A Novel System of Disaggregate Models for Travel Demand Modelling, Using Decision Tree and Random Forest Concepts

摘要: This research attempts to address the gap between and practical transport demand models. Evolution of travel models started from early paper pencil versions conventional four-step late 50s proceeded activity-based During this transition emphasis shifted aggregate disaggregate models, whereby researchers increasingly paid attention individual decision making regarding daily activities. alteration engendered a conspicuous disparity their practicality, precision policy sensitivity level. is main paper. Unlike highly modelling structure proposed in which trip purpose, mode transport, time day, commute distance attributes destination are modelled. method computational efficient cost effective it attractive for small or medium size cities. Destination choice problem decomposed framework by using rather than its actual location. All decisions modelled Decision Tree (DT), Modified version (MDT), Random Forest (RF). The model estimation practice done Victorian Integrated Survey Travel Activity (VISTA) 2007 as training set, transferability examined utilizing 2009 year same dataset. As results show RF DT system have high accuracy, whereas MDT predict patterns test dataset more similar observations.