Stable Voronoi Treemaps for Software System Visualization

Abstract

In this thesis we introduce stable Voronoi treemaps, a visualization technique that is designed for software quality monitoring. When monitoring the quality of a software system the analysis results of multiple versions of a software system have to be interpreted. Stable Voronoi treemaps help with interpreting the analysis results by creating stable and deterministic pictorial representations. This means that insight gained in one pictorial representation can easily be carried over to another.

As part of our implementation of stable Voronoi treemaps we introduce a sweep line algorithm for additively weighted power Voronoi diagrams. The algorithm extends Fortune's algorithm for Voronoi diagrams by adding weights to the Voronoi sites and using the power distance function that takes those weights into account.

To provide stability to Voronoi treemaps we introduce an algorithm based on scaled Hilbert curves that places Voronoi sites in a deterministic manner. By also enforcing a strict order on the data being visualized we can ensure that the pictorial representations remain visually close to each other while clearly showing the difference.

Using an empirical study we validate our result and conclude that stable Voronoi treemaps are useful for software quality monitoring and software quality assessment in general.