CN104021665A - Gene search short time traffic flow forecasting method - Google Patents

Abstract

The invention discloses a gene search short time traffic flow forecasting method. The method includes the following steps that (1) when traffic flow data are collected, it is guaranteed that the collected data include a certain number of working day data, non-working day data and major holiday and festival data, namely the diversity of a database is guaranteed; (2) binary coding is conduced on collected continuous time period traffic flow data, so that a short time traffic flow forecasting gene pool is established; (3) binary coding is conduced on current continuous time period traffic flow data; (4) the most similar chromosome populations are searched for by comparing current traffic flow data chromogenes with historical traffic flow data chromogenes in the short time traffic flow forecasting gene pool; (5) the deviation values of all the chromogenes in the similar populations are figured out; (6) forecasting gene segments of three chromosomes with the smallest deviation values are decoded, the mean value of the decoded values of the three forecasting gene segments is taken to be used as a final forecasting result, hence, the forecasting precision is higher than that of a model which is forecasted through the data of the same day or the historical data of the past several days, and the method is an effective short time traffic flow forecasting method.

Description

Gene search Short-time Traffic Flow Forecasting Methods

Technical field

The present invention relates to a kind of gene search Short-time Traffic Flow Forecasting Methods that can realize intelligent traffic administration system.

Background technology

Since entering the new century, China's economy grows continuously and fast, and a large amount of automobiles enter daily life, has also brought the problems such as traffic congestion, traffic pollution, traffic hazard simultaneously.Under current road conditions, intelligent transportation system is the method for relevant issues in a kind of effective solution field of traffic.In Intelligent traffic management systems, forecasting traffic flow control method, as fundamental research, is the key that realizes Intelligent traffic management systems, significant.

Single Short-time Traffic Flow Forecasting Methods all requires unique information characteristics and specific applicable elements at present, cause single forecast model not high to complicated traffic flow forecasting precision, and often needed to carry out large quantitative analysis and judge to select the best approach before prediction.Some short-time traffic flow forecast combined methods often can not be taken into account accuracy and the real-time of prediction simultaneously, although these some combined methods wherein improve precision of prediction to a certain extent, but its complicated operation, labor capacity are larger, be unfavorable for carrying out arithmetic for real-time traffic flow prediction.The open text of patent that for example number of patent application is 200910100395.1, adopts the method for intelligences combination, by the prediction output of the historical method of average and neural network model, carries out fuzzy combination and obtain the prediction output of built-up pattern.The weak point that the method exists is more complicated for the anabolic process of two Seed models, and workload is large, and need to carry out working day, weekend, great festivals or holidays of classification prediction, the prediction accuracy that guarantee is higher to traffic flow data.Number of patent application is 201210186056.1 patent, adopts the method for weighted array, and predicting the outcome of k nearest neighbor method is weighted to summation with predicting the outcome of fuzzy neural network, predicts the outcome as final built-up pattern.The weak point that the method exists is that k nearest neighbor predicted method is made a prediction to the traffic flow situation of next time period by the distance between sample in historical data base and current traffic flow situation, but distance only reflects current point and k nearest neighbor " near property ", can not directly reflect " shape similarity " between them, shape similarity directly reflects the development and change rule of traffic flow.And in the time that traffic flow data storehouse is larger, k nearest neighbor method predetermined speed is slower.

Summary of the invention

The object of the invention is to overcome the deficiencies in the prior art and provide a kind of to improve the accuracy of short-time traffic flow forecast and the gene of real-time search Short-time Traffic Flow Forecasting Methods.

The object of the present invention is achieved like this: a kind of gene search Short-time Traffic Flow Forecasting Methods, it is characterized in that, and comprise the following steps:

Step 1), data acquisition:

When traffic flow data is gathered, ensure to comprise in the data that collect a certain amount of nonworkdays data and great festivals or holidays data, ensure the diversity of database;

Step 2), historical traffic flow data coding:

Utilize current t-3, t-2, the traffic flow data of t-1 time period is predicted the magnitude of traffic flow of t time period, choose the traffic flow data q (t-3) of continuous 4 time periods, q (t-2), q (t-1), the assortment of genes of q (t) is as a chromosome C, specifically for different traffic flow data storehouse sizes and the prediction accuracy requiring, select section traffic flow data continuous time of different length to predict current traffic flow situation, utilize t-m, t-m+1, t-2, the traffic flow data q (t-m) of t-1 time period, q (t-m+1), q (t-2), q (t-1) predicts the magnitude of traffic flow of t time period, wherein m is positive natural number, in chromosome C, comprise comparison genetic fragment D ₁with predicted gene fragment D ₂, D ₁by the gene of q (t-3), q (t-2), q (t-1) correspondence combine D ₂by gene corresponding to q (t) composition,

Step 3), current traffic flow data coding:

When the current t time period, the magnitude of traffic flow was predicted, t-3, the t-2 before the t time period, the assortment of genes of the traffic flow data of tri-time periods of t-1 are as a chromosome Y ₁this chromogene is 36 binary coded value, wherein first 12 is that hundred corresponding binary coded value of traffic flow numeral combine, middle 12 is that ten corresponding binary coded value of traffic flow numeral combine, and latter 12 is that a traffic flow numeral corresponding binary coded value combines;

Chromosome C in short-time traffic flow forecast gene pool is because of with predicted gene fragment D ₂so it is 48 binary coded value; For the ease of comparing with chromosome C, to chromosome Y ₁repair, increase by 12 and be encoded to 0 amorph fragment 000000000000, the chromosome after repairing is 48 binary coded value, is designated as match-on criterion chromosome Y;

Step 4), search for similar chromosome population:

In short-time traffic flow forecast gene pool, search for the chromosome C identical with front 18 genes of match-on criterion chromosome Y as initial population, in the time that short-time traffic flow forecast gene pool is larger, select the more homologous genes of multidigit, to improve search efficiency, initial population quantity is defined as P, in the time of P > 10, taking the 19th genic value of match-on criterion chromosome Y as standard, in initial population, continue the search chromosome C identical with the 19th genic value of match-on criterion chromosome Y, then again add up qualified similar chromosome sum, if P is still greater than 10, taking the 20th genic value of match-on criterion chromosome Y as standard, until when P < 10, stop, this is initial population quantity,

Step 5), similar chromosome deviate calculate:

Each chromosome in initial population all contains 48 genes, and according to search condition, front 18 genes are just the same, and 48 genes of chromosome C are divided into 16 genetic fragments, and every three genes are a genetic fragment, and the deviate of each genetic fragment is designated as because front 18 genes are just the same, front 6 genetic fragments are the same, when calculation deviation value, calculate since the 7th genetic fragment;

Every three genes are a genetic fragment, are that the position of the binary value that these three gene pairss are answered is the same because in t-3, t-2, corresponding traffic flow data binary code of t-1 time period, and in the time of calculation deviation value, they have congruency; For each chromosome C in initial population _n, n is positive natural number, and n < 10, chromosome C _narbitrary genetic fragment calculating principle in the time calculating with the deviation value of match-on criterion chromosome Y all different; Concrete formula is as follows:

Chromosome C in initial population P _n, n is positive natural number, and n < 10, a i genetic fragment and i genetic fragment of match-on criterion chromosome Y have a gene different (a=0,1,2,3), chromosome C in initial population P _ni its deviate of genetic fragment be:

${\mathrm{\&Delta;}}_n^i=a\&times;2^b\&times;{10}^c,$

In formula 2 ^bfor decimal system transforming relationship corresponding to this genetic fragment, 10 ^crepresent that this gene position is that binary coded value corresponding to ten of traffic flow numerals or individual position combines;

Chromosome C in initial population P _ntotal departure value:

${\mathrm{\&Delta;}}_n={\mathrm{\&Delta;}}_n^7+{\mathrm{\&Delta;}}_n^8+{\mathrm{\&Delta;}}_n^9+{\mathrm{\&Delta;}}_n^{10}+{\mathrm{\&Delta;}}_n^{11}+{\mathrm{\&Delta;}}_n^{12};$

Step 6), predicted gene fragment decoding and prediction:

Choose deviate Δ _nthree minimum chromosome C _n, n=1,2,3, by the predicted gene fragment D in every item chromosome ₂be decoded as decimal number and be respectively S ₁, S ₂, S ₃, they are the magnitude of traffic flow of time period similar to the current t time period magnitude of traffic flow in historical data base, the value S that utilizes them to predict the magnitude of traffic flow of current t time period is:

$S=\frac{S_1+S_2+S_3}{3}.$

The present invention has following good effect:

Accuracy: the gene search procedure that the present invention proposes, by chromosomal binary code in current traffic flow situation chromosome and short-time traffic flow forecast gene pool is compared by turn, is searched for the most similar chromosome.Because the singularity of this law coding, so the essence of search is that the size of data of continuous three time periods of short-term traffic flow is approached by turn simultaneously, the sample standard deviation searching is like this to possess the high quality samples of " shape similarity " with current traffic flow situation.Therefore the prediction accuracy of this method is also high compared with k nearest neighbor, and this method can make a prediction to current traffic flow situation based on infinitely great traffic flow data storehouse in theory, and therefore precision of prediction is the highest in all methods.

Rapidity: gene search procedure utilizes scale-of-two to carry out gene code to historical traffic flow data, the sample after coding is arranged according to binary sized in short-time traffic flow forecast gene pool.Therefore in search when the most similar sample, can realize in short-time traffic flow forecast gene pool location fast according to the size of current traffic flow situation binary code, extract the most similar initial sample population.

Simplification: Classical forecast method essence is to predict by the rule between mathematical formulae computational data, and conventionally working day, nonworkdays, great festivals or holidays separately need to be predicted for the precision that improves prediction.Gene search procedure is different from Classical forecast algorithm principle, by traffic flow situation is carried out to binary coding, utilize the method for gene coupling and predict, having avoided a large amount of mathematical computations, and without to separately predicting working day, nonworkdays, great festivals or holidays.Because the traffic flow situation between them does not possess similarity, in the time that searching for, gene is independent of each other.So the gene search procedure simple possible that the present invention proposes, efficiency is high, and precision of prediction is high, is a kind of effective Short-time Traffic Flow Forecasting Methods, and it predicts the outcome can provide foundation with the service of control for vehicle supervision department carries out traffic guidance.

Portable: gene search procedure is different from Classical forecast algorithm mechanism, only need a large amount of traffic flow data storehouses as support, just can predict fast and accurately current traffic flow situation.Therefore this method can be transplanted under the situation of different time interval, different vehicle flowrate radixes and predict.

Brief description of the drawings

Fig. 1 is gene search model Forecasting Methodology process flow diagram provided by the invention.

Embodiment

As shown in Figure 1, gene search Short-time Traffic Flow Forecasting Methods of the present invention comprises the following steps:

Step 1), data acquisition:

When traffic flow data is gathered, ensure to comprise in the data that collect a certain amount of nonworkdays data and great festivals or holidays data, i.e. the diversity of database.Because when every class traffic flow data is predicted, all need to just can make accurately prediction based on a certain amount of such data, the diversity of database can ensure nonworkdays and festivals or holidays forecasting traffic flow accuracy.

Step 2), historical traffic flow data coding:

Choose the assortment of genes of traffic flow data q (t-3), q (t-2), q (t-1), q (t) of continuous four time periods as a chromosome C, Duan Yue continuous time wherein choosing is many, and it is more accurate to predict.In chromosome C, comprise comparison genetic fragment D ₁with predicted gene fragment D ₂.D ₁by the gene of q (t-3), q (t-2), q (t-1) correspondence combine D ₂by gene corresponding to q (t) composition.Specific coding process is as follows:

Adopt binary coding to encode respectively to the traffic flow data of continuous four time periods.In the process of coding, the traffic flow data of establishing four time periods is three figure places, mends work 0 if indivedual data are double-digit hundred.The traffic flow digit order number of each time period is 9 to the maximum, and corresponding binary coding is 1001, so bits per inch word should be made up of tetrad code in the process of coding, the traffic flow numeral of each time period is 12 binary codes.

First of q (t-3), q (t-2), binary code that three traffic flow datas of q (t-1) are corresponding is successively placed on together, as 1-3 the gene loci of chromosome C.Wherein the first bit binary number of binary code corresponding to t-3 time period is as first gene loci of chromosome C, the first bit binary number of binary code corresponding to t-2 time period is as second gene loci of chromosome C, and the first bit binary number of string of binary characters corresponding to t-1 time period is as the 3rd gene loci of chromosome C.The second of binary code corresponding three traffic flow datas is successively placed on together, as chromosomal 4-6 gene loci.The rest may be inferred, completes the comparison genetic fragment D of chromosome C ₁coding work.The binary code that q (t) is corresponding is directly added in comparison genetic fragment D ₁after, as the predicted gene fragment D in chromosome C ₂.Successively the data of continuous four time periods of historical traffic flow are completed to coding work, set up short-time traffic flow forecast gene pool.

Using q (t-3), q (t-2), q (t-1) and q (t) separately encode be because the binary code of q (t-3), q (t-2), q (t-1) in the time that gene mate as comparing genetic fragment D ₁, for identifying fast the similar chromosome C of short-time traffic flow forecast gene pool.After the match is successful, the predicted gene fragment D that q (t) is corresponding ₂represent the magnitude of traffic flow of corresponding historical time section similar with the magnitude of traffic flow of current t time period in short-term traffic flow historical data base, therefore directly to predicted gene fragment D ₂decode and can obtain the magnitude of traffic flow of the corresponding historical time section similar with the current t time period magnitude of traffic flow, be convenient to next step and fast the magnitude of traffic flow of current t time period predicted.

Coding example is as follows, is provided with traffic flow sequence [181,186,203,251]:

To after 181 binary codings being:

000110000001

To after 186 binary codings being:

000110000110

To after 203 binary codings being:

001000000011

To after 251 binary codings being:

001001010001

First of the binary code of first three traffic flow data is 0, and the chromosome 1-3 position gene after restructuring is 000.Three binary code seconds are 0, and the chromosome 4-6 position gene after restructuring is 000.Three the 3rd of binary codes are 0,0,1, and the chromosome 7-9 position gene after restructuring is 001.The rest may be inferred compares genetic fragment D ₁be 000000001110110000000000000010011101.251 binary coding be 001001010001, it is as the predicted gene fragment D in chromosome ₂directly be added in comparison genetic fragment D ₁below.So the gene of final chromosome C is:

000000001110110000000000000010011101001001010001。

Step 3), current traffic flow data coding:

When the magnitude of traffic flow of current t time period is predicted, by t-3, t-2 before the t time period, the traffic flow data of tri-time periods of t-1 according to comparison genetic fragment D ₁coding method encode, obtain a chromosome Y after restructuring ₁; This chromogene is 36 binary codes; Wherein first 12 is that hundred corresponding binary coded value of traffic flow numeral combine, and middle 12 is that ten corresponding binary coded value of traffic flow numeral combine, and latter 12 is that a traffic flow numeral corresponding binary coded value combines.

Chromosome C in short-time traffic flow forecast gene pool is because of with predicted gene fragment D ₂so it is 48 binary codes.For the ease of with short-time traffic flow forecast gene pool in chromosome C compare, so to chromosome Y ₁repair, increase by 12 and be encoded to 0 amorph fragment 000000000000.

This genetic fragment, without in all senses, just increases for the ease of gene coupling.Chromosome after repairing is designated as Y, and it is called to match-on criterion chromosome.

Step 4), search for similar chromosome population:

While coupling in short-time traffic flow forecast gene pool, that utilizes chromosome C in front 36 genes of chromosome Y and short-time traffic flow forecast gene pool compares genetic fragment D ₁carry out Rapid matching, detailed process is as follows:

In short-time traffic flow forecast gene pool, search for the chromosome C identical with front 18 genes of match-on criterion chromosome Y as initial population (in the time that short-time traffic flow forecast gene pool is larger, can select the more homologous genes of multidigit).Initial population quantity is defined as P, in the time of P > 10, taking the 19th genic value of match-on criterion chromosome Y as standard, in initial population, continue search and the 19th the chromosome C that genic value is identical of match-on criterion chromosome Y, then again add up the sum of qualified chromosome C.If P is still greater than 10, taking the 20th genic value of match-on criterion chromosome Y as standard, until when P < 10, stop, this is initial population quantity.

Chromogene in short-time traffic flow forecast gene pool after binary coding is arranged according to binary code size order, and therefore initial population is a fixing continuum in short-time traffic flow forecast gene pool, and location is simple.

Step 5), similar chromosome deviate calculate:

In the initial population forming after search, each chromosome C _n(n is positive natural number and n < 10) is all different with the similarity of match-on criterion chromosome Y, and the non-similarity degree standard of weighing both is deviate Δ _nsize.

Each chromosome in initial population all contains 48 genes, and according to search condition, front 18 genes are just the same.48 genes are divided into 16 genetic fragments, and the deviate of each genetic fragment is designated as because just the same according to front 18 genes of search condition, front 6 genetic fragments are the same, and deviate is 0, therefore calculate since the 7th genetic fragment when calculation deviation value.

The binary number that every three gene pairss are answered is a genetic fragment, because in t-3, t-2, the corresponding binary code of t-1 time period traffic flow data, the position of the binary number that these three gene pairss are answered is the same, and in the time of calculation deviation value, they have congruency.For each chromosome C in initial population _narbitrary genetic fragment of (n is positive natural number and n < 10) calculating principle in the time of the deviate of calculating and match-on criterion chromosome Y is all different.Concrete formula is as follows:

Chromosome C in initial population P _n, n is positive natural number, and n < 10, a i genetic fragment and i genetic fragment of match-on criterion chromosome Y have a gene different (a=0,1,2,3), chromosome C in initial population P _ni its deviate of genetic fragment be:

In formula 2 ^bfor decimal system transforming relationship corresponding to this genetic fragment, 10 ^crepresent that this gene position is that binary coded value corresponding to ten of traffic flow numerals or individual position combines;

Chromosome C in initial population P _ntotal departure value:

${\mathrm{\&Delta;}}_n={\mathrm{\&Delta;}}_n^7+{\mathrm{\&Delta;}}_n^8+{\mathrm{\&Delta;}}_n^9+{\mathrm{\&Delta;}}_n^{10}+{\mathrm{\&Delta;}}_n^{11}+{\mathrm{\&Delta;}}_n^{12};$

Chromosome C in initial population _nthe 7th genetic fragment of (n is positive natural number and n < 10) is that the 7th genetic fragment of 19-21 position genic value and match-on criterion chromosome Y is that 19-21 position genic value has a the different (a=0 of gene, 1,2,3), chromosome C in initial population _nthe 7th its deviate of genetic fragment of (n is positive natural number and n < 10) is:

${\mathrm{\&Delta;}}_n^7=a\&times;2^1\&times;10,$

In formula 2 ¹for decimal system transforming relationship corresponding to this genetic fragment, 10 represent that this gene position is that ten corresponding binary coded value of traffic flow numeral combine.

Chromosome C in initial population _nthe 8th its deviate of genetic fragment of (n is positive natural number and n < 10) is:

${\mathrm{\&Delta;}}_n^8=b\&times;2^0\&times;10,$

B in formula (b=0,1,2,3) represents chromosome C in this initial population _nin the 8th genetic fragment of (n is positive natural number and n < 10) from gene number different in the 8th genetic fragment of match-on criterion chromosome Y, 2 ⁰for decimal system transforming relationship corresponding to this genetic fragment, 10 represent that this gene position is that ten corresponding binary coded value of traffic flow numeral combine.

Chromosome C in initial population _nthe 9th its deviate of genetic fragment of (n is positive natural number and n < 10) is:

${\mathrm{\&Delta;}}_n^9=b\&times;2^3\&times;1,$

C in formula (c=0,1,2,3) represents chromosome C in this initial population _nin the 9th genetic fragment of (n is positive natural number and n < 10) from gene number different in the 9th genetic fragment of match-on criterion chromosome Y, 2 ³for decimal system transforming relationship corresponding to this genetic fragment, 1 represents that this gene position is that a traffic flow numeral binary coded value that position is corresponding combines.

Chromosome C in initial population P _nthe 10th its deviate of genetic fragment of (n is positive natural number and n < 10) is:

${\mathrm{\&Delta;}}_n^{10}=d\&times;2^2\&times;1,$

D in formula (d=0,1,2,3) represents chromosome C in this initial population _nin the 10th genetic fragment of (n is positive natural number and n < 10) from gene number different in the 10th genetic fragment of match-on criterion chromosome Y, 2 ²for decimal system transforming relationship corresponding to this genetic fragment, 1 represents that this gene position is that a traffic flow numeral binary coded value that position is corresponding combines.

Chromosome C in initial population P _nthe 11st its deviate of genetic fragment of (n is positive natural number and n < 10) is:

${\mathrm{\&Delta;}}_n^{11}=e\&times;2^1\&times;1,$

E in formula (e=0,1,2,3) represents chromosome C in this initial population _nin the 11st genetic fragment of (n is positive natural number and n < 10) from gene number different in the 11st genetic fragment of match-on criterion chromosome Y, 2 ¹for decimal system transforming relationship corresponding to this genetic fragment, 1 represents that this gene position is that a traffic flow numeral binary coded value that position is corresponding combines.

Chromosome C in initial population P _nthe 12nd its deviate of genetic fragment of (n is positive natural number and n < 10) is:

${\mathrm{\&Delta;}}_n^{12}=f\&times;2^0\&times;1,$

F in formula (f=0,1,2,3) represents chromosome C in this initial population _nin the 12nd genetic fragment of (n is positive natural number and n < 10) from gene number different in the 12nd genetic fragment of match-on criterion chromosome Y, 2 ⁰for decimal system transforming relationship corresponding to this genetic fragment, 1 represents that this gene position is that a traffic flow numeral binary coded value that position is corresponding combines.

Chromosome C in initial population P _nthe total departure value of (n is positive natural number and n < 10):

${\mathrm{\&Delta;}}_n={\mathrm{\&Delta;}}_n^7+{\mathrm{\&Delta;}}_n^8+{\mathrm{\&Delta;}}_n^9+{\mathrm{\&Delta;}}_n^{10}+{\mathrm{\&Delta;}}_n^{11}+{\mathrm{\&Delta;}}_n^{12}.$

Example is as follows:

Be provided with traffic flow sequence [181,186,203,251], this traffic flow sequence chromogene is:

00000000111011000000000 00000 10 011 101001001010001

If the traffic flow situation of current continuous three time periods is:

[182183212] genic value after coding is:

000000001110000000001000000111010

Deviate is calculated as follows:

First after increasing the amorph position of 12 0, be:

00000000111011000000000 10000 00 111 010000000000000

Then mate with chromogene C:

00000000111011000000000 00000 10 011 101001001010001

By coupling can find out, two chromogenes front 18 just the same, therefore this chromosome is similar chromosome.Article two, chromosomal the 8th, 10,11 genetic fragments have a binary code different, and the 12nd genetic fragment has 3 binary codes different, show that according to each genetic fragment deviate computing formula this chromosomal deviate is:

Δ _n＝1×2 ⁰×10+1×2 ²×1+1×2 ¹×1+3×2 ⁰×1＝19。

Step 6), predicted gene fragment decoding and prediction:

Choose deviate Δ _nthree minimum chromosome C _n(n=1,2,3), by the predicted gene fragment D in every item chromosome ₂be decoded as decimal number and be respectively S ₁, S ₂, S ₃.They are the magnitude of traffic flow of time period similar to the current t time period magnitude of traffic flow in historical data base, and the value S that utilizes them to predict the magnitude of traffic flow of current t time period is:

$S=\frac{S_1+S_2+S_3}{3}.$

Example is as follows:

Three chromogenes that rear gene matching value maximum is calculated in design are respectively:

00000000ll101100000000000000l00l110l00l00l010001

00000000l1l0l100000000000l0l00000l0000l00l00l00l

00000000l1l0l1000000001000000l1l001100l00l0l0l0l

These three chromosomal predicted gene fragments are decoded, calculate 12 tens digits that binary code is corresponding after every chromosome.Computation rule is the corresponding decimal system traffic flow numeral of every tetrad code, after three chromosome predicted gene fragment decodings, is respectively 251,249,255, substitution formula can be calculated genetic search algorithm predicted value is 252.

When concrete enforcement, utilize toroid winding vehicle detection card to add up the magnitude of traffic flow of every day, statistical interval is 10min.Vehicle flowrate after statistics is utilized to matlab programming, set up short-time traffic flow forecast gene pool.In order to improve the precision of prediction, when traffic flow data is gathered, ensure to comprise in the data that collect a certain amount of nonworkdays data and great festivals or holidays data.

Forecasting traffic flow result is mainly used in road network and coordinates to control the dynamic traffic guidance of implementing and shielding based on LED.System in controlled region all main crossing and and part critical path on toroid winding vehicle detection card has been installed, for detecting in real time various transport information, be transferred to traffic control center by dedicated network, and be stored in traffic information database.

Claims (1)

1. a gene search Short-time Traffic Flow Forecasting Methods, is characterized in that, comprises the following steps:

Step 1), data acquisition:

When traffic flow data is gathered, ensure to comprise in the data that collect a certain amount of nonworkdays data and great festivals or holidays data, ensure the diversity of database;

Step 2), historical traffic flow data coding:

Utilize current t-3, t-2, the traffic flow data of t-1 time period is predicted the magnitude of traffic flow of t time period, choose the traffic flow data q (t-3) of continuous 4 time periods, q (t-2), q (t-1), the assortment of genes of q (t) is as a chromosome C, specifically for different traffic flow data storehouse sizes and the prediction accuracy requiring, select section traffic flow data continuous time of different length to predict current traffic flow situation, utilize t-m, t-m+1, t-2, the traffic flow data q (t-m) of t-1 time period, q (t-m+1), q (t-2), q (t-1) predicts the magnitude of traffic flow of t time period, wherein m is positive natural number, in chromosome C, comprise comparison genetic fragment D ₁with predicted gene fragment D ₂, D ₁by the gene of q (t-3), q (t-2), q (t-1) correspondence combine D ₂by gene corresponding to q (t) composition,

Step 3), current traffic flow data coding:

When the current t time period, the magnitude of traffic flow was predicted, t-3, the t-2 before the t time period, the assortment of genes of the traffic flow data of tri-time periods of t-1 are as a chromosome Y ₁this chromogene is 36 binary coded value, wherein first 12 is that hundred corresponding binary coded value of traffic flow numeral combine, middle 12 is that ten corresponding binary coded value of traffic flow numeral combine, and latter 12 is that a traffic flow numeral corresponding binary coded value combines;

Chromosome C in short-time traffic flow forecast gene pool is because of with predicted gene fragment D ₂so it is 48 binary coded value; For the ease of comparing with chromosome C, to chromosome Y ₁repair, increase by 12 and be encoded to 0 amorph fragment 000000000000, the chromosome after repairing is 48 binary coded value, is designated as match-on criterion chromosome Y;

Step 4), search for similar chromosome population:

In short-time traffic flow forecast gene pool, search for the chromosome C identical with front 18 genes of match-on criterion chromosome Y as initial population, in the time that short-time traffic flow forecast gene pool is larger, select the more homologous genes of multidigit, to improve search efficiency, initial population quantity is defined as P, in the time of P > 10, taking the 19th genic value of match-on criterion chromosome Y as standard, in initial population, continue the search chromosome C identical with the 19th genic value of match-on criterion chromosome Y, then again add up qualified similar chromosome sum, if P is still greater than 10, taking the 20th genic value of match-on criterion chromosome Y as standard, until when P < 10, stop, this is initial population quantity,

Step 5), similar chromosome deviate calculate:

Each chromosome in initial population all contains 48 genes, and according to search condition, front 18 genes are just the same, and 48 genes of chromosome C are divided into 16 genetic fragments, and every three genes are a genetic fragment, and the deviate of each genetic fragment is designated as because front 18 genes are just the same, front 6 genetic fragments are the same, when calculation deviation value, calculate since the 7th genetic fragment;

Every three genes are a genetic fragment, are that the position of the binary value that these three gene pairss are answered is the same because in t-3, t-2, corresponding traffic flow data binary code of t-1 time period, and in the time of calculation deviation value, they have congruency; For each chromosome C in initial population _n, n is positive natural number, and n < 10, chromosome C _narbitrary genetic fragment calculating principle in the time calculating with the deviation value of match-on criterion chromosome Y all different; Concrete formula is as follows:

Chromosome C in initial population P _n, n is positive natural number, and n < 10, a i genetic fragment and i genetic fragment of match-on criterion chromosome Y have a gene different (a=0,1,2,3), chromosome C in initial population P _ni its deviate of genetic fragment be:

In formula 2 ^bfor decimal system transforming relationship corresponding to this genetic fragment, 10 ^crepresent that this gene position is that binary coded value corresponding to ten of traffic flow numerals or individual position combines;

Chromosome C in initial population P _ntotal departure value:

Step 6), predicted gene fragment decoding and prediction:

Choose deviate Δ _nthree minimum chromosome C _n, n=1,2,3, by the predicted gene fragment D in every item chromosome ₂be decoded as decimal number and be respectively S ₁, S ₂, S ₃, they are the magnitude of traffic flow of time period similar to the current t time period magnitude of traffic flow in historical data base, the value S that utilizes them to predict the magnitude of traffic flow of current t time period is: