The High Frequency Traffic Flow Analysis

The statistical analysis of traffic flow is the fundamental of road traffic forecasting with data mining. In the common traffic flow theory, macroscopic flow model and data presentation shows chaos and stochastic distribution. However, in the microscopic based traffic model, it is revealed that in high frequency sampled test, the variables such as flow, density, and lane usage have common trend of U shape in daytime, e.g. high in peak time and low in off-peak time. Based on the traffic database of cross-river channel in LuJiaZui, PuDong, Shanghai, the statistics method is used for different sampling frequency to validate the mean, standard variance, and kurtosis. The variant ratio is used to analyze the first order negative correlation among high frequency sampled traffic data. The result shows the models (e.g. ARMA model) used in low frequency sampling data cannot directly apply to high frequency samples. In order to validate the U shape distribution of traffic data in daytime, the wavelet based nerve network is used for the analysis of `calendar phenomena´ of cross-river tunnel traffic.