Pattern recognition of daily activity patterns using human mobility motifs and sequence analysis

In this paper, we develop a new joint pattern recognition method that combines network motif-based analysis with activity sequence-based analysis. We use the advantages of both methods in creating groups of patterns that have within them distinct pattern homogeneity and across-pattern heterogeneity. The first portion of the analysis here applies a more traditional approach to identify unique network motifs, with 16 of them capturing 83.05% of the 2017 NHTS-California workday data. Multivariate analysis of grouped motifs data shows different preference of motifs for students, part-time workers, retirees, telecommuters, drivers, women, and younger adults. In the second portion of the analysis, motifs are grouped into categories based on the number of locations a person visits in a day and their correlation with time use and travel is explored. Time use and travel are analyzed based on minute-by-minute time allocation pattern identification using sequence analysis and hierarchical clustering. The correlation between motifs group and sequence analysis finds substantial heterogeneity within the motif groups. The within motif group clusters of activity-based sequences show typical commuting, going to school, and resting patterns. We also find seven patterns that are not typical but have similarities across motifs in their temporal footprint and the variety of activities in each sequence. The paper provides a summary of the analytical steps and findings as well as next steps.