M.C. Barioni, E. Botelho (Brazil), C. Faloutsos (USA), H. Razente, A.J.M. Traina, and C. Traina, Jr. (Brazil)
Data Visualization, Dimensionality Reduction, Distance Functions
Completely automated data analysis techniques often fail to meet their requirements, due to their inability to exploit peripheral knowledge associated with the data. Human beings are very good at interpreting data represented in graphical format, and usually have the wisdom to recognize the associated knowledge. This paper addresses this dichotomy through a data visualization tool which displays, in a graphical manner, data stored in database relations, without requiring any native spatial data distribution, thus involving human beings in the main stream of KDD processes. It develops the conceptual framework which supports the data transformations enabling the visualization of data composed by attributes of many data types (numbers, dates and texts). This is achieved through the mapping of the attributes taken as multidimensional data into a 3-dimensional space, applying a user-defined distance function. Experimental evaluation shows that this tool is scalable to any database size, regarding number of tuples and attributes.
Important Links:
Go Back
