Cluster Analysis of Long-Distance Person Travel in Alberta

Abstract

Transportation is the movement of goods and people throughout a network of links and nodes and consists of short-distance SD (intra-city) and long-distance LD (inter-city) trips. A component of the latter which is responsible for a relatively large portion of total kilometres travelled – that is long-distance person trips LDPT not including LD trucking – has yet to be profoundly investigated, measured, and modelled to become manageable and benefit relevant policies and funding allocations. In LDPT modelling, the traditional methods of LD trips representation are those used for SD trips modelling in which some explanatory factors such trip purpose are selected for LDPT segmentation as a means to distinguish among sectors of such trips. This in turn has resulted in the development of LDPT models that fall short of correctly representing such trips. In this regard, the work described herein hypothesizes and demonstrates how LD trips are not merely longer versions of SD trips, but their relevant datasets possess a natural structure and inherent groupings or clusters of trips that needs to be deeply investigated using appropriate methods (cluster analysis used in computer science studies for network related data) in order to contribute to much needed refinements of LDPT modelling exercises. The method along with several novel avenues of input selection, validity, and significance measurements, revealed ten heterogeneous clusters of LDPT to exist in the 2010 Alberta to Alberta trips in the 2010 Travel Survey of Residents of Canada TSRC dataset. The clusters exhibiting a mix of trip purpose, person, and activity features, with proportions of total person trips shown in brackets are: short economical getaways (31%), same-day shopping (22%), personal business (12%), visiting friends and relatives (10%), business/casino trips (9%), young adult team sport players (4%), same-day trips of snow/festival loving young families with kids (4%), costly cottage trips (3%), high educated multiple city visitors (3%), and seniors with medical appointments (2%). The existence of clusters which were detected through the approach proposed herein carries a fundamental meaning showing there exist different segments of LD travel behaviour in reality which need to be studied if one is to accurately model them.