CN105185106B - A kind of road traffic flow parameter prediction method based on Granule Computing - Google Patents
A kind of road traffic flow parameter prediction method based on Granule Computing Download PDFInfo
- Publication number
- CN105185106B CN105185106B CN201510408459.XA CN201510408459A CN105185106B CN 105185106 B CN105185106 B CN 105185106B CN 201510408459 A CN201510408459 A CN 201510408459A CN 105185106 B CN105185106 B CN 105185106B
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- data
- traffic flow
- particle
- flow parameter
- 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.)
- Active
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to Traffic information demonstration and traffic administration and control field, the invention discloses a kind of road traffic flow parameter prediction method based on Granule Computing, its feature is specifically included：First, using messenger particle alternate data point as the elementary cell of data mining analysis.2nd, overall prediction framework is run through with Granule Computing thought, grain treatment is used as a kind of data processing method with unified structure, policymaker is allowed to interact residing status between being more clearly understood that a variety of formal systems, their communication way is grasped, enhanced harmonious environment is established between each different approach；3rd, with Fuzzy time sequence and Gath Geva Clustering Theories to rely on, by focusing on the general character of existing official method, the orthogonality of existing good framework is recognized（As probability is touched upon the probability density function of various variables）, and based on varying granularity concept, the siding-to-siding block length analysis model that particle size range is built according to numerical value entity is set up, implementation pattern identification and supposition accordingly.
Description
Technical field
It is the present invention relates to be related to Traffic information demonstration and traffic administration and control field more particularly to a kind of based on Granule Computing
Road traffic flow parameter prediction method.
Background technology
Traffic flow parameter prediction is the important evidence of Traffic flow guidance and Traffic information demonstration.Under urban transportation background,
Traffic flow is based on time-domain dynamic change, so that the prediction of traffic flow parameter is asking based on time-domain dynamic treatment
Topic.Traditional Forecasting Methodology, in precision and dynamic treatment, it is difficult to search out equalization point, causes its deviation that predicts the outcome actual
Traffic parameter data variation trend.
Song and Chissom (1993) propose the concept of Fuzzy time sequence first, compare with traditional fuzzy collection, its
There is good variation characteristic in dynamic time domain.In recent years, Fuzzy time sequence has successfully solved the decision-making of multiple fields
Problem, the concept based on Fuzzy time sequence, scholars propose the Fuzzy time sequence mould based on Fixed Time Interval in succession
The Fuzzy Time Series Model of type and dynamic change time interval.The application study of Fuzzy Time Series Model has been achieved for very
Big achievement, but still there is certain deficiency.In the past research in, scholars seldom consider time variable in itself influence (or
The time is not considered as variable in itself), and research is often confined to data information content in itself, in little mining data pass
Connection information, such as the data regularity of distribution.Scholars have used SVMs, the rough set based on entropy, granular computing etc. to calculate
Method delimit gap length in research range, and, to seek precision of prediction higher, the prediction for but only being obtained in research range is imitated
Really, after research range is with actual demand or time dynamic, estimation effect is decreased obviously, deficient in applicability
Lack.
For traffic flow parameter prediction, CN201210102006.0 discloses a kind of traffic based on Fuzzy Kalman Filter
Stream Forecasting Methodology, it is characterised in that comprise the following steps：Detector is laid in each track in each track in section upstream and section downstream
Collection arithmetic for real-time traffic flow supplemental characteristic；Obtain each lane detector same period historical traffic stream supplemental characteristic；Using Kalman filtering
Technique construction Dynamic Kalman Filtering traffic flow parameter forecast model；Updated by Kalman filtering time update equation and state
Equation, obtains Kalman filtering parameter prediction result；History same period average traffic stream parameter is introduced into formula, fuzzy card is built
Kalman Filtering traffic flow parameter forecast model；The arithmetic for real-time traffic flow parameter and history same period average traffic collected according to detector
Stream parameter, traffic flow parameter to following time interval and afterwards is predicted.CN201210102006.0 is disclosed to be based on mould
The Forecasting Methodology for pasting Kalman filtering is to utilize kalman filter method, and the dynamic prediction of parameter is completed with reference to obscure idea,
Its precision of prediction achieves certain raising compared with conventional method, but precision is non-very accurate, that is, predict the outcome and truly count
It is unable to estimate according to the random error that there is fluctuation, and error.Can be clearer and more definite, it is pre- disclosed in CN201210102006.0
Although the value accuracy of survey method prediction is high, deviation is but certainly existed, be a kind of mistake of accurate type.
The content of the invention
For not high for the road traffic flow parameter prediction reliability in the prior art and adaptable dynamic of shortage
The technical problem of prediction, the invention discloses a kind of road traffic flow parameter prediction method based on Granule Computing.
The purpose of the present invention is realized by following technical proposals：
A kind of road traffic flow parameter prediction method based on Granule Computing, it is characterised in that specifically include following step：
A kind of road traffic flow parameter prediction method based on Granule Computing, it specifically includes following step：Step one, root
According to the number range span of the traffic flow parameter for detecting, research range U=[U are definedl,Uu] and determine messenger particle number h,
Wherein UlRepresent the integer value more arbitrarily small than numerical value minimum value in overall data, UuRepresent and appoint than numerical value maximum in overall data
The big integer value of meaning；Step 2, fuzzy set delimited in research range, and determine the traffic flow parameter data that detect and fuzzy
Membership between collection；Wherein fuzzy set number is identical with messenger particle number；Step 3, determine that logic between fuzzy set is closed
System, obtains fuzzy relation group；Step 4, the trend according to fuzzy relation group, messenger particle area is carried out using Fuzzy time sequence
Between estimate, so as to predict the traffic flow parameter of subsequent time period.
Further, the process of above- mentioned information particle interval estimation is specially：Assuming that logical relation, Aj→Aj1,
Aj2,...,AjpIf logical relation group trend is ascendant trend, and the lower limit of interval estimation is mj, the upper limit of interval estimation is
(mj1+mj2,...,mjp)/p；If logical relation group trend is downward trend, the lower limit of interval estimation is (mj1+mj2,...,
mjp)/p, the upper limit of interval estimation is mj；If logical relation group trend is not only on the rise but also there is downward trend, interval is estimated
The lower limit of meter is (mj1+mj2,...,mjk)/k, the upper limit of interval estimation is (mjk+1+mjk+2,...,mjp)/(p-k), wherein, j is
The subscript of logical relation front end fuzzy set, AjRepresent j-th fuzzy set, Aj1,Aj2,...,AjpRepresent that logical relation front end obscures
Collect corresponding rear end fuzzy set, common p, k is the intermediate point between 1~p, mjIt is AjThe corresponding u of fuzzy setjMedian point.
Further, above- mentioned information particle is by data set D={ xk| k=1 ..., n } constitute, particle characteristic includes area
Between extent length and cover two features of data point number, messenger particle is expressed as Ω=[a, b], and wherein a and b is data set D
={ xk| k=1 ..., n } boundary, the boundary refers to that the boundary up and down of particle, i.e. particle include the bound of data set
Limit.
Further, for Fuzzy time sequence F (the t-1)=A of different time pointsiWith F (t)=Aj, fuzzy logic pass
System regards logical relations of the F (t-1) and F (t) between as, is designated as Ai→Aj, AiIt is relation front end, AjIt is relation rear end；For same
One fuzzy set A, when several fuzzy logical relationship front ends are identical, merges into fuzzy logical relationship group.
Further, as time series χ={ xk| k=1 ..., n }, common n sample, corresponding time coordinate be θ=
{tk| k=1 ..., n }, by degree of membership ui,kWith cluster centre ηiComposition minimize function complete on time data collection
Fuzzy clustering, minimize construction of function be：Wherein zk=
[tk,xk T]TIt is the data point comprising time coordinate；M is weighted index, m ＞ 1；C is cluster species number, c ＞ 2；D2(zk,ηi) be
Data point zkWith cluster centre ηiBetween distance；ui,kRepresent that k-th data point in n sample is under the jurisdiction of the degree of membership of the i-th class.
Further, above-mentioned U=[Ul,Uu] it is research range, the research range is divided into the h area of unequal length
Between, i.e., h messenger particle, the method for determining this h siding-to-siding block length is as follows：Step1. species number c is determined, calculating is subordinate to accordingly
Degree；Species number c=[h/2], represents the maximum integer no more than h/2 values, and h is interval number, and calculates cluster centre η1,η2,...,
ηcAnd corresponding degree of membership ui,k(i=1 ..., c；K=1 ..., n)；Step2. data subset is constructed according to degree of membership；For
Cluster centre η1,η2,...,ηcAnd corresponding degree of membership ui,k(i=1 ..., c；K=1 ..., n), construction data subset is as follows： Step3. messenger particle is built；It is data subset DiCluster centre, asCalculated using messenger particle building method and believed
The optimal upper and lower boundary of particle is ceased, messenger particle is Ωi=[ai,bi]；Step4. corresponding research range interval u is determined1,
u2,...,uh。
Further, the above method also includes pre-processing the traffic flow parameter for detecting.
Further, above-mentioned pretreatment is specially the benefit for carrying out data as interpolation using the average for closing on for two time points
Fill and reparation.
By the technical scheme using the above, the present invention has following beneficial effect：
In urban transportation background of today, road traffic stream mode is always in real-time dynamic change.Traffic flow parameter
Prediction, it is impossible to depart from the influence of traffic flow modes.Traditional Forecasting Methodology, in precision and dynamic treatment, it is difficult to search out flat
Weighing apparatus point, causes its deviation actual traffic supplemental characteristic variation tendency that predicts the outcome.Present forecast demand is moved under time-domain
The process of state treatment, the present invention constructs the Fuzzy Time Series Model based on messenger particle to adapt to this feature：In terms of grain
Thought is calculated, using messenger particle alternate data point as the elementary cell of data mining analysis, by building with varying granularity
Messenger particle, studies the dynamical forecasting problem of traffic flow parameter.In actual traffic environment, the frequency of traffic flow parameter dynamic change
Degree causes that the Accurate Prediction based on numerical value becomes for the difficult point of sphere of learning.But traffic decision-making person is often to parameter prediction
Demand be interval range, traffic flow modes are judged with this.Forecast model of the invention, just in actual demand and dynamic
Treatment is sought to appropriate prediction scheme, it is ensured that prediction accuracy, and the person that preferably serves communications policy decision-making
Information is supplied.
Compared with Forecasting Methodology disclosed in CN201210102006.0, Forecasting Methodology of the invention is based on Granule Computing thought,
It is theoretical with reference to Fuzzy time sequence and relevant cluster, complete the motion interval prediction of traffic flow parameter, that is, the result predicted is
A kind of interval range form.Although this patent does not provide a kind of accurate numerical value, by case verification, this patent can be with
Necessarily within this range, i.e., this patent provides a kind of the accurate of fuzzy type to guarantee actual value.With CN201210102006.0
Compare, difference substantially is that the mistake of accurate type provides numerical value, but but certainly exists error, and dynamic prediction mistake
Error is random fluctuation in journey.And fuzzy type is accurately to replace numerical value with interval range, numerical value is any of interval range
Value, but ensure that numerical value is certain without departing from this interval range, compared with the mistake of accurate type, equivalent to by numerical prediction and ripple
The estimation of dynamic (error) is completed simultaneously.In actual applications, traffic decision-making person is often according to Assessment of Service Level for Urban Roads
Analytical standard and traffic flow parameter contrast, so as to judge traffic flow modes.In Assessment of Serviceability of Roads analytical standard, it is carried
It is interval to the demand of parameter prediction to have supplied the affiliated scope of traffic flow parameter under various traffic behaviors, i.e. traffic decision-making person
Scope.Therefore, in two kinds of Forecasting Methodologies of patent, the interval range that this patent is proposed is predicted closer in actual need
Ask, practicality is higher.
Brief description of the drawings
Fig. 1 is the structural representation with data point as data analysis unit.
Fig. 2 is the structural representation with information as data analysis unit.
Fig. 3 is the structure flow chart of traffic flow parameter forecast model of the present invention.
Fig. 4 is trend analysis figure in traffic flow parameter of the present invention prediction example.
Specific embodiment
With reference to Figure of description, specific embodiment of the invention is described in detail.
The invention discloses a kind of road traffic flow parameter prediction method based on Granule Computing, its feature specifically includes following
The step of：Step one, the number range span according to the traffic flow parameter for detecting, define research range and determine messenger particle
Number；Step 2, fuzzy set delimited in research range, and determined between the traffic flow parameter data that detect and fuzzy set
Membership；Wherein fuzzy set number is identical with messenger particle number；Step 3, the logical relation for determining between fuzzy set, obtain
Fuzzy relation group；Step 4, messenger particle interval estimation is carried out using Fuzzy time sequence, so as to predict subsequent time period
Traffic flow parameter.Fuzzy time sequence is attached to traffic flow parameter prediction field by the present invention, realizes the standard of traffic flow parameter
Really prediction, predicts the interval range of the traffic flow parameter of next time period, and it predicts the outcome 100% accurately, compared to existing
There are the modes such as recursive prediction, neural network prediction in technology, its accuracy rate brings up to 100%, while according to traffic parameter
Feature, such as speed, it only needs to carry out judging whether congestion by obtaining a value range, it is not necessary to accurate data point
Value, it is clear that obtain a point value that there is error with it, it is one that the present invention can obtain a 100% true scope value
Preferably selection.The present invention using messenger particle alternate data point as the elementary cell of data mining analysis, with Granule Computing thought
Through overall prediction framework, with Fuzzy time sequence and Gath-Geva Clustering Theories to rely on, by focusing on existing formal side
The general character of method, recognizes the orthogonality of existing good framework, based on varying granularity concept, sets up according to numerical value entity structure
The siding-to-siding block length analysis model of particle size range is built, accordingly implementation pattern identification and supposition.
Above-mentioned messenger particle is different from conventional data analysis side as data mining analysis elementary cell of the invention
Method, the introducing of messenger particle is easy to be identified from extremely different problem aspects the general character of data essence.Messenger particle is by counting
According to collection D={ xk| k=1 ..., n } constitute, particle characteristic is comprising particle interval extent length and covers two spies of data point number
Levy.Messenger particle is expressed as Ω=[a, b], and wherein a and b is considered as data set D={ xk| k=1 ..., n boundary.Its
Middle L (Ω) is the siding-to-siding block length of particle, and the function of computational length can be set to F1, then the feature representation of messenger particle can define
It is F1(L(Ω)).Particle also needs to describe the data amount check contained by particle, is designated as Card { xk|xk∈ Ω }, function can be used
F2(Card{xk|xk∈ Ω }) represent.Then the computational methods of boundary can be expressed as follows：
V (b)=F1(|b-med(D)|)·F2(Card{xk∈D|med(D)≤xk≤b}) (1)
V (a)=F1(|a-med(D)|)·F2(Card{xk∈D|a≤xk≤med(D)}) (2)
Wherein med (D) is the median of data set D.On this basis, optimal messenger particle bound computational methods are such as
Under：
Find out a for meeting above formulaoptAnd boptAs the bound of messenger particle.On length and data amount check letter
Number, may be calculated as：
F1(u)=exp (- α u), F2(u)=u (4)
Further, the above-mentioned construction to messenger particle (in description believe by formula (1)~(3) and front and rear related text
The statement of feature and feature that breath particle should be included and calculating, wherein calculating section is related to theory) it is related to Fuzzy Time sequence
(formula (6)~(10) are exactly Clustering Theory thought to row, how are described in detail in time series with Gath-Geva Clustering Theories
On the basis of, using the Clustering Theory, complete cluster, i.e. fuzzy partition.Wherein degree of membership inherently fuzzy clustering is basic
Concept, probability density is calculated to be using Gaussian function analysis and adjusted.Construction in order to be applied to messenger particle of the invention, to two
Kind theory is adjusted as follows：
Fuzzy time sequence, is the effect that this factor of time is added on the basis of fuzzy set, redefines fuzzy set,
Construction method is as follows：
1. are defined with U={ u1, u2..., unProblem of representation research range, then under scope U fuzzy set definition such as
Under：
Wherein fAIt is the membership function for being under the jurisdiction of fuzzy set A, fARepresent uiIt is under the jurisdiction of the degree of fuzzy set A, 1≤i≤n, i.e.,
fA→[0,1]。
One time variable set Y (t) of the hypothesis of definition 2. (t=..., 0,1,2 ...), by corresponding data time point,
It is applied to each time subset f of fuzzy set Ai(t) (i=1,2 ...) in, F (t) is considered as the f at correspondence momenti(t)(i
=1,2 ...) set, then F (t) be referred to as the Fuzzy time sequence in time variable Y (t) (t=..., 0,1,2 ...).
Fuzzy time sequence F (t-1)=A of the definition 3. for different time pointsiWith F (t)=Aj, fuzzy logical relationship can
To regard logical relations of the F (t-1) and F (t) between as, A is denoted asi→Aj, AiIt is relation front end, AjIt is relation rear end.
Define 4. and be directed to same fuzzy set A, when several fuzzy logical relationship front ends are identical, can merge into fuzzy
Logical relation group, reduces and calculates amount of analysis.There is relation Ai→Aj1,Ai→Aj2..., fuzzy logical relationship group can be formed
Ai→Aj1,Aj2,...。
Fuzzy partition is carried out to time series using Gath-Geva fuzzy clusterings, temporal information is added thereto, obtain one
The individual fuzzy partition with time domain information.It is assumed that a time series χ={ xk| k=1 ..., n }, common n sample is corresponding
Time coordinate is θ={ tk| k=1 ..., n }, the Clustering Theory essence is by degree of membership ui,kWith cluster centre ηiConstitute most
The fuzzy clustering on time data collection that smallization function is completed, minimizes construction of function as follows：
Take local minimum, wherein zk=[tk,xk T]TIt is the data point comprising time coordinate；M is weighted index, general m
＞ 1；C is cluster species number, general c ＞ 2；D2(zk,ηi) it is data point zkWith cluster centre ηiBetween distance.Minimize function
Constraints is as follows：
Because tkWith xkIt is separate, apart from D2(zk,ηi) can be expressed with following formula：
For time variable and the independence of numerical variable, the probability density in Gath-Geva Fuzzy Clustering Theories is entered
Row adjustment, will be apart from D2(zk,ηi) in the probability density that is related to adopt and calculate with the following method：
Probability density functionRepresent tkBelong to the probability of the i-th class, it is possible to use Gaussian function is calculated：
WhereinWithAverage and variance are represented respectively.And probability density functionRepresent xkBelong to the i-th class
Probability, it is possible to use Gaussian function is calculated：
WhereinWithAverage and covariance are represented respectively, and r is covariance matrixLine number.αiIt is coefficient, meets：
In sum, when the division hop count of given research range, the theory can complete cluster and phase based on time series
The fuzzy partition answered.
Further, it is theoretical based on above two, the determination of the length (siding-to-siding block length and data amount check) of messenger particle
Flow is as follows：
The delimitation of messenger particle siding-to-siding block length can influence the precision for finally predicting the outcome, the present invention to use a kind of unequal intervals
The division methods of length, to improve the degree of accuracy for predicting the outcome.
Research range needs the interval number for assuming to divide, i.e. messenger particle number first when dividing.So as to provide c classes cluster
Center,Then according to corresponding degree of membership, c class subsets are built, and calculate the messenger particle of maximum
Subset, finally determines the siding-to-siding block length of messenger particle.
Specific division methods are as follows：Assuming that preset time sequence sets D={ xk| k=1 ..., n }, define Dmin=min { xi|
xi∈ D }, Dmax=max { xi|xi∈D}.Assuming that U=[Ul,Uu] it is research range, the research range is divided into h Length discrepancy
The interval (messenger particle number) of degree, the method for determining this h siding-to-siding block length is as follows：
Step1. determine species number c, calculate corresponding degree of membership.Species number c=[h/2], represents no more than h/2 values most
Big integer, h is interval number (messenger particle number) and calculates cluster centre η1,η2,...,ηcAnd corresponding degree of membership ui,k(i=
1,...,c；K=1 ..., n)；
Step2. data subset is constructed according to degree of membership.For cluster centre η1,η2,...,ηcAnd be subordinate to accordingly
Degree ui,k(i=1 ..., c；K=1 ..., n), construction data subset is as follows：
Step3. messenger particle is built.It is data subset DiCluster centre, as
The optimal upper and lower boundary of messenger particle is calculated using messenger particle building method, messenger particle is Ωi=[ai,bi]；
Step4. corresponding research range interval u is determined1,u2,...,uh。
In sum, the determination of siding-to-siding block length can be reduced to following two kinds of situations：
(I) when h is odd number, ifThen u1=[Ul,(Ul+med(D1))/2], u2=[(Ul+
med(D1))/2,med(D1)], u2i-1=[med (Di-1),(bi-1+ai)/2], u2i=[(bi-1+ai)/2,med(Di)], uh=
[med(D(h-1)/2),U0] (i=2 ..., (h-1)/2), otherwise u1=[Ul,med(D1)], u2i=[med (Di),(bi+ai+1)/
2], u2i+1=[(bi+ai+1)/2,med(Di+1)], uh-1=[med (D(h-1)/2),(Uu+b(h-1)/2)/2], uh=[(Uu+
b(h-1)/2)/2,Uu] (i=2 ..., (h-3)/2).
(II) when h is even number, u1=[Ul,med(D1)], u2i=[med (Di),(bi+ai+1)/2], u2i+1=[(bi+
ai+1)/2,med(Di+1)], uh=[med (Dh/2),Uu] (i=1 ..., (h-2)/2).
Further, based on above-mentioned prediction thought, basic theory, particle interval length configuration model, based on grain
The traffic flow parameter forecast model of calculating builds and is made up of following four step：
Step1. define research range and determine messenger particle siding-to-siding block length
Research range is an information matrix, the traffic flow that A points are detected to each moment traffic detector between B points
Parameter is as an element in matrix and completes the division of messenger particle siding-to-siding block length：
Step2. fuzzy set and Fuzzy time sequence are built
According to the concept of Fuzzy time sequence, on the basis of Given information particle siding-to-siding block length, can be as follows
Set up fuzzy set：
Ai=1/ui+0.5/ui-1+0.5/ui+1, (i=1 ..., h) (12)
Wherein u0And uh+1Take the numerical value of infinity.
Step3. the logical relation between fuzzy set is determined
Logical relation and bad direct description between fuzzy set, but due between fuzzy set and messenger particle build when when
Between domain corresponding relation, can be described by the degree of consistency between fuzzy set and messenger particle, be denoted as Poss (Ω, Ai),
Computational methods are as follows：
Poss(Ω,Ai)=supx∈Ω[Ω(x)tAi(x)] (13)
Wherein sup function representations capping, equivalent to seeking degree of association maximization problems.By the statement of the method, time
Sequence can be changed into messenger particle time series.
In above-mentioned analysis, there are a kind of special circumstances, it is assumed that Ω7(expression is divided into 7 intervals) corresponding AiConsistent journey
Degree is respectively 0,0,0,0,0.5,1,1, and maximum is A6And A7, then Ω7Logical relation mapping should be 0.5A6+0.5A7.It is determined that
After logical relation, logic need to be carried out according to the mapping of each messenger particle logical relation and write, it is assumed for example that each information
The mapping of grain is respectively A1,A3,A3,A4,A3,A4,0.5A6+0.5A7, then logical relation group can be written as：A1→A3；A3→A3,
A4；A4→A3,0.5A6+0.5A7Three groups.
Step4. interval estimation
On the basis of the logical relation between obtaining fuzzy set, it is assumed that logical relation, Aj→Aj1,Aj2,...,AjpEstimate in interval
Meter method is mainly three kinds of principles：
If the logical relation group of principle 1. trend is ascendant trend, i.e. j1, j2 ..., jp ＞ j, then the lower limit of interval estimation
It is mj(ujMedian point), the upper limit of interval estimation is (mj1+mj2,...,mjp)/p, so that the interval for obtaining current point in time is estimated
Meter result.
If the logical relation group of principle 2. trend is downward trend, i.e. j1, j2 ..., jp≤j, then the lower limit of interval estimation
It is (mj1+mj2,...,mjp)/p, the upper limit of interval estimation is mj, so as to obtain the interval estimation result of current point in time.
If the logical relation group of principle 3. trend is not only on the rise but also have a downward trend, i.e. j1, j2 ..., jk≤j and
Jk+1 ..., jp ＞ j, then the lower limit of interval estimation is (mj1+mj2,...,mjk)/k, the upper limit of interval estimation is (mjk+1+
mjk+2,...,mjp)/(p-k), so as to obtain the interval estimation result of current point in time.
The following detailed description of how building forecast model and complete dynamic prediction.It specifically includes following step：Data
Pretreatment, delimit fuzzy set, build the relation between fuzzy set, interval estimation.
The detailed description of model construction and application is carried out by example：Using in March, 2011 Beijing's Three links theory microwave
Detection data carries out traffic flow parameter dynamic prediction analysis.Microwave detection data raw data form is as shown in table 1：
The microwave of table 1 detection initial data (partial data) Table 1 microwave detection data
Shown in upper table be only Beijing Three links theory partial period record for obtaining partial data.The original detection number of microwave
There are the parameters such as the magnitude of traffic flow (being obtained by the statistics number of CAR_ID), occupation rate, speed according to the traffic flow parameter for containing, in order to
It is easy to sample calculation analysis, herein only by data based on speed this parameter, verifies model estimation effect.When initial data takes 1-19
Between point data, 20-22 times point data as checking data.
On the basis of existing basic data, the pretreatment of data, such as part-time point are carried out for speed parameter information
Speed data be 0 value, be primarily due to there is the situation of missing inspection when detector is recognized, the average that can close on for two time points is made
For interpolation carries out the supplement of data and repairs, the integrity degree of data message is improved.After data prediction, the result such as institute of table 2 is obtained
Show：
The preprocessing of of table 2 microwave detection speed preprocessed data (partial data) Table 2
microwave detection data(snapshot)
Using model of the invention, the estimation prediction of speed is carried out, calculating process is as follows：
Step1. define research range and determine messenger particle number.
According to number range span, research range U=[35,95] is defined, assume to be divided into h=7 information in example
Grain, i.e., in research range, the species number c=[7/2]=3 of cluster.Research range dividing condition is：
U={ u1,u2,...,u7}。
Step2. fuzzy set delimited
Fuzzy set delimited in research range, 7 messenger particles assume that in example, general fuzzy set number be same
.Determine that FUZZY SET APPROACH TO ENVIRONMENTAL is as follows using messenger particle：
Ai=1/ui+0.5/ui-1+0.5/ui+1, (i=1 ..., h) (14)
According to said method, with reference to institute's established model of the present invention, the corresponding membership result of calculation of fuzzy set is as shown in table 3：
The data membership situation of table 3
Table 3 data’s degree of membership
Parameter alpha value is different, and data membership situation there may be certain difference, using historical data in practical application
The value of calibrating parameters α, choosing value is generally arrayIn element value.
Step3. the logical relation between fuzzy set is built
Determine logical relation between fuzzy set and arrange these relations using formula (13) to arrange, obtain what is needed in model
Fuzzy relation group, as shown in table 4：
The logic relationship groups between of logical relation group Table 4 between the fuzzy set of table 4
fuzzy sets
Step4. interval estimation
What Fuzzy time sequence prediction was completed is the estimation of messenger particle feature structure, the i.e. estimation of particle length.Information
The estimate at the time point included in particle can be substituted with any value in siding-to-siding block length, and the error of generation allows model in model
In enclosing.Messenger particle number is determined according to actual needs, and the time point that messenger particle is included is fewer, and particle degree of refinement is got over
Height, estimated accuracy is also higher, but amount of calculation fast can also increase therewith.During in face of huge data, time point contained by messenger particle compared with
It is many, but the typically care of huge data is not specific time point, but the estimated accuracy of time period, the interval with messenger particle is estimated
Not contradiction is counted, while can also reduce workload.According to set forth herein interval estimation principle, with reference to the fuzzy set for having obtained
Between logical relation group, 7 time period (messenger particle) interval estimation results can be obtained as follows：
The estimation results in Intervals of 5 interval estimation result Table of table 5
It can be seen from upper table, predicting the outcome for subsequent time period is [79.5,98.5], due in example by 19 time points
Data be divided into 7 messenger particles, it is 2-3 to represent the time point that a time period covers.So the result that upper table is estimated
Represent that the data value at following 2-3 time point should be in [79.5,98.5].The True Data value of observing time point 20-22
Understand that estimated result is accurate.
Certainly, due to the influence of the factor such as accidental, 3 points of general forecast are not enough to prove the applicability of model.Using micro-
Wave number evidence, proceeds prediction experiment again, and preceding 19 data of predicted time point are only taken every time as historical data, draws prediction
Trend analysis figure is as shown in Figure 4.
The broken line graph trend that can be seen that prediction data from anticipation trend figure is basic with original True Data broken line graph trend
It is identical.The confidential interval of model selective analysis data of the invention, emphasis is contrasted with the variation tendency of actual value.Continuous prediction
When, mode input is updated according to historical data, dynamic prediction is capable of achieving as shown in example.
The coefficient and parameter gone out given in the above embodiments, are available to those skilled in the art to realize or use
Of the invention, the present invention is not limited and only takes foregoing disclosed numerical value, without departing from the present invention in the case of the inventive idea, this
The technical staff in field can make various modifications or adjustment to above-described embodiment, thus protection scope of the present invention does not go up
State embodiment to be limited, and should be the maximum magnitude for meeting the inventive features that claims are mentioned.
Claims (7)
1. a kind of road traffic flow parameter prediction method based on Granule Computing, it is characterised in that specifically include following step：Step
Rapid one, according to the number range span of the traffic flow parameter for detecting, research range U=[U are definedl,Uu] and determine messenger particle
Number h, wherein UlRepresent the integer value more arbitrarily small than numerical value minimum value in overall data, UuRepresent than numerical value in overall data most
It is worth arbitrarily large integer value greatly；Step 2, fuzzy set delimited in research range, and determine the traffic flow parameter data for detecting
Membership between fuzzy set；Wherein fuzzy set number is identical with messenger particle number；Step 3, determine fuzzy set between
Logical relation, obtains fuzzy relation group；Step 4, the trend according to fuzzy relation group, row information is entered using Fuzzy time sequence
Particle interval estimation, so as to predict the traffic flow parameter of subsequent time period；The process of described information particle interval estimation is specific
For：Assuming that logical relation, Aj→Aj1,Aj2,...,AjpIf logical relation group trend is ascendant trend, under interval estimation
It is limited to mj, the upper limit of interval estimation is (mj1+mj2,...,mjp)/p；If logical relation group trend is downward trend, interval
The lower limit of estimation is (mj1+mj2,...,mjp)/p, the upper limit of interval estimation is mj；If logical relation group moves towards existing rising become
Gesture has downward trend again, then the lower limit of interval estimation is (mj1+mj2,...,mjk)/k, the upper limit of interval estimation is (mjk+1+
mjk+2,...,mjp)/(p-k), wherein, j is the subscript of logical relation front end fuzzy set, AjRepresent j-th fuzzy set, Aj1,
Aj2,...,AjpThe corresponding rear end fuzzy set of logical relation front end fuzzy set is represented, common p, k is the intermediate point between 1~p, mjFor
AjThe corresponding u of fuzzy setjMedian point.
2. the road traffic flow parameter prediction method of Granule Computing is based on as claimed in claim 1, it is characterised in that described information
Particle is by data set D={ xk| k=1 ..., n } constitute, particle characteristic is comprising interval range length and covers data point number two
Individual feature, messenger particle is expressed as Ω=[a, b], and wherein a and b is data set D={ xk| k=1 ..., n } boundary, it is described
Boundary refers to the boundary up and down that the boundary up and down of particle, i.e. particle include data set.
3. the road traffic flow parameter prediction method of Granule Computing is based on as claimed in claim 1, it is characterised in that for difference
Fuzzy time sequence F (the t-1)=A at time pointiWith F (t)=Aj, fuzzy logical relationship regards as F (t-1) and F (t) between
Logical relation, is designated as Ai→Aj, AiIt is relation front end, AjIt is relation rear end；For same fuzzy set A, when several fuzzy logics
When relation front end is identical, fuzzy logical relationship group is merged into.
4. the road traffic flow parameter prediction method of Granule Computing is based on as claimed in claim 1, it is characterised in that when time sequence
Row χ={ xk| k=1 ..., n }, common n sample, corresponding time coordinate is θ={ tk| k=1 ..., n }, by degree of membership
ui,kWith cluster centre ηiThe fuzzy clustering on time data collection for minimizing function completion of composition, minimizes construction of function
For：Wherein zk=[tk,xk T]TIt is the data comprising time coordinate
Point；M is weighted index, m ＞ 1；C is cluster species number, c ＞ 2；D2(zk,ηi) it is data point zkWith cluster centre ηiBetween distance；
ui,kRepresent that k-th data point in n sample is under the jurisdiction of the degree of membership of the i-th class.
5. the road traffic flow parameter prediction method of Granule Computing is based on as claimed in claim 1, it is characterised in that U=[Ul,Uu]
It is research range, the research range is divided into the h interval of unequal length, i.e., h messenger particle determines this h interval length
The method of degree is as follows：Step1. determine species number c, calculate corresponding degree of membership；Species number c=[h/2], represents and is no more than h/2
The maximum integer of value, h is interval number, and calculates cluster centre η1,η2,...,ηcAnd corresponding degree of membership ui,k(i=1 ..., c；
K=1 ..., n)；Step2. data subset is constructed according to degree of membership；For cluster centre η1,η2,...,ηcAnd be subordinate to accordingly
Degree ui,k(i=1 ..., c；K=1 ..., n), construction data subset is as follows：
Step3. messenger particle is built；It is data subset DiCluster centre, asThe optimal upper and lower boundary of messenger particle, messenger particle are calculated using messenger particle building method
It is Ωi=[ai,bi]；Step4. corresponding research range interval u is determined1,u2,...,uh。
6. the road traffic flow parameter prediction method of Granule Computing is based on as claimed in claim 1, it is characterised in that methods described
Also include pre-processing the traffic flow parameter for detecting.
7. the road traffic flow parameter prediction method of Granule Computing is based on as claimed in claim 1, it is characterised in that the pre- place
Reason is specially to be carried out the supplement of data and repaired using the average for closing on for two time points as interpolation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510408459.XA CN105185106B (en) | 2015-07-13 | 2015-07-13 | A kind of road traffic flow parameter prediction method based on Granule Computing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510408459.XA CN105185106B (en) | 2015-07-13 | 2015-07-13 | A kind of road traffic flow parameter prediction method based on Granule Computing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105185106A CN105185106A (en) | 2015-12-23 |
| CN105185106B true CN105185106B (en) | 2017-07-04 |
Family
ID=54907149
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510408459.XA Active CN105185106B (en) | 2015-07-13 | 2015-07-13 | A kind of road traffic flow parameter prediction method based on Granule Computing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105185106B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108519993B (en) * | 2018-03-02 | 2022-03-29 | 华南理工大学 | Social network hotspot event detection method based on multi-data-stream calculation |
| CN109543046A (en) * | 2018-11-16 | 2019-03-29 | 重庆邮电大学 | A kind of robot data interoperability Methodologies for Building Domain Ontology based on deep learning |
| CN109598935B (en) * | 2018-12-14 | 2020-12-15 | 银江股份有限公司 | Traffic data prediction method based on ultra-long time sequence |
| CN111696367B (en) * | 2020-06-11 | 2022-06-07 | 西安电子科技大学 | Multi-geomagnetic-sensor speed measurement system and speed measurement method thereof |
| CN112215396A (en) * | 2020-09-03 | 2021-01-12 | 中交西安筑路机械有限公司 | Data prediction algorithm based on fuzzy time series analysis and particle swarm technique |
| CN113963539B (en) * | 2021-10-19 | 2022-06-10 | 交通运输部公路科学研究所 | Highway traffic accident identification method, module and system |
| CN114240687A (en) * | 2021-11-08 | 2022-03-25 | 国电南瑞科技股份有限公司 | Energy hosting efficiency analysis method suitable for comprehensive energy system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102004039283A1 (en) * | 2004-08-13 | 2006-03-02 | Daimlerchrysler Ag | Forecasting journey time in road network, by taking into account time-space associations and/or patterns when selecting proportion of measured data as predicted parameter |
| JP3932383B2 (en) * | 2002-09-27 | 2007-06-20 | 財団法人生産技術研究奨励会 | Traveling time prediction device |
| CN101783075A (en) * | 2010-02-05 | 2010-07-21 | 北京科技大学 | System for forecasting traffic flow of urban ring-shaped roads |
| TW201039264A (en) * | 2009-04-24 | 2010-11-01 | Univ Nat Taiwan Science Tech | Forecasting TAIFEX based on particle swarm optimization |
| CN102087787A (en) * | 2011-03-11 | 2011-06-08 | 上海千年工程建设咨询有限公司 | Prediction device and prediction method for short time traffic conditions |
| CN102629418A (en) * | 2012-04-09 | 2012-08-08 | 浙江工业大学 | Fuzzy kalman filtering-based traffic flow parameter prediction method |
-
2015
- 2015-07-13 CN CN201510408459.XA patent/CN105185106B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3932383B2 (en) * | 2002-09-27 | 2007-06-20 | 財団法人生産技術研究奨励会 | Traveling time prediction device |
| DE102004039283A1 (en) * | 2004-08-13 | 2006-03-02 | Daimlerchrysler Ag | Forecasting journey time in road network, by taking into account time-space associations and/or patterns when selecting proportion of measured data as predicted parameter |
| TW201039264A (en) * | 2009-04-24 | 2010-11-01 | Univ Nat Taiwan Science Tech | Forecasting TAIFEX based on particle swarm optimization |
| CN101783075A (en) * | 2010-02-05 | 2010-07-21 | 北京科技大学 | System for forecasting traffic flow of urban ring-shaped roads |
| CN102087787A (en) * | 2011-03-11 | 2011-06-08 | 上海千年工程建设咨询有限公司 | Prediction device and prediction method for short time traffic conditions |
| CN102629418A (en) * | 2012-04-09 | 2012-08-08 | 浙江工业大学 | Fuzzy kalman filtering-based traffic flow parameter prediction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105185106A (en) | 2015-12-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105185106B (en) | A kind of road traffic flow parameter prediction method based on Granule Computing | |
| Zheng et al. | Deep and embedded learning approach for traffic flow prediction in urban informatics | |
| Celikoglu et al. | Extension of traffic flow pattern dynamic classification by a macroscopic model using multivariate clustering | |
| Yao et al. | A support vector machine with the tabu search algorithm for freeway incident detection | |
| CN108492555A (en) | A kind of city road net traffic state evaluation method and device | |
| CN101216998B (en) | An urban traffic flow information amalgamation method of evidence theory based on fuzzy rough sets | |
| Ma et al. | Short-term traffic flow forecasting by selecting appropriate predictions based on pattern matching | |
| CN106448151B (en) | Short-term traffic flow prediction method | |
| CN106781489B (en) | A kind of road network trend prediction method based on recurrent neural network | |
| Lu et al. | Lane-level traffic speed forecasting: A novel mixed deep learning model | |
| CN103839409A (en) | Traffic flow state judgment method based on multiple-cross-section vision sensing clustering analysis | |
| Chen et al. | A multiscale-grid-based stacked bidirectional GRU neural network model for predicting traffic speeds of urban expressways | |
| CN113033899B (en) | Unmanned adjacent vehicle track prediction method | |
| Gao et al. | Traffic speed forecast in adjacent region between highway and urban expressway: based on MFD and GRU model | |
| Bartlett et al. | A machine learning based approach for the prediction of road traffic flow on urbanised arterial roads | |
| Zhang et al. | Traffic flow prediction under multiple adverse weather based on self-attention mechanism and deep learning models | |
| Yu et al. | Speed‐Density Model of Interrupted Traffic Flow Based on Coil Data | |
| Djenouri et al. | Intelligent deep fusion network for urban traffic flow anomaly identification | |
| Bao | A multi-index fusion clustering strategy for traffic flow state identification | |
| Li et al. | Traffic fatalities prediction based on support vector machine | |
| Xue et al. | Traffic status evolution trend prediction based on congestion propagation effects under rainy weather | |
| Xu et al. | PIGAT: Physics-Informed Graph Attention Transformer for Air Traffic State Prediction | |
| CN109887290B (en) | Traffic flow prediction method based on balance index smoothing method and stack type self-encoder | |
| CN114463978B (en) | Data monitoring method based on track traffic information processing terminal | |
| Dong et al. | An identification model of urban critical links with macroscopic fundamental diagram theory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |