Abstract—Nowadays, the large datasets become more and more common. However, traditional visualization techniques, which although allow to visually analyze and explore data, can not scale well with the large one. This restrains the ability of detecting, recognizing and classifying phenomena of interest, such as patterns, clusters, trends, etc. This Paper proposes a method for interactive multi-resolution visualization to overcome the problem of traditional visualization techniques when working with a large dataset via hierarchical clustering. Based on hierarchical clustering, users can not only examine the dataset at different levels of detail, but also can explore many regions of interest. The basic idea underlying this method is to choose multiple scales from the hierarchical tree for representing the data at different levels of abstraction, which creates an easy environment for interactive exploration without re-run the clustering algorithm. Moreover, we also define a criterion for evaluating the multiple scale representation based on the concept of split factor. An experiment of applying the proposed method into the task of interactive visualization in a clinical case study of the large dMRI (diffusion magnetic resonance imaging) data is also carried out. The results show that our proposed method efficiently provides a friendly tool for visualization the large data.

Index Terms—Visualization, hierarchical clustering, machine learning, tract segmentation, diffusion MRI.

The authors are with the Bruno Kessler Foundation (FBK) and Trento University, Trento, 38122, Italy (e-mail: tbnguyen@fbk.eu, olivetti@fbk.eu, avesani@fbk.eu).

 

Cite:Bao T. Nguyen, Emanuele Olivetti, and Paolo Avesani, "Multiple-Scale Visualization of Large Data Based on Hierarchical Clustering," International Journal of Computer and Electrical Engineering vol. 6, no.2, pp. 77-82, 2014.