Modeling Daily Traffic Counts: Analyzing the Effects of Holidays

In this chapter, two modeling philosophies for forecasting daily traffic counts are compared. The starting point for the first philosophy is the fact that successive traffic counts are correlated, and that therefore past values provide a solid base to forecast future traffic counts. The second philosophy presupposes that daily traffic counts can be explained by other variables. Special attention is paid to the investigation of holiday effects. The analysis is performed on data originating from single inductive loop detectors, collected in 2003, 2004 and 2005. Results from both modeling philosophies show that weekly cycles predetermine the variability in daily traffic counts. The Box-Tiao modeling approach, which exploits the underlying preposition that explanatory variables can be used for forecasting future traffic counts, provides the required framework to quantify holiday effects. The results indicate that daily traffic counts are significantly reduced during holiday periods. When the forecasting performance of the different modeling techniques was assessed, the Box-Tiao modeling approach outperformed the other modeling strategies, especially when a large forecast horizon was considered. Simultaneous modeling of travel motives and revealed traffic patterns is a key challenge for further research.