In the last years, visual analytics got an important research topic to keep track of the vast amounts of electronically stored data and gain new information out of the data. This thesis arose from several real application areas and deals with visual analytics of two data types, multi-dimensional time-related event-based data, and multi-dimensional data without timestamp, which is very heterogeneously. In the first part of the thesis, a flexible approach to finding significant events, event clusters, and event patterns is introduced. The system has built-in functions for ordering event groups according to the similarity of their event sequences, temporal gap alignments, and stacking of co-occurring events. Three different case studies dealing with business process events, news articles, and time-related 3D data demonstrate the flexible capabilities of this approach. In the second part, an automatic and interactive approach for improving the quality of projections in terms of both structural preservation and class separation by feature selections and transformations is introduced. Quality measures for assessing the structural preservation quality and the visual quality of the projections are proposed. The effectiveness of the approach is evaluated by applying it to several widely used projection techniques using a set of benchmark data sets. A data example for which it can be shown how well the two parts fit together analyzes a common data set. It shows the combination of both approaches and the benefit that can be achieved with them in a sequential visual analytics process. Furthermore, there exists a close interaction between the visual and the algorithmic parts of the approaches, and a combination of algorithmic optimization with user interaction guides the user to find an optimal projection in terms of user satisfaction and quality measures. This results in a task defined better projection via user interaction as a step-wise optimization. But the approaches also cover other benefits like a descriptive real-time presentation of the measures visually and by numbers at once. Furthermore, a selectable stress value visualization leads to a better understanding of the data exploration and projection techniques.