Destination and activity estimation

Household travel surveys contain, besides travel information, sociodemographic information which plays an important role in analyzing travel dynamics. Disadvantages of such surveys are though that these are costly to conduct, time-consuming and often do not cover more than 2-3% of the population leading to possible biases and errors in reporting. On the other hand, smart card data provide real-time, accurate and detailed on board transactions records of each user, however, only cover a subset of all trips. Further, smart card data do not usually contain socio-demographic information and hence, one of the most important parameter “trip purpose or activity” is missing. This problem is of growing interest and several algorithms to infer the trip purpose are proposed. In this chapter, the discussion is on trip destination and purpose estimation methods and a 1-day Household Person Travel Diary Survey (HHPTD) is utilized to generate a framework for assigning trip purposes to trips. Activities studied are “home”, “work”, “school”, “academy” and “shopping”. Decision trees and rule-based models, namely, R-Tree and C50, are used to estimate the trip purpose. HHPTD is utilized to train the datasets and applied on the smart card dataset to calculate the probabilities of different activities. Activity duration, location and start time are the parameters which were analysed for developing the inference framework. The spatial context of this study is Seoul Metropolitan Area (SMA) with a focus on Seoul City. The smart card data analysed in this study dates back to June 11, 2012 and over 28 million transactions were made during that day.