Visualization of N-Dimensional Databases Using Parallel Coordinates

Conventional methods of exploratory data analysis employ tools ranging from two and three dimensional scatterplots and descriptive statistical summaries to multivariate methods employing factoring and model formulation. Much of this kind of effort is directed toward distilling the important parameters and observations from noisy, and sometimes sparse, databases. Parallel coordinates provides a method for examining the relationships between variables in the context of the complete set of observations. Unlike conventional graphical methods. the'parallelcoordinate method is not restricted to three dimensions or four if one considers animation. Since parallel coordInate representations are based on the methods of projective geometry, fundamental relationships such as pos~ive or negative correlations between variables can be discerned directly from graphs in a manner similar to conventional scatterplots but for more variables simu~aneously. This results in a more compact and efficient tool for browsing and familiarizing onese~ w~h a complex database w~h practical benef~s in quality assurance and analysis of databases. Introduction