Modeling weekly representative activity patterns of population groups: a bioinspired multiple sequence alignment approach

This research introduces a multi-module framework to derive weekly representative travel patterns from single-day travel diaries. The methodology first uses hierarchical clustering to group samples with similar activity patterns, followed by progressive multiple sequence alignment to construct day-level representative patterns. These day-level patterns are then merged based on their similarity to create week-level representative activity patterns, ultimately producing archetypal weekly pseudo-diaries. The proposed approach accounts for sequential patterns, activity transitions, and cross-day similarities, providing deeper insights into travel behavior beyond traditional statistical methods. Analyzing weekly activity patterns from this longitudinal data revealed significant insights into travel behavior, including distinct work and non-work patterns across the week. For working groups, shorter work durations on Fridays were observed, and the weekly work duration for teleworkers was found to be lower than that of workplace workers. Although the detailed exploration of weekly activity and time-use patterns provides valuable policy insights, this research primarily focuses on advancing activity-based travel demand modeling by introducing a mathematically robust approach to capturing sequential and temporal activity patterns.