Application of Two-Step Discriminant Model on Multi-Choice Urban Chained Travel Analysis － A Case Study of Taipei Metropolitan Area

• Main Authors: Shih-Hsien Yang, 楊士賢
• Other Authors: Yu-Chun Liao
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/6u5m6s

碩士 === 中原大學 === 土木工程研究所 === 91 === In order to participate in planned activities within the limited time budget and to maximize the total utility, the urban commuters tend to link trips into trip-chains to reduce negative effects induced by wasteful travel. The trip makers’ arrangements of daily activities and trip chains mainly depend on capability constraints, coupling constraints, and authority constraints; therefore individuals make different activity/travel patterns accordingly. This study was based on the activity theory and incorporated the concept of trip chaining by using “tours” as analysis units instead of trips. The process of research was comprised of three successive sections: (1) pattern identification, (2) influence factor specification and (3) model development. The study was expected to present additional details useful to understand the complicated travel behavior of urban commuters. The conventional approach in travel behavior studies mostly use the logit models to estimate the possible choices of individuals. Since the logit model is probability-based, the utility function of each tour pattern is hard to formulate. By using discriminant analysis, not only can the utility function of each tour pattern been obtained, but also the possible choices been recognized by using multi-logit functions. The choice of variables in the model was guided by previous empirical studies. In addition to the socio-economic characteristics and the household-member interrelationships, we used the cluster analysis to emphasize that individual’s trip scheduling was constrained by time window factors. Based on these hypotheses, the two-step model was established to explore the effect of time windows on improving the model accuracies. By using the travel data collected in Taipei metropolitan area in year 2000, the modeling results showed that in addition to the conventional attributes, the time window variables played an important role in category discrimination. Furthermore, the complexity of trip scheduling growed with the increasing trip legs in the tour, and the biased estimation was therefore frequently generated between patterns with similar complexity. Acknowledging this problem, the cluster analysis was adopted to revise tour patterns by joint consideration of trip scheduling and time windows; the second-step modification largely rectified the problems mentioned above and the results were satisfied.