parvis parallel coordinates visualisation

"Nicht Parvis, sondern Parsifal müsste er heissen, denn das hätte zum Gral gepasst."
- Stanislaw Lem, "Fiasko"

Introduction

Features

Parallel Coordinate Visualisation of arbitrary data sets

click to enlarge screenshot

Multidimensional datasets with arbitrary number of dimensions and records can be loaded into parvis and automatically rendered as a Parallel Coordinate visualisation of the data.

Interactive rerdering, scaling and translating of axes

click to enlarge screenshot

The rendered data can be manipulated intuitively and in real-time. Axes can be reordered (to show different relationships between values), scaled and translated by simple drag operations. The progressive rendering provides the possibility of real-time manipulation of the display even with large datasets.

Brushing of records with boolean operations (add, subtract)

click to enlarge screenshot

Records can be selected in the display by dragging over them with a "brush". Brushes can be expanded, or other records can be subtracted from the brush. This allows to create boolean queries graphically, e.g. "all cars with 4 cylinders built before 1980" (shown here).

2-step progressive, selective rendering on two layers for interactive handling of large datasets

click to enlarge screenshot

Since real-time high quality rendering of large datasets can become very time consuming, the application provides the user with a fast preview rendering of the data, to enable her to continue with her work. The high-quality rendering is continued in the background, and automatically replaces the preview image when finished. Shown here is a high-quality rendered brush in front of a preview-quality rendered background, still being rendered in the background (as indicated by the progress bar).

Axis histogram for better understanding of data density

click to enlarge screenshot

With this new feature you immediately get an overview of the number of records in an area in comparison to the total number of records and the brushed records. The number of histogram "bins" that an axis is divide into is configurable.