Performance evaluation of travel demand forecasting models for transportation network analysis

Abstract

This paper presented a performance evaluation of travel demand forecasting techniques on transportation networks in Upper Northern Provincial Cluster 2, Thailand. The study compared multiple regression analysis and four-step sequential decision models. The findings revealed that the four-step sequential decision model forecasted person-trip generation in the study area to be 346,506, 373,422, 404,356, and 440,132 person-trips/day for the years 2029, 2034, 2039, and 2044, respectively. In comparison, the multiple regression model predicted approximately 320,245, 328,678, 338,123, and 349,567 person-trips/day for the same years, showing differences of 8.20%, 13.61%, 19.59%, and 25.91%, respectively. This variation can be attributed to the four-step sequential decision model's superior capability in comprehensively considering the impacts of future infrastructure development projects in the area compared to the multiple regression model. While both models forecast total person-trip generation, the four-step model additionally provides spatial distribution, modal allocation, and network assignment of these trips, enabling detailed analysis of traffic volumes on specific corridors. However, when evaluating model development convenience and time requirements, the multiple regression analysis approach offers faster problem-solving capabilities due to its more straightforward development process, while providing reasonably accurate forecasts of total person-trips.