抄録
To identify travel patterns and examine their influential factors in the context of developing countries, this paper first develops a hybrid model, which consists of a data mining approach (Exhaustive CHAID), an aggregate logit model and a structural equation model. Data mining approach is used to systematically identify travel patterns. Aggregate logit model is used to describe the zonal shares of travel patterns, reflecting the stochastic characteristics of travel pattern choices. Structural equation model is adopted to represent not only the correlated structure of unobserved factors related to choice of travel patterns, but also the influence of complex cause-effect relationships in a much more sophisticatedly statistical and practical way. The effectiveness of the proposed hybrid model is then examined using person-trip data from 10 developing cities collected by JICA. Furthermore, a comparative analysis is conducted in order to explore the similarities and differences among travel patterns in different cities.
