The explosive growth in the availability of information is overwhelming traditional information management systems. Although individual pieces of information have become easy to find, the larger context in which they exist has become harder to track. These contextual questions are ideally suited to visualization since the humrex visual system is remarkably adept at interpreting large quantities of information, and at detecting patterns and anomalies. The challenge is to present the information in a manner that maximally leverages our v/sual skills. This paper discusses a set of properties that such a presentation should have, and describes the design and functionality of Vxlnsight, a visualization tool built to these principles.

Abstract:- Unsupervised learning (clustering) deals with instances, which have not been pre-classified in any way and so do not have a class attribute associated with them. The scope of applying clustering algorithms is to discover useful but unknown classes of items. Unsupervised learning is an approach of learning where instances are automatically placed into meaningful groups based on their similarity. This paper introduces the fundamental concepts of unsupervised learning while it surveys the recent clustering algorithms. Moreover, recent advances in unsupervised learning, such as ensembles of clustering algorithms and distributed clustering, are described.

Unsupervised data classification can be considered one of the most important initial steps in the process of data mining. Numerous algorithms have been developed and are being used in this context in a variety of application domains. Albeit, only little evidence is available as to which algorithms should be used in which context, and which techniques offer promising results when being combined for a given task. In this paper we present an empirical evaluation of some prominent unsupervised data classification techniques with respect to their usability and the interpretability of their result representation.

In this paper, we present a new hierarchical clustering and visualization algorithm called H-BLOB, which groups and visualizes cluster hierarchies at multiple levels-of-detail. Our method is fundamentally different to conventional clustering algorithms, such as C-means, K-means, or linkage methods that are primarily designed to partition a collection of objects into subsets sharing similar attributes. These approaches usually lack an efficient level-ofdetail strategy that breaks down the visual complexity of very large datasets for visualization. In contrast, our method combines grouping and visualization in a two stage process constructing a hierarchical setting. In the first stage a cluster tree is computed making use of an edge contraction operator. Exploiting the inherent hierarchical structure of this tree, a second stage visualizes the clusters by computing a hierarchy of implicit surfaces. We believe that HBLOB is especially suited for the visualization of very large datasets and for visual decision making in information visualization. The versatility of the algorithm is demonstrated using examples from visual data mining.