Reduction of a Forecasting Database for Nationwide Traffic Information Service

摘要

We propose an enhanced method of "Feature Space Projection Method Method" for long-term traffic information forecasting. The traditional method can deal with various "day factors factors" which traffic condition depends on. Such as days, seasons, and vacations. It also achie achieve vess an accurate forecasting with aa little amount of calculation. Now, internet service using this method provides drivers with the traffic information forecast. However, it is difficult to apply this method to PC for home use and in-car navigation unit because the size of aa forecasting database is large. This paper in introduce troduce the method that decreases the size of the forecasting database greatly by performing forecast process in the feature space shared by several links keeping the accuracy of forecast data. In case of Japanese nationwide database, it makes one-tenth as large as that of usual one got by regression analysis. Using this new method, it can decrease the total size of Japanese forecasting database to 820MB.