MatrixExplorer: a Dual-Representation System to Explore Social Networks
No visualization currently enables both a clear overview and details of associations between elements in large categorical data sets. Many approaches have been taken to address the problem, but each has produced limited results due to the inherent challenge in visualizing manydimensional data sets. In computer and social networks, for example, every node has potential connections to each of the other nodes, and so every node is a dimension in the dataset. Many visualizations, including the popular nodelink representation, quickly become occluded as networks grow beyond a trivial size. While no single visualization has succeeded in providing a complete insight into categorical data, several approaches are successfully used to view specific aspects of the domain. Interesting features that are occluded in one visualization may become readily apparent in another. With this understanding, the Information Visualization tool developed in conjunction with this research provides multiple coordinated views that present an organized visualization into categorical data sets. Each view takes a different approach to identifying clusters in the domain, and each uses size, position, color, and shading elements to communicate associations within and between clusters. Combined with the interactive features to explore and dynamically alter the canvas, the visualization system is named ConceptMap. The application includes two main innovations: First, a modified approach to hierarchical clustering that arranges nodes within a cluster at the same time it aggregates higher levels of cluster groupings. And second, A dendrogram coordinated with a reordered adjacency matrix in a fashion that the overview, clusters, and elements are visible simultaneously.