Novel methods for visualising graphs

Abstract

Graphs/networks are fundamental mathematical constructs that play a crucial role in representing and analysing diverse real-world phenomena across various disciplines. However, rapidly increasing quantities of data pose significant challenges for informed decision-making. Visual clutter increases proportionately with the graph’s size and complexity, obscuring semantic relationships and limiting human comprehension. Additionally, graph visualisation research has primarily focused on depicting graphs based on their primary values without considering the uncertainty inherent in the data. This could yield visual representations that lead to overlooking unrevealed trends and patterns or misinterpretations of the underlying data by human viewers. Consequently, there is an increasing need for methodologies that assist end users in understanding their data and its inherent structure, thereby facilitating an effective analysis and better decisionmaking procedure under uncertainty. This thesis outlines two research objectives: • Addressing graph summarisation challenges through proposing a summarisation algorithm. The first objective is to understand the network data pertaining to links and nodes, with less emphasis on the network’s structure and connectivity. Users require simplified visualisations that clearly convey the relationship between network structure and associated data. We developed an algorithm to summarise graph-based data by extracting the maximum information in readable and informative forms of the original graph to end-users. • Exploring the development of a novel visualisation approach to aid the design, implementation, and operation of visual search tasks on node-link diagrams. The second objective is to identify and address existing approaches’ limitations and challenges. This research introduces a node-link visualisation model designed for visually representing and analysing bivariate networks. We demonstrate it effectively addresses the challenges associated with these approaches. The major contributions of the thesis are as follows: 1. We present a novel node-link visual model — visual entropy (Vizent) graph — to effectively represent both primary and secondary values, such as uncertainty, on the edges simultaneously. 2. We present the novel Vizent edge design and empirically demonstrate (in collaboration with Lucy McLaughlin) that different edge glyphs have a perceived order through pairwise testing. x 3. We perform two task-based usability studies to demonstrate the efficiency and effectiveness of our approach for visualising bivariate networks using static node-link diagrams. 4. We compare the Vizent design against three visual encodings selected from the literature on various graphs ranging in complexity from 5 to 25 edges for three different tasks. Keywords: Information Visualisation, Graph Visualisation, Graph Summarisation, Node-Link Diagram, Edge Visualisation