Discovery and Visualization of Interesting Patterns

Data storage space comes almost at no costs today. Accumulating data is therefore an ubiquitous task in basically every business organization. However, this collection process needs to be complemented with sophisticated data analysis techniques in order to detect patterns inside these data. Such patterns may indicate problems or opportunities. In both cases it is of paramount importance to detect the formation and development of such patterns early enough in order to take timely countermeasures. To reach a large range of users, such analysis methods have to be intuitively controllable, must provide instant feedback and offer suitable visualizations. In this thesis, I propose a framework to visualize and filter the temporal evolution of sets of association rules. I will show how linguistic terms (represented by fuzzy sets) can be used to quantify a rule's history (w.r.t. certain quantitative measures) and subsequently rank them to present only the most relevant ones to the user for further assessment. I will transfer the suggested filtering method to other model types, present the software platform on which the methods are implemented and provide empirical evaluations on real-world business data.