Direct manipulation of parallel coordinates

This paper introduces two novel techniques to manipulate parallel coordinates. Both techniques are dynamic in nature as they encourage one to experiment and discover new information through interacting with a data set. The first technique, polyline averaging, makes it possible to dynamically summarise a set of polylines and can hence replace computationally much more demanding methods, such as hierarchical clustering. The other new technique interactively visualises correlation coefficients between polyline subsets, helping the user to discover new information in the data set. Both techniques are implemented in a Java-based parallel coordinate browser. In conclusion, examples of the use of these techniques in visual data mining are also explored