Activity Patterns Identification Based on National Household Travel Survey Data

To better understand residents' activity patterns, it is necessary to make detailed analyses on individuals' activity information. This paper first extracts activity chains of residents in the New York metropolitan area from National household travel surveys (NHTS) data including 1995, 2001, 2009, and 2017, and determines the main activity types for 48 half-hour segments in a day. Secondly, based on the residents' activity sequences, the latent dirichlet allocation (LDA) topic model has been applied to obtain the activity-topic distribution, from which the activity patterns are identified with affinity propagation (AP) clustering method. Lastly, residents' representative activity patterns (RAPs) have been identified and compared with the indices of activity characteristics and social demographic attributes of residents. The results show that the RAPs of residents in New York metropolitan area have certain commonalities. For example, there are work-regular (W-R), school (S), and home (H) activity patterns on weekdays during 1995-2017. In addition, other activity patterns emerged, such as work-late return (W-L) and business (B) activity patterns.