Interactive Visualization Tools for Meta-Clustering

In this work we propose a scientific data exploration methodology and software environment that permits to obtain both data meta-clustering and interactive visualizations. Our approach is based on an elaboration pipeline, including data reading, multiple clustering solution generation, meta clustering and consensus clustering. Each stage is supported by dedicated visualization and interaction tools. Involved techniques include a Price based global optimization algorithm able to build a set of solutions that are local minima of the K-means objective function; different consensus methods aimed to reduce the set of solutions; tools for the interactive hierarchical agglomeration of clusterings and for the exploration and visualization of the space of clustering solutions.