Research On Urban Short-time Traffic State Forecasting Model Based On Floating Car Technology

In recent years, the number of vehicle is growing fast with the development of society economy. It makes the traffic more convenient, but the phenomenon of the road congestion is more common and serious. Limited in land resource, it cannot satisfy the traffic demand by expanding the road in cities. Therefore, the intelligent transportation system has become popular to solve the congestion issue in the whole world. The short-term traffic state forecasting is the key of traffic guidance and control in ITS. There are a lot of methods of short-term traffic state forecasting application, and got a certain achievements. As most of the methods are based on the fixed detector to forecast, many of them can hardly adapt to the characteristics of the floating car data. The prediction model based on floating car data prefers to ignore the missing data, which causes the low prediction accuracy, so that it cannot satisfy the real needs.In order to improve the accuracy of the short-term traffic state forecasting, the method that can fill the vacancy is given in the paper, which is based on the historical and real-time data. Then it chooses the different forecasting models according to the loss situation of the input layer data. The travel speed is selected as the characterization of traffic state parameter in the paper. The preprocessing methods given, which includes filtering, fitting, filling vacancy and denoising. For the part of the filling vacancy, it is used K-means method based on historical and real-time data. The result is verified by actual data. Depending on the analysis of the spatial and temporal correlation of the travel speed, short-time traffic state is forecasted respectively based on the time, space and space-time dimension data. The forecasted results are verified with part of the actual data of taxies in Dalian as the floating car data. The comprehensive prediction model is realized based on the loss data of input layer.From the experience results, the proposed model can estimate the short-time traffic state of Dalian well, which has the higher accuracy and reliability. The result of verified examples data also meets the real traffic condition of Dalian, which can satisfy the requirement of forecasting the short-time traffic condition for the travelers and proves the theoretical and practical values of the method in the paper.