Entwicklung und Bewertung eines Ansatzes zur visuellen Analyse in zeitabhängigen Sequenzdaten

This diploma thesis considers the possibility of a visual analysis of time series consisting of discrete symbols. It aims at finding and evaluating suitable visualizations for efficient pattern identification by the human eye. Pattern visualization will use the notion of symbol similarity to enable fuzzy search of pattern. This thesis combines data mining and information visualization to perform visual analysis. To proof the applicability of the results a prototype has been developed, to show that they are suited to support human pattern finding. Data dimentionality reduction enables the use of simple but effective 2D visualizations which show clearly pattern distribution in time and place. To enhance the effectiveness of the analysis the user has multiple abilities to adjust the visualizations to his needs. In addition this prototype shows a simple way of an interactive manipulation tool for the symbol similarities. Unfortunately the rest of this thesis will be written in german