Modeling Human Mobility Entropy as a Function of Spatial and Temporal Quantizations

Abstract

The knowledge of human mobility is an integral component of several different branches of research and planning, including delay tolerant network routing, cellular network planning, disease prevention, and urban planning. The uncertainty associated with a person's movement plays a central role in movement predictability studies. The uncertainty can be quantified in a succinct manner using entropy rate, which is based on the information theoretic entropy. The entropy rate is usually calculated from past mobility traces. While the uncertainty, and therefore, the entropy rate depend on the human behavior, the entropy rate is not invariant to spatial resolution and sampling interval employed to collect mobility traces. The entropy rate of a person is a manifestation of the observable features in the person's mobility traces. Like entropy rate, these features are also dependent on spatio-temporal quantization. Different mobility studies are carried out using different spatio-temporal quantization, which can obscure the behavioral differences of the study populations. But these behavioral differences are important for population-specific planning. The goal of dissertation is to develop a theoretical model that will address this shortcoming of mobility studies by separating parameters pertaining to human behavior from the spatial and temporal parameters.