Improving The Visualization And Animation Of Weighted Dynamic Networks Using Force-Directed Graph Drawing Algorithms

Abstract: The visualization of networks as graphs composed of nodes and vertices benefits many fields of science including social network analysis. The use case of visualizations is twofold. Firstly, easy initial visualization of networks will help researchers find and specify their hypotheses before having to do any technical analysis. Secondly, once hypotheses are con confirmed, visualizations can be used to support these findings, making it possible to explain them to a broad audience. This thesis will expand upon the tools currently available for visualizing undirected graphs in two ways. Modern force-directed graph drawing algorithms are adjusted in order to approximate visualizing graphs' edges' weights as their respective lengths. A number of adjustments of Yifan Hu's spring-electrical force-directed graph drawing algorithm are compared and evaluated. Even though this is an NP-hard problem, results show that simple adjustments can improve a layout's edges' weight-length (ewl) relationship significantly. In order to evaluate whether graph's ewl scores improve from running the weight-adjusted Yifan Hu algorithm, a novel method is introduced. A number of experiments are conducted to investigate the effects of degree and variety of edge weights on a graph's ewl score. The second contribution concerns the design and implementation of functions aimed at visualizing the transitions between different timepoints of the same graph. Different approaches to ensure insightful animation of dynamic graphs are discussed and a method for the animation of dynamic graphs is implemented. Finally, both contributions are combined and applied to a real-world offline dynamic graph, resulting in visualisation of the co-occurrence of popular Twitter hashtags during the COVID-19 pandemic in Sweden. This application will visually highlight the contributions' strengths and weaknesses.