Examining travelers "optimal strategies" in transit trip choices, applying a classification tree approach on transit quality of service indicators

Abstract

According to Spiess and Florian (1989) [1], at each transfer point passengers may shape the "optimal" for them strategy taking into account minimum generalized travel cost. The generalized cost of trip choices is formulated based on objective and subjective criteria of travelers that reflect their perception of the generalized cost of their trip. Quality of Service (QS) parameters affect users?perceptions and may influence the formulation of the perceived cost and furthermore may influence travel choices. This paper examines the transit QS formulation by analyzing 26 quality indicators which have been assessed by transit users through a Customer Satisfaction Survey (CSS). The survey took place in the city of Thessaloniki, Greece in the summer of 2012. This paper examines travelers optimal strategies based on transit QS by applying two classification trees on transit quality of service indicators. A Classification and Regression tree has been used to classify the important quality indicators among male and female travelers and a J48 classification algorithm has been applied to classify all quality attributes based on their importance in the general assessment formulation of transit QS. Research pointed out that stop location is the most important attribute to a gender classification of transit QS attributes, followed by the indicator of information by phone, internet. For the general assessment of the transit QS as perceived by transit users, the root node of the classification tree is the security against crimes on bus, which denotes the most important predictor in such an analysis.