Visual Analytics Approaches to Exploring Multivariate Time Series Data

Time-oriented data analysis has attracted the attention of researchers for decades, across many research domains, including but not limited to medical records, business, science, engineering, biographies, history, planning, and project management. However, the complexities of time-oriented data with a large number of variables and varying time scales hinder scientists from completing more than the most basic analyses. In this dissertation, I present two design studies where multivariate time series data are involved. In the first design study, I developed an interactive interface, \textit{t}-RadViz, for a manufacturing company to visually monitor and analyze real-time streaming multivariate testbench data with continuous numeric values. In the second design study, I developed a visual analytics prototype named EVis for analyzing and exploring how recurring environmentally driven events are related to high dimensional time series of continuous numeric environmental variables. In both design studies, I closely collaborated with domain users in the whole process of requirement analysis, design, and evaluation. Besides a rich set of fundamental graphic charts for supporting basic analysis functions, new visual analytics techniques were developed in the design studies for addressing challenging tasks, such as a novel trajectory-based multivariate time series visual analytics approach in EVis for exploring temporally lagging relationships between events and antecedent conditions. The effectiveness and efficiency of the prototypes are illustrated by case studies conducted with real users and feedback from domain experts.