Extension of the 24-hour activity models to weeklong models and generating better routine activity skeletons, which are later filled in with non-routine activity episodes are identified as two areas of improvement in current activity-based modeling techniques. This paper utilizes multidimensional sequence alignment methods to measure similarities between routine weekly activity sequences of 282 surveyed individuals, as reported in a specialized survey of routine weekly schedules conducted in Toronto, Canada. Similar activity patterns are classified into a number of clusters. General behavioral patterns of the resulting clusters are described and analyzed based on socioeconomic activities of members of each cluster. Significant differences are found in a variety of socioeconomic variables that describe individual membership in each cluster, including age, income, gender, employment status, student status, marital status, drivers license, cell phoneusage and education level.
Abstract