Publikationsserver der Universitätsbibliothek Marburg

Titel:Efficient Point Clustering for Visualization
Autor:Beilschmidt, Christian
Weitere Beteiligte: Seeger, Bernhard (Prof. Dr.)
Veröffentlicht:2019
URI:https://archiv.ub.uni-marburg.de/diss/z2019/0472
URN: urn:nbn:de:hebis:04-z2019-04729
DOI: https://doi.org/10.17192/z2019.0472
DDC: Informatik
Publikationsdatum:2019-08-28
Lizenz:https://rightsstatements.org/vocab/InC-NC/1.0/

Dokument

Schlagwörter:
Geovisualisierung, Datenverarbeitung, Informatik, Visual Clustering, Clusteranalyse, Geovisualization, Aggregation von Punkten, Point Clustering, Visualisierung

Summary:
The visualization of large spatial point data sets constitutes a problem with respect to runtime and quality. A visualization of raw data often leads to occlusion and clutter and thus a loss of information. Furthermore, particularly mobile devices have problems in displaying millions of data items. Often, thinning via sampling is not the optimal choice because users want to see distributional patterns, cardinalities and outliers. In particular for visual analytics, an aggregation of this type of data is very valuable for providing an interactive user experience. This thesis defines the problem of visual point clustering that leads to proportional circle maps. It furthermore introduces a set of quality measures that assess different aspects of resulting circle representations. The Circle Merging Quadtree constitutes a novel and efficient method to produce visual point clusterings via aggregation. It is able to outperform comparable methods in terms of runtime and also by evaluating it with the aforementioned quality measures. Moreover, the introduction of a preprocessing step leads to further substantial performance improvements and a guaranteed stability of the Circle Merging Quadtree. This thesis furthermore addresses the incorporation of miscellaneous attributes into the aggregation. It discusses means to provide statistical values for numerical and textual attributes that are suitable for side-views such as plots and data tables. The incorporation of multiple data sets or data sets that contain class attributes poses another problem for aggregation and visualization. This thesis provides methods for extending the Circle Merging Quadtree to output pie chart maps or maps that contain circle packings. For the latter variant, this thesis provides results of a user study that investigates the methods and the introduced quality criteria. In the context of providing methods for interactive data visualization, this thesis finally presents the VAT System, where VAT stands for visualization, analysis and transformation. This system constitutes an exploratory geographical information system that implements principles of visual analytics for working with spatio-temporal data. This thesis details on the user interface concept for facilitating exploratory analysis and provides the results of two user studies that assess the approach.


* Das Dokument ist im Internet frei zugänglich - Hinweise zu den Nutzungsrechten