Traffic Prediction Based on Correlation of Road Sections

Road section data packet is very necessary for the estimation and prediction in short-time traffic condition. However, previous researches on this problem are lack of quantitative analysis. A section correlation analyzing method with traffic flow microwave data is proposed for this problem. It is based on the metric multidimensional scaling theory. With a dissimilarity matrix, scalar product matrix can be calculated. Subsequently, a reconstructing matrix of section traffic flow could be got with principal components factor analysis, which could display section groups in low dimension. It is verified that the new method is reliable and effective. After that, Auto Regressive Moving Average (A RMA) model is used for forecasting traffic flow and lane occupancy. Finally, a simulated example has shown that the technique is effective and exact. The theoretical analysis indicates that the forecasting model and algorithms have a broad prospect for practical application.