time of each prediction window Φ
tCorresponding observation state is O
t, (1≤t≤992), and formed observation state (value) sequence O=(O
1, O
2..., O
992).
Below a kind of trend prediction method of short-term traffic flow based on Hidden Markov Model (HMM) is described in detail, comprises the steps:
(1), determine hidden state set and the observation state set of Hidden Markov Model (HMM), specific as follows:
I, with collection period δ (30s), a certain traffic flow modes parameter by certain highway section xsect (for example any one in traffic flow speed, vehicle flowrate, occupation rate) is gathered, obtain in the period, take the data sequence of collection period δ as interval corresponding to this parameter detecting.
II, set when fixing segment length as prediction window Φ (5min), it is the short-term prediction duration, described prediction window Φ is the integral multiple of collection period δ, so prediction window Φ contains the data sequence that the individual a certain traffic flow modes parameter value of Φ/δ (Φ/δ=10) forms.
Set the transition window Δ, mean that prediction window Φ take the transition window Δ and as unit, slide backward successively transfer on time shaft, described transition window Δ is the integral multiple of collection period δ, and scope is δ≤Δ≤Φ; Determine accordingly 992 of the quantity that detects prediction window Φ in the period (500min).
Initial time parameter value θ in each prediction window Φ
tAs observed value, all observed values form observed value sequence O={O
1, O
2..., O
T.
Utilize gray scale united symbiosis Matrix C, determine the contrast C ON of data sequence in each prediction window Φ.Specific as follows:
Utilize the mean parameter of a certain traffic flow parameter within certain period (having detected period 500min) of prediction window
(first-order statistics variable) and the characterising parameter situation contrast C ON(second-order statistic that fluctuates on time shaft) combining of the two traffic behavior is described.Wherein, mean parameter can represent intuitively to the current or historical static information of traffic flow, variation tendency and degree of fluctuation that contrast can the characterising parameter sequence.
To the variation tendency on time shaft analyze traffic flow parameter, from the angle of statistics, just must carry out the second-order statistics analysis to argument sequence.
The present invention has expanded the gray scale united symbiosis matrix (GLCM) of widespread use in the picture research field and the definition of contrast (Contrast, CON).In to graphical analysis, the brightness of gray level expressing pixel and intensity, and for the traffic flow parameter sequence, the traffic parameter value just is equivalent to gray-scale value.The acquisition interval of the data acquisition in the present embodiment is 30s, the data sequence that 10 parameter values in 5min form is analyzed, and then generated gray scale united symbiosis matrix, thereby try to achieve contrast C ON.Given one group of data sequence, the element c in gray scale united symbiosis matrix (GLCM) C
IjThe intensity level (traffic flow parameter value) that means data point is the frequency that the such data combination of j occurs for the intensity level of i and adjacent data point thereof.Be element c
IjThe intensity level (traffic flow parameter value) that means data point in sequence is the number of times that the combination of j composition data occurs for the intensity level of i and adjacent data point thereof, the ratio of the sum of all possible data combination in the data sequence formed with 10 parameter values, as the formula (1).Herein, gray scale united symbiosis Matrix C is a N
g* N
gMatrix, N wherein
gThe number that means gray level.
In image, weigh the correlative value that pixel is adjacent the intensity of pixel, and the amount of image local grey scale change is called contrast C ON, its expression formula is as follows:
CON=∑
i,j|i-j|
2c
ij (2)
In the present invention, following formula (2) is improved, as follows:
CON=∑
i,j|j-i|(j-i)c
ij (3)
As can be known by formula (3), in the traffic sequence, contrast C ON means parameter positive and negative variation in time, can reflect the variation tendency of traffic parameter.Its absolute value | the size of CON| has meaned the severe degree of velocity variations trend.
The one group of data sequence of take now is example, intuitively gray scale united symbiosis matrix (GLCM) and contrast (Contrast, CON) is described.If 10 data values in the prediction window of 5min have formed one group of sequence: O=(10,10,7,10,8,9,10,10,8,8), according to formula (1), can try to achieve the united symbiosis Matrix C, as shown in Figure 3.In Matrix C, element c
10,10=0.22 because in data sequence O, the combination of (10,10) has occurred 2 times, and sequence always total number of combinations be 9, therefore its ratio 2/9=0.22.
As can be known by formula (3), contrast C ON has positive and negative variation.In the traffic sequence, contrast C ON just can mean parameter positive and negative variation in time, for example has the trend of rising when speed, contrast C ON be just on the occasion of, otherwise for bearing, and its absolute value | the size of CON| has meaned the severe degree of velocity variations trend.The contrast C ON of data sequence in above-mentioned 5min:
CON=0.11×(10-7)
2+0.11×(9-8)
2+0.11×(10-9)
2
-0.11×(10-7)
2-0.22×(10-8)
2=0.67。
The variation range of III, all observed values of statistics, according to statistics, carry out discrete M the interval that turn to by the variation range of observed value, obtains simultaneously corresponding to interval grade, is about to the discrete M of the turning to level of all observed values, sets grade and be observation state o
i(i=1,2 ..., M), obtain observation state set O={o
1, o
2..., o
M.
In the present embodiment, the parametric speed of take is example, and the velocity variations interval is approximately (0mph, 60mph), and the speed observer value is divided for 11 grades, as shown in table 1.
Table 1 speed observer sate discretization
In like manner, the mean parameter in statistical forecast window Φ
With the variation range of contrast C ON, according to statistics, by mean parameter
With contrast C ON, carry out respectively discrete turn to m and n interval, obtain simultaneously corresponding to interval grade, be about to mean parameter
The discrete m of turning to level, the discrete n level that turns to of contrast C ON.
When the hidden state of structure THMM model, speed average in prediction window is carried out discretely turning to 6 grades, similar with classic method, the friction speed level is corresponding to the different parameters interval of traffic flow.It is as shown in table 2,
Table 2 average velocity discretize
As can be known by the average velocity discretize table in prediction window, the friction speed descriptive grade be the static information of traffic flow modes, can explain intuitively the traffic capacity of current traffic flow.
In the present embodiment, the variation range of speed contrast C ON is mainly (70,70).For the ease of the structure to the hidden state of traffic Hidden Markov Model (HMM), similar with the discretize of average velocity in prediction window, turn to 7 grades by contrast C ON is discrete, 7 grades correspond respectively to contrast from negative value on the occasion of 7 intervals, as shown in table 3.
Table 3 speed contrast discretize
When structure traffic flow Hidden Markov Model (HMM), by serial mean in prediction window
Discrete some grades that represent the friction speed interval and the window intrinsic parameter sequence contrast C ON discretize grade of turning to is respectively as two factors of hidden state.Utilize contrast C ON and data sequence mean value
The hidden state set S={s of two-dimentional total divisor associating descriptive model of grade after discretize
1, s
2..., s
N.This statement of combining to hidden state, overcome classic method and only described the more unilateral shortcoming of traffic behavior static information.
For example,, to the mean parameter in prediction window
And contrast C ON discrete m level and the n level of turning to respectively, the hidden state of its describable traffic flow is exactly m * n.The present embodiment take " speed " be example, to speed average in prediction window and contrast discrete respectively be 6 grades and 7 grades, so the hidden state of model just has 6 * 7=42 kind, hidden state set is S={s
1, s
2..., s
42.
For easy, hidden state is meaned with arabic numeral, as shown in table 3.Thereby, the represented traffic of the data sequence in prediction window (5min) (containing 10 detected values in 5min) can be made as to certain state.
The hidden state of table 4 forms table
Annotate: in table 4, some states are the road conditions that often run in the process of moving, and for example state 17 is exactly crowdedly to cause the situation that traffic flow speed is less peak period.State 39 means the coast is clear, and road conditions are good, the less situation of impact between vehicle.Yet some states are rarely found, for example state 1, and speed is very low, continues the possibility descended very little.However, contain all states in order to make model, will it count.
(2), after the hidden state set of having determined Hidden Markov Model (HMM) and observation state set, Hidden Markov Model (HMM) is trained, obtain being suitable for the Hidden Markov Model (HMM) of traffic flow
Specific as follows:
At first utilize random assignment to carry out initialization to the Hidden Markov Model (HMM) parameter, obtained Hidden Markov initialization model λ
Initial=(Π, A, B), according to λ
Initial=(Π, A, B) and known observed value sequence O={O
1, O
2..., O
T, utilize Hidden Markov revaluation formula iteration to obtain new Hidden Markov Model (HMM)
Can prove
The revaluation process is continued to iteration until
Convergence, now
Be the required Hidden Markov Model (HMM) that is suitable for traffic flow
(3), in given being applicable to, exchange on the basis of the Hidden Markov Model (HMM) of stream and observed value sequence, utilize the Viterbi algorithm to try to achieve the hidden status switch of the optimum corresponding with the observed value sequence, the final state in hidden status switch is the traffic flow modes of predicting detected after the period; And pass and carry out the short-term traffic flow status predication successively.Because observed value is corresponding with the initial time measured value of each prediction window, hidden state is corresponding with the traffic flow modes in prediction window, and predicted time length is exactly the length of prediction window.
At first, in given Hidden Markov Model (HMM)
And observed value sequence O=(O
1, O
2..., O
t) basis on, utilize the Viterbi algorithm to try to achieve the (S with the hidden status switch S=of the optimum of observation sequence
1, S
2..., S
t).
Secondly, utilize Viterbi variable δ
t(i) and the memory variable
By the Viterbi algorithm iteration, obtain the hidden status switch S corresponding to the O optimum.
Because set the measured value of prediction window initial time corresponding to observation state O
t, and the last hidden state S in the corresponding hidden status switch of optimum
tBe in prediction window in the description of traffic behavior, so state S of last moment in the status switch S of Optimum Matching
tBe the state of prediction, and pass and carry out the short-term traffic flow status predication successively.
Claims (5)
1. the trend prediction method of the short-term traffic flow based on Hidden Markov Model (HMM), is characterized in that: comprise the steps:
(1), determine hidden state set and the observation state set of Hidden Markov Model (HMM), specific as follows:
I, with collection period δ, a certain traffic flow modes parameter by certain highway section xsect is gathered, obtain in the period, take the data sequence of collection period δ as interval corresponding to this parameter detecting;
II, set when fixing segment length as prediction window Φ, i.e. short-term prediction duration, described prediction window Φ is the integral multiple of collection period δ, so prediction window Φ contains the data sequence that Φ/δ a certain traffic flow modes parameter value forms;
Set the transition window Δ, mean that prediction window Φ take the transition window Δ and as unit, slide backward successively transfer on time shaft, described transition window Δ is the integral multiple of collection period δ, and scope is δ≤Δ≤Φ; Determine accordingly in the quantity that detects prediction window Φ in the period;
Utilize gray scale united symbiosis Matrix C, determine the contrast C ON of data sequence in each prediction window Φ; As follows: the element c in gray scale united symbiosis Matrix C
IjThe intensity level that means data point is that the intensity level of i and adjacent data point thereof is the frequency that the such data combination of j occurs, namely
Initial time parameter value θ in each prediction window Φ
tAs observed value, all observed values form observed value sequence O={O
1, O
2..., O
T;
The variation range of III, all observed values of statistics, according to statistics, carry out discrete M the interval that turn to by the variation range of observed value, obtains simultaneously corresponding to interval grade, is about to the discrete M of the turning to level of all observed values, sets grade and be observation state o
i(i=1,2 ..., M), obtain observation state set O={o
1, o
2..., o
M;
In like manner, the mean parameter in statistical forecast window Φ
With the variation range of contrast C ON, according to statistics, by mean parameter
With contrast C ON, carry out respectively discrete turn to m and n interval, obtain simultaneously corresponding to interval grade, be about to mean parameter
The discrete m of turning to level, the discrete n level that turns to of contrast C ON, utilize mean parameter
Combine with the two-dimentional total divisor of the grade of contrast C ON that to describe hidden state be exactly m * n, obtain hidden state set S={s
1, s
2..., s
N, N=m * n;
(2), after the hidden state set of having determined Hidden Markov Model (HMM) and observation state set, Hidden Markov Model (HMM) is trained, obtain being suitable for the Hidden Markov Model (HMM) of traffic flow
Specific as follows:
At first utilize random assignment to carry out initialization to the Hidden Markov Model (HMM) parameter, obtained Hidden Markov initialization model λ
Initial=(Π, A, B), according to λ
Initial=(Π, A, B) and known observed value sequence O={O
1, O
2..., O
T, utilize Hidden Markov revaluation formula iteration to obtain new Hidden Markov Model (HMM)
Can prove
The revaluation process is continued to iteration until
Convergence, now
Be the required Hidden Markov Model (HMM) that is suitable for traffic flow
(3), on the basis of the given Hidden Markov Model (HMM) that is applicable to traffic flow and observed value sequence, utilize the Viterbi algorithm to try to achieve the hidden status switch of the optimum corresponding with the observed value sequence, the final state in hidden status switch is the traffic flow modes of predicting detected after the period, and passes and carry out the short-term traffic flow status predication successively.
2. the trend prediction method of the short-term traffic flow based on Hidden Markov Model (HMM) according to claim 1, it is characterized in that: the traffic flow modes parameter of described collection is traffic flow speed or vehicle flowrate or occupation rate.
3. the trend prediction method of the short-term traffic flow based on Hidden Markov Model (HMM) according to claim 1 and 2, it is characterized in that: the duration that detects the period in described step I is 500min or 1440min.
4. the trend prediction method of the short-term traffic flow based on Hidden Markov Model (HMM) according to claim 3, it is characterized in that: described collection period is that δ is 30s or 1min.
5. the trend prediction method of the short-term traffic flow based on Hidden Markov Model (HMM) according to claim 4, it is characterized in that: the duration of described prediction window Φ is 5min or 10min.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310276581.7A CN103413443B (en) | 2013-07-03 | 2013-07-03 | Short-term traffic flow forecasting method based on hidden Markov model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310276581.7A CN103413443B (en) | 2013-07-03 | 2013-07-03 | Short-term traffic flow forecasting method based on hidden Markov model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103413443A true CN103413443A (en) | 2013-11-27 |
CN103413443B CN103413443B (en) | 2015-05-20 |
Family
ID=49606448
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310276581.7A Expired - Fee Related CN103413443B (en) | 2013-07-03 | 2013-07-03 | Short-term traffic flow forecasting method based on hidden Markov model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103413443B (en) |
Cited By (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103745602A (en) * | 2014-01-16 | 2014-04-23 | 银江股份有限公司 | Traffic flow predicating method based on sliding window average |
CN104021665A (en) * | 2014-06-06 | 2014-09-03 | 河南理工大学 | Gene search short time traffic flow forecasting method |
CN104408203A (en) * | 2014-12-18 | 2015-03-11 | 西安电子科技大学宁波信息技术研究院 | Method for predicting path destination of moving object |
CN104464291A (en) * | 2014-12-08 | 2015-03-25 | 杭州智诚惠通科技有限公司 | Traffic flow predicting method and system |
CN104616498A (en) * | 2015-02-02 | 2015-05-13 | 同济大学 | Markov chain and neural network based traffic congestion state combined prediction method |
CN105303835A (en) * | 2015-11-13 | 2016-02-03 | 西安邮电大学 | Short-time prediction method of road traffic flow state |
CN106097726A (en) * | 2016-08-23 | 2016-11-09 | 苏州科达科技股份有限公司 | The detection determination in region, traffic information detection method and device |
CN106205213A (en) * | 2014-12-30 | 2016-12-07 | 江苏理工学院 | Ship track prediction method |
CN106373410A (en) * | 2016-09-21 | 2017-02-01 | 青岛大学 | City traffic signal optimization control method |
CN106530715A (en) * | 2016-12-24 | 2017-03-22 | 浙江工业大学 | Road network traffic state prediction method based on fuzzy Markov process |
CN106649122A (en) * | 2016-12-28 | 2017-05-10 | Tcl集团股份有限公司 | Model construction method and device for terminal application |
CN107437339A (en) * | 2017-06-20 | 2017-12-05 | 北京交通大学 | Variable information advices plate control method for coordinating and system under a kind of information guidance |
CN108629457A (en) * | 2018-05-09 | 2018-10-09 | 西南交通大学 | The method and apparatus predicted trip mode and build prediction model |
US10140854B2 (en) | 2017-04-03 | 2018-11-27 | Here Global B.V. | Vehicle traffic state determination |
CN109147325A (en) * | 2018-09-04 | 2019-01-04 | 广州视源电子科技股份有限公司 | Road condition prediction method and device, storage medium and processor |
CN109255492A (en) * | 2015-03-31 | 2019-01-22 | 江苏理工学院 | Subway track real-time prediction method based on robust strategy |
CN109344195A (en) * | 2018-10-25 | 2019-02-15 | 电子科技大学 | Pipe safety event recognition and Knowledge Discovery Method based on HMM model |
CN109637128A (en) * | 2018-12-14 | 2019-04-16 | 南通大学 | A kind of grey Verhulst Short-time Traffic Flow Forecasting Methods and system based on Markov |
CN109981358A (en) * | 2019-03-13 | 2019-07-05 | 南京理工大学 | A kind of adaptive network performance method for early warning based on built-up pattern |
CN110020475A (en) * | 2019-04-03 | 2019-07-16 | 北京工业大学 | A kind of Markov particle filter method of forecasting traffic flow |
CN110084112A (en) * | 2019-03-20 | 2019-08-02 | 太原理工大学 | A kind of traffic congestion judgment method based on image procossing |
CN110085026A (en) * | 2019-03-28 | 2019-08-02 | 中国公路工程咨询集团有限公司 | A kind of traffic status prediction method based on clustering and Markov model |
CN110415521A (en) * | 2019-07-31 | 2019-11-05 | 京东城市(北京)数字科技有限公司 | Prediction technique, device and the computer readable storage medium of traffic data |
CN111444302A (en) * | 2020-04-17 | 2020-07-24 | 中国传媒大学 | Mobility prediction method, system and device based on user classification |
CN111951553A (en) * | 2020-08-17 | 2020-11-17 | 上海电科智能系统股份有限公司 | Prediction method based on traffic big data platform and mesoscopic simulation model |
CN113380026A (en) * | 2021-05-28 | 2021-09-10 | 南京理工大学 | HMM-based highway traffic prediction method |
CN113762644A (en) * | 2021-09-26 | 2021-12-07 | 中国联合网络通信集团有限公司 | Congestion state prediction method and device based on Markov chain |
CN114282169A (en) * | 2021-10-12 | 2022-04-05 | 腾讯科技(深圳)有限公司 | Abnormal data detection method and related device |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1707544A (en) * | 2005-05-26 | 2005-12-14 | 上海交通大学 | Method for estimating city road network traffic flow state |
CN1776739A (en) * | 2004-11-16 | 2006-05-24 | 微软公司 | Traffic predictiong using model-setting and analuzing to probability relativity and environment data |
US20070208492A1 (en) * | 2006-03-03 | 2007-09-06 | Inrix, Inc. | Dynamic time series prediction of future traffic conditions |
CN101188002A (en) * | 2007-12-24 | 2008-05-28 | 北京大学 | A city traffic dynamic prediction system and method with real time and continuous feature |
US20090002195A1 (en) * | 2007-06-29 | 2009-01-01 | Microsoft Corporation | Sensing and predicting flow variance in a traffic system for traffic routing and sensing |
US20120072096A1 (en) * | 2006-03-03 | 2012-03-22 | Chapman Craig H | Dynamic prediction of road traffic conditions |
CN102695637A (en) * | 2009-08-31 | 2012-09-26 | 丰田自动车欧洲股份有限公司 | Vehicle or traffic control method and system |
-
2013
- 2013-07-03 CN CN201310276581.7A patent/CN103413443B/en not_active Expired - Fee Related
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1776739A (en) * | 2004-11-16 | 2006-05-24 | 微软公司 | Traffic predictiong using model-setting and analuzing to probability relativity and environment data |
CN1707544A (en) * | 2005-05-26 | 2005-12-14 | 上海交通大学 | Method for estimating city road network traffic flow state |
US20070208492A1 (en) * | 2006-03-03 | 2007-09-06 | Inrix, Inc. | Dynamic time series prediction of future traffic conditions |
US20120072096A1 (en) * | 2006-03-03 | 2012-03-22 | Chapman Craig H | Dynamic prediction of road traffic conditions |
US20090002195A1 (en) * | 2007-06-29 | 2009-01-01 | Microsoft Corporation | Sensing and predicting flow variance in a traffic system for traffic routing and sensing |
CN101188002A (en) * | 2007-12-24 | 2008-05-28 | 北京大学 | A city traffic dynamic prediction system and method with real time and continuous feature |
CN102695637A (en) * | 2009-08-31 | 2012-09-26 | 丰田自动车欧洲股份有限公司 | Vehicle or traffic control method and system |
Non-Patent Citations (2)
Title |
---|
丁恒,等: "基于小波分析的随机交通流组合预测方法研究", 《系统仿真学报》, vol. 24, no. 2, 8 February 2012 (2012-02-08), pages 377 - 381 * |
丁栋,等: "基于影响模型的短时交通流预测方法", 《计算机工程》, vol. 38, no. 10, 20 May 2012 (2012-05-20), pages 164 - 167 * |
Cited By (44)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103745602B (en) * | 2014-01-16 | 2016-04-13 | 银江股份有限公司 | A kind of traffic flow forecasting method average based on sliding window |
CN103745602A (en) * | 2014-01-16 | 2014-04-23 | 银江股份有限公司 | Traffic flow predicating method based on sliding window average |
CN104021665A (en) * | 2014-06-06 | 2014-09-03 | 河南理工大学 | Gene search short time traffic flow forecasting method |
CN104021665B (en) * | 2014-06-06 | 2016-09-07 | 河南理工大学 | Genetic search Short-time Traffic Flow Forecasting Methods |
CN104464291A (en) * | 2014-12-08 | 2015-03-25 | 杭州智诚惠通科技有限公司 | Traffic flow predicting method and system |
CN104408203A (en) * | 2014-12-18 | 2015-03-11 | 西安电子科技大学宁波信息技术研究院 | Method for predicting path destination of moving object |
CN106205213A (en) * | 2014-12-30 | 2016-12-07 | 江苏理工学院 | Ship track prediction method |
CN106251704A (en) * | 2014-12-30 | 2016-12-21 | 江苏理工学院 | Ship track prediction method |
CN104616498A (en) * | 2015-02-02 | 2015-05-13 | 同济大学 | Markov chain and neural network based traffic congestion state combined prediction method |
CN109447327A (en) * | 2015-03-31 | 2019-03-08 | 江苏理工学院 | Subway train track prediction method |
CN109255492A (en) * | 2015-03-31 | 2019-01-22 | 江苏理工学院 | Subway track real-time prediction method based on robust strategy |
CN105303835A (en) * | 2015-11-13 | 2016-02-03 | 西安邮电大学 | Short-time prediction method of road traffic flow state |
CN106097726A (en) * | 2016-08-23 | 2016-11-09 | 苏州科达科技股份有限公司 | The detection determination in region, traffic information detection method and device |
CN106373410B (en) * | 2016-09-21 | 2018-12-21 | 青岛大学 | A kind of Optimal Method of Urban Traffic Signal Control |
CN106373410A (en) * | 2016-09-21 | 2017-02-01 | 青岛大学 | City traffic signal optimization control method |
CN106530715A (en) * | 2016-12-24 | 2017-03-22 | 浙江工业大学 | Road network traffic state prediction method based on fuzzy Markov process |
CN106530715B (en) * | 2016-12-24 | 2019-05-31 | 浙江工业大学 | Road network traffic state prediction method based on fuzzy Markov process |
CN106649122A (en) * | 2016-12-28 | 2017-05-10 | Tcl集团股份有限公司 | Model construction method and device for terminal application |
US10140854B2 (en) | 2017-04-03 | 2018-11-27 | Here Global B.V. | Vehicle traffic state determination |
CN107437339A (en) * | 2017-06-20 | 2017-12-05 | 北京交通大学 | Variable information advices plate control method for coordinating and system under a kind of information guidance |
CN108629457A (en) * | 2018-05-09 | 2018-10-09 | 西南交通大学 | The method and apparatus predicted trip mode and build prediction model |
CN108629457B (en) * | 2018-05-09 | 2021-09-28 | 西南交通大学 | Travel prediction mode and method and device for building prediction model |
CN109147325A (en) * | 2018-09-04 | 2019-01-04 | 广州视源电子科技股份有限公司 | Road condition prediction method and device, storage medium and processor |
CN109147325B (en) * | 2018-09-04 | 2022-01-28 | 广州视源电子科技股份有限公司 | Road condition prediction method and device, storage medium and processor |
CN109344195B (en) * | 2018-10-25 | 2021-09-21 | 电子科技大学 | HMM model-based pipeline security event recognition and knowledge mining method |
CN109344195A (en) * | 2018-10-25 | 2019-02-15 | 电子科技大学 | Pipe safety event recognition and Knowledge Discovery Method based on HMM model |
CN109637128B (en) * | 2018-12-14 | 2021-06-18 | 南通大学 | Markov-based gray Verhulst short-time traffic flow prediction method and system |
CN109637128A (en) * | 2018-12-14 | 2019-04-16 | 南通大学 | A kind of grey Verhulst Short-time Traffic Flow Forecasting Methods and system based on Markov |
CN109981358A (en) * | 2019-03-13 | 2019-07-05 | 南京理工大学 | A kind of adaptive network performance method for early warning based on built-up pattern |
CN110084112A (en) * | 2019-03-20 | 2019-08-02 | 太原理工大学 | A kind of traffic congestion judgment method based on image procossing |
CN110084112B (en) * | 2019-03-20 | 2022-09-20 | 太原理工大学 | Traffic jam judging method based on image processing |
CN110085026A (en) * | 2019-03-28 | 2019-08-02 | 中国公路工程咨询集团有限公司 | A kind of traffic status prediction method based on clustering and Markov model |
CN110020475A (en) * | 2019-04-03 | 2019-07-16 | 北京工业大学 | A kind of Markov particle filter method of forecasting traffic flow |
CN110020475B (en) * | 2019-04-03 | 2023-10-10 | 北京工业大学 | Markov particle filtering method for traffic flow prediction |
CN110415521A (en) * | 2019-07-31 | 2019-11-05 | 京东城市(北京)数字科技有限公司 | Prediction technique, device and the computer readable storage medium of traffic data |
CN111444302A (en) * | 2020-04-17 | 2020-07-24 | 中国传媒大学 | Mobility prediction method, system and device based on user classification |
CN111951553B (en) * | 2020-08-17 | 2022-11-11 | 上海电科智能系统股份有限公司 | Prediction method based on traffic big data platform and mesoscopic simulation model |
CN111951553A (en) * | 2020-08-17 | 2020-11-17 | 上海电科智能系统股份有限公司 | Prediction method based on traffic big data platform and mesoscopic simulation model |
CN113380026B (en) * | 2021-05-28 | 2022-07-22 | 南京理工大学 | HMM-based highway traffic prediction method |
CN113380026A (en) * | 2021-05-28 | 2021-09-10 | 南京理工大学 | HMM-based highway traffic prediction method |
CN113762644A (en) * | 2021-09-26 | 2021-12-07 | 中国联合网络通信集团有限公司 | Congestion state prediction method and device based on Markov chain |
CN113762644B (en) * | 2021-09-26 | 2023-11-24 | 中国联合网络通信集团有限公司 | Congestion state prediction method and device based on Markov chain |
CN114282169A (en) * | 2021-10-12 | 2022-04-05 | 腾讯科技(深圳)有限公司 | Abnormal data detection method and related device |
CN114282169B (en) * | 2021-10-12 | 2024-07-12 | 腾讯科技(深圳)有限公司 | Abnormal data detection method and related device |
Also Published As
Publication number | Publication date |
---|---|
CN103413443B (en) | 2015-05-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103413443B (en) | Short-term traffic flow forecasting method based on hidden Markov model | |
CN109035761B (en) | Travel time estimation method based on auxiliary supervised learning | |
CN110570651B (en) | Road network traffic situation prediction method and system based on deep learning | |
CN105185115B (en) | A kind of traffic forecast method and forecasting system | |
CN102110365B (en) | Road condition prediction method and road condition prediction system based on space-time relationship | |
Mazloumi et al. | Prediction intervals to account for uncertainties in neural network predictions: Methodology and application in bus travel time prediction | |
CN107085943B (en) | Short-term prediction method and system for road travel time | |
CN105303882A (en) | Parking management system and method based on space-time dimension dynamic parking charging strategy | |
Olayode et al. | Prediction and modeling of traffic flow of human-driven vehicles at a signalized road intersection using artificial neural network model: A South African road transportation system scenario | |
Theja et al. | Short term prediction of traffic parameters using support vector machines technique | |
CN105550789A (en) | Method for predicting bus taking passenger flow | |
CN102346964A (en) | Real-time jam prediction and intelligent management system for road traffic network area | |
Olayode et al. | PREDICTION OF VEHICULAR TRAFFIC FLOW USING LEVENBERG-MARQUARDT ARTIFICIAL NEURAL NETWORK MODEL: ITALY ROAD TRANSPORTATION SYSTEM. | |
CN110400462B (en) | Track traffic passenger flow monitoring and early warning method and system based on fuzzy theory | |
CN103632546A (en) | Floating car data-based urban road traffic accident influence prediction method | |
CN105303856A (en) | Variable message sign information release method of prediction model | |
Chen et al. | A review on traffic prediction methods for intelligent transportation system in smart cities | |
Tian et al. | Dynamic evolution of demand fluctuation in bike-sharing systems for green travel | |
Li et al. | Real-time congestion prediction for urban arterials using adaptive data-driven methods | |
CN112863182A (en) | Cross-modal data prediction method based on transfer learning | |
Liu et al. | Highway traffic congestion detection and evaluation based on deep learning techniques | |
CN112669595B (en) | Network taxi booking flow prediction method based on deep learning | |
CN107403241A (en) | A kind of urban track traffic for passenger flow second order fluctuation computational methods | |
Li et al. | Graph neural networks in urban intelligence | |
Zou et al. | Application of advanced traffic information systems: field test of a travel-time prediction system with widely spaced detectors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150520 |