We propose a travel time forecast method for nationwide traffic information services. This method can deal with various "day factors" which traffic condition depends on, e.g. days, seasons, and vacations, without complicated local parameter setting. It also has the advantage in both the size of the database and the amount of the calculation for forecasting. The basic idea of the method is that travel time data consists of the weighted sum of several "feature bases" which represent characteristic traffic patterns such as congestion in morning or evening. The feature bases are the information which is independent of the day factors; in contrast, the information to be forecasted according to the day factors is weighting coefficients of the feature bases. The feature bases are given by the feature space projection using principal component analysis, and the forecasting process for the weighting coefficients is performed in the feature space. Since the dimension of the feature space is smaller than that of the travel time data space, the method can achieve an accurate forecasting process with a small database and a little amount of calculation. The information of the feature bases are restored to the forecast data of travel time by the inverse projection from the feature space. Evaluation results show that the method efficiently improves the forecast accuracy for wide-area applications.
Development of long-term travel time forecast method for nationwide traffic information services
Titel japanisch
Transactions of the Information Processing Society of Japan ; 45 , 12 ; 2696-2705
2004
10 Seiten, 10 Quellen
Article (Journal)
Japanese
Short-Term Traffic Forecast System of Beijing
Transportation Research Record | 2010
|Automotive engineering | 2000
|Traffic estimation method using average travel time information and traffic estimation device
European Patent Office | 2020