The Research Of Traffic Flow Based On Logarithmic Transformation

With the development of the social progress and economic development, the demand for transportation is growing. However, the contradiction between the situation of the existing traffic road and the growth of the traffic flow is increasingly severe. A real-time and accurate forecast is necessary to the traffic control and induction. The theoretical research shows that the intelligent transportation system can solve this problem and the main method is based on the parameters forecasting and the incident detection about the traffic flow. So, a better traffic flow parameters and better traffic incident detection method are the serious problems to be solved.The main contents and results of this paper are as follows:(1)The wavelet thresholding through the logarithmic transformation has been studied. In the research, the abnormal points are applied on the standard signal, when the abnormal value is in the range of [-0.7340, 0.4314], the range becomes [-0.7428, 0.4685] after logarithmic transformation, the length changes from 1.1654 to 1.2113, namely 3.94% growth. It shows that the non-linear transform has some good effect on the wavelet thresholding, and further some theoretical analysis is conducted. The research belongs to stationary method research of "mechanism + identification" prediction strategy.(2)The mathematical proof shows that the logarithmic transformation can improve the wavelet de-noising effect. The Taylor expansion of the logarithmic transformation reflects its effects on the non-linear: the logarithmic transformation can reduce the MSE(Mean Squared Error), but increase the ME(Mean Error), which are verified by the numerical experiments and the examples of traffic flow prediction.(3)The comparison about the old and new de-noising methods of traffic flow and the five prediction models shows that the non-linear transformation is good for wavelet de-noising. The signal has been greatly improved after non-linear transformation.