Investigating the effects of daily travel time patterns on short-term prediction

Abstract

Various studies on dealing with the prediction of traffic variables have been conducted in the field of intelligent transportation systems (ITS) because of its significant role in facilitating traffic operation and management strategies. In particular, travel time has been recognized as an important parameter of such traffic variables for fully facilitating many ITS applications. However, an important research issue that existing studies have disregarded is to understand the characteristics of travel time data prior to performing traffic predictions. This study attempts to characterize the temporal evolution of travel times (referred to in this paper as travel time patterns), and explore the relationship between the characteristics and the prediction accuracy. Entropies used in information theory are utilized to characterize travel time data obtained from an advanced traffic surveillance system. Furthermore, three different types of prediction techniques are employed to perform one-step-ahead travel time prediction. Statistically modeled relationships based on regression analysis imply that better prediction can be performed by identifying travel time patterns. It is believed that the proposed approach would effectively support to conducting better travel time prediction.