Interactive Graph Clustering based on User Hints

Graph clustering is a difficult optimization problem that arises in Software Engineering. This paper presents a framework for graph clustering where users play a strong role. In the framework, a soft computing method produces a clustering of the graph and a visualization of it is provided using some graph drawing techniques. Through the visualization the user can then analyze the clustering and give “hints” that help the soft computing method to find better solutions. Hints include a variety of constraints for solutions, as well as direct manipulation of the previously computed clustering. The framework is flexible: it can accommodate several kinds of hints, clustering algorithms, and visualization techniques.