Developing a Parsimonious Classification of Traffic Analysis Zones Using a Large Number of Accessibility Indicators and Transportation Level of Service.

Accessibility, the ease or difficulty with which activity opportunities can be reached from a given location, can be measured using the cumulative number of opportunities from an origin within a given amount of travel time and/or distance. Estimating accessibility indicators is used to show the level of service quality offered by a transportation system and as explanatory factors of travel behavior equations in transportation modeling and simulation. With the availability of large quantities of data, detailed representation of transportation networks (a set of nodes and links forming a graph), and level of service often measured in movement speed on network links, analysts can compute hundreds of accessibility indicators. Reporting of all these indicators is tedious, cumbersome, and too complicated. In this paper, the combination of principal component analysis to extract summaries of service quality from 190 indicators is demonstrated. Each of these indicators is computed for 1822 traffic analysis zones (TAZ - a convenient geographic subdivision used to model transportation systems) covering the entire country of Qatar in the Arabian Peninsula. Then, a small number of principal components (e.g., 27 components capturing 91.6% of the variance of 190 variables) are used to classify each TAZ in a group with clearly identifiable characteristics using hierarchical clustering. In this paper experiments with different subsets of the 190 indicators are also used to show how one can distinguish among different levels of service offered at each of the TAZs analyzed by time of day and means of travel.