Meta Data for Visual Data Mining

Abstract

Increasing amounts of large data and information sets require new analysis techniques. The domain of data mining investigates new paradigms and methods adapted for scalability, flexibility and problem abstraction for large data sets. In particular the field of visual data mining offers valuable methods for analyzing large amount of data intuitively. Visual data mining combines several visual and non-visual methods to reveal patterns, coherences and other features of data sets. The application of meta data can support the selection of suitable mining methods as well as of appropriate parameter values to control these methods. This paper defines a variety of meta data for visual data mining purposes, for instance by specifying ranges of values, cluster structures and regions of interest. These meta data can be applied for a more efficient visualization of the data set. We introduce a framework for effective extraction of these meta data.