Visual Analytics Techniques for Clustering of High-dimensional Data

Clustering of large data sets is an important research area with a large variety of applications. Due to the curse of dimensionality, most clustering algorithms are only able to determine clusters with some rather limited set of predefined characteristics. The exponential search space does not allow an effective computation of more general clusterings. Visualization technology can help to solve this problem by guiding the clustering process and allowing a visual inspection of the intermediate results. In the presentation, the HD-Eye (high-dim. eye) system will be introduced which provides a tight integration of an advanced density-based clustering algorithm with state-of-the-art visualization techniques, allowing a better understanding and an effective guidance of the clustering process and therefore helps to improve the clustering results.