A Framework for Analysing Trajectories of Movement in a Dynamic Geographic Context

Recent advances in location-aware technologies enable the collection of trajectories of moving entities, which can be useful in different application domains such as urban planning, transportation and environment management. The analysis of these trajectories has mainly focused on discovering movement patterns. However, the usefulness of the discovered patterns depends on the possibility to interpret and understand them. Recent studies have shown that the consideration of the movement context, while analysing trajectories, has the potential to support the understanding of movement patterns. However, the integration of movement context into the analysis of trajectories is still in its infancy and most of the available work considers only a static geographic context. This thesis develops a comprehensive conceptual and methodological framework for integrating a dynamic geographic context into the analysis of trajectories. In the first step, a conceptual model relating the movement to its dynamic geographic context is developed. The thesis establishes a classification of geographic context elements and then proposes a set of qualitative relations, termed movement interactions, between the movement and the context. In the second step, the thesis proposes an analysis framework which exploits the conceptual model developed. The analysis framework is based on the process of Knowledge Discovery in Database (KDD). The thesis focuses on two steps, which correspond to the steps of the KDD process aimed at discovering and interpreting patterns. The first step applies data mining and spatial analysis methods to extract interactions from trajectories and context data. The second step quantifies the extracted interactions and explores the correlation or dependence between the quantified interactions and dynamic attributes of the movement and the context. In order to evaluate the framework developed, the thesis executes three experiments using real trajectories of vehicle movement in urban environment. Each experiment focusses on specific challenges addressed by the thesis. The first experiment focuses on the temporal dynamics of the dynamic geographic context while the second experiment focusses on its spatial dynamics. While the first two experiments involve context data in pattern discovery, the third experiment involves context data for post-processing already discovered patterns. The experiments show that the integration of context data supports not only the interpretation of movement patterns but also a deeper understanding of the movement context. Furthermore, the experiments show that context data can be integrated at the pattern discovery stage or for post-processing already discovered patterns. The choice of the integration step depends on the data being analysed and the type of patterns being mined