A novel pattern recognition technique to characterize multi-day shopping and entertainment trip activities

Abstract

The driving force behind individuals’ travel behavior is closely linked to the need to engage in various activities, such as working, shopping, and entertainment. While the importance of shopping and entertainment activities is well-documented in activity-based modeling research, there is no existing literature specifically addressing different shopping activities and entertainment trips over long time periods, such as an entire week, with a granular level of investigation. This study introduces a novel framework using comprehensive pattern recognition modeling, aiming to identify households’ level weekly shopping and entertainment trip activity patterns and to identify their representative patterns. Utilizing data from the 2019 Puget Sound Regional Household Travel Survey, the one-week activity patterns are split into 336 30-minute intervals. Each interval is comprised of information on trip activity types, duration, and start time. Pattern complexity of activity sequences in the dataset is recognized using the two-staged clustering process involving affinity propagation (AP) and k-means algorithms, which results in six unique clusters of homogeneous weekly activity patterns. These clusters exhibit a heterogeneous diversity in the temporal distribution of trip activity durations and significant differences in a variety of sociodemographic variables. Moreover, using sequence alignment techniques, we identified the representative trip activity pattern of the households in each cluster. Notably, younger individuals tend to shop on weekends, while older adults (age 65+) maintain consistent daily shopping habits. Households with higher incomes and vehicle access typically shop midweek, whereas a significant portion of high-income households without vehicles opt for Monday shopping. This comprehensive analysis highlights the intricate relationship between recreational travel behavior and sociodemographic factors, shedding light on nuanced patterns of activity engagement over extended time periods.