2
Abstract
Abstract
Abstract
Abstract
The prediction of short time traffic flow is the key of the Traffic Guidance
System(TGS),and it
’
s an important part of the Intelligent Transportation System(ITS).There
have been many research methods for short time traffic flow prediction. The paper studies the
theory of wavelet transformation and algorithms of decompose and reconstruction with Mallat,
and the theory of Gray-Markove forecasting model and the methods of prediction in detail.
Then we put forward the Gray-Markove forecasting Model based on wavelet transformation
with Daubechies 5 in Matlab.
Th
e basic thought of this paper is followed:
Firstly, with the wavelet toolbox in the Matlab we decompose the original traffic flow
data in 5 levels with the db5 and we can get the 5 detail parts and 1 aprroximation part. Then
we reconstruct the approximation part to the original level, and so do
th
e detail parts.
Secondly, we make the Gray–markov forecasting model on the reconstructed
aprroximation part .This paper use the GM(1,1),which is a single factor model to simulate the
history data and predict the traffic flow data of the next signal period. On the basis of the
fitting data ,we can acquire the errors .And now according to the fitting errors we partition to
five states and work out the transition probability of several shifts from one to five. According
to the the transition probability and the times that every state appears, we can verify whether or
not traffic flow data can satisfy the exclusive characteristic of the Markov Model that the
prediction is independent of the history. Experiment prove that the traffic flow data fill this
exclusive characteristic. .In this paper ,we apply the Markov
’
s chain Model with weight to
forecast the traffic flow. According to the transition probability matrix, the autocorrelation
coefficients can be calculated, and then we can get the weights of every rank .On the basis of
the results above, we can calculate the transition probability with weight. The maximum in the
transition probability matrix is the forecasting state of the error of the next signal period. Then
we can work out the final predicted traffic flow data.
In this paper the experimental data comes from Taian of Shandong province, including the
data acquired in the locale and the traffic monitoring video. From all the data we pick up 200
traffic flow data of continuous signal periods at the Post port of Dongyue Street to test the
