Multidimensional data mapping - integrating mining and visualization

Projection or point placement techniques, which seek to map multidimensional data onto visual spaces, have been the interest of the visual data analysis community for a long time due to their ability for exploratory tasks based on similarity and correlation. However, many problems still persist that impair their application, particularly those caused by the compromises involving computational costs, capability of separating groups of correlated points and exploratory power of the resulting display. The thesis that this paper describes achieved further understanding of the projection problem and resulted in the development of projection techniques that are faster than previous ones, appropriately define groups of highly similar data instances, separate these groups in the final layout, and allow the data exploration on different levels of detail. In addition, we integrate some data mining features to the process of multidimensional visualization, mainly for the application of projections to the visualization of document collections. Keywords-Information visualization; Visual data mining; Multidimensional projection; Documents map.