Optimal Sankey Diagrams Via Integer Programming

David Cheng Zarate, Pierre Le Bodic, Tim Dwyer, Graeme Gange, Peter J. Stuckey

Abstract: We present the first practical Integer Linear Programming model for Sankey Diagram layout. We show that this approach is viable in terms of running time for reasonably complex diagrams and also that the quality of the layout is measurably and visibly better than heuristic approaches in terms of crossing reduction. Finally, we demonstrate that the model is easily extensible through the addition of constraints, such as arbitrary grouping of nodes.