E-Locus - Institutional Repository of the University of Crete

Home Collections Type of Work Doctoral theses

Doctoral theses

Current Record: 1 of 2318

[Add to Basket]

Identifier

000454651

Title

Analysis and visualization of directed graphs

Alternative Title

Ανάλυση και οπτικοποίηση κατευθυνόμενων γράφων

Author

Λιονάκης, Παναγιώτης Κ.

Thesis advisor

Τόλλης, Ιωάννης

Reviewer

Αργυρός, Αντώνιος
Στεφανίδης, Κωνσταντίνος
Τσαμαρδινός, Ιωάννης
Πρατικάκης, Πολύβιος
Liotta, Giuseppe
Μανασάκης, Κωνσταντίνος

Abstract

The display of Graphs, is widely used for the visualization of data or information. Although it is important to show all the information represented by the edges and the nodes of the graph, there are cases where this is either impossible due to the complexity and the size of the graphs or even misleading since the graph could be very dense, while other reasons may be related to privacy protection (e.g., social network graphs). To overcome this challenge, different approaches imply to hide the unnecessary or redundant information or even replace the original graph with a subgraph or a summary. The main objective of this dissertation is to investigate and elaborate on the visualization and analysis of large graphs. More specifically, regarding vizualization we design and implement various versions of a new sophisticated framework of graph layout techniques that focus on the idea of improving the visualization aesthetics, in order to reduce the visualization complexity of the graph, on top of sophisticated graph drawing layouts. These techniques can be categorized in two main groups. The first category is with respect to bends while the second is based on edge removal. To this respect, we perform a set of experiments that show that these techniques produce progressively more abstract drawings of the input graph. No dummy vertices are introduced and the vertices of each path/channel are vertically aligned. Subsequently, the value of the introduced approach is that it also provides a generic and parameterized visualization method based on the given scenario. Towards this direction, we also elaborate on methods for grouping nodes with similar characteristics that naturally decomposes the graph based on the information derived by the edges within the graph, which can be used in order to have better visualization of complex networks. In order to evaluate our drawing layout techniques, we assess the usability of various versions of our new layout compared with the one produced by other hierarchical drawing techniques. As a result, we provide insights regarding the factors that affect efficiency so introduce a set of metrics in order to evaluate the performance of these techniques. Our algorithms require almost linear time. Moreover, we design and conduct a comparative task-based evaluation with users in which we ask the participants to carry them out in order to extract the user's satisfaction level and exploit any possible problems and difficulties as regards the reachability information. Generally, the drawings produced by our algorithms have lower number of bends and are significantly smaller in area. The user evaluation also reveals that the performance of the participants is slightly better in the drawings of our proposed model and that our model is preferred in overall rating compared to the other model.

Language

English

Subject

Abstraction of edges

Bends

Crossings

Experimental and user study

Graph algorithms

Graph drawing

Hierarchical graph drawing

Information visualization

Reachability