Enhancements for Visualizing Temporal and Geospatial Datasets

Abstract

In this thesis, we will discuss enhancements for the analysis and visualization of temporal and geospatial data. Techniques for both research domains have a long history and wide applicability, but at the same time suffer from basic issues as overplotting or hidden patterns. In combination, space and time are even more challenging with respect to the Visual Analytics design however enable new perspectives. The main idea of all enhancements presented in this thesis is focusing on interesting aspects of the data and visually conveying concepts by abstraction. The importance is in our case defined by subject matter experts and consequently our methods are parametrized in a way allowing user influence. In detail, we will improve analysis, prediction, and visualization techniques for time series by mechanisms enhancing the visual saliency of important points in time. Additionally, our goal is to implement inspectable models and explain why our system believes something being important for the analyst. As a second step, we investigate how to enhance geospatial visualizations avoiding and reducing overplotting issues. Overplotting often occurs in geospatial visualizations because of unequal density distributions. We discuss techniques to reduce overplotting in point-based visualizations and present simplifying methods for line-based representations, as in general removing all overplotting lines is not possible. Combinations of both geospatial and temporal data are analyzed in the domain of recorded soccer data. We enhance the way domain experts analyze soccer matches and present methods enabling the expert to focus only on the interesting parts of a match by appropriate Visual Analytics techniques.