Dissertation
Attributed graph analysis with usability
Doctor of Philosophy (PhD), Washington State University
01/2020
Handle:
https://hdl.handle.net/2376/111094
Abstract
Attributed graphs are now been widely used in expressing world wide web, social network, knowledge base, biological structure, etc. Analyzing heterogeneous and large-scale graphs is expensive and writing queries to search entities or rank nodes is nevertheless a nontrivial task for end users. It is hard for end-users to write precise queries that will lead to meaningful answers without any prior knowledge of the underlying data graph. Users often need to revise the queries multiple times to find desirable answers. Given a large number of entities in the graph, users often require efficient predictive models that can effectively suggest the nodes as answers for analytical queries. Analyzing such graphs is challenging due to the ambiguity in queries, the inherent computational complexity (e.g., subgraph isomorphism) and resource constraints (e.g., response time) for large graphs.
In order to solve these challenges, the dissertation focuses on the problem of attributed graph analysis with usability. It solves two typical graph analysis tasks, entity search and node ranking. We provide usability for attributed graph analysis which contains (1) a query construction method that helps users to write precise queries without any specific query languages, (2) an optimization strategy that improves the query evaluation process, (3) a mechanism that helps users to understand the result and fine-tune the queries, and (4) a supervised model that captures users’ interest and automatically rank entities. The thesis experimentally verifies the efficiency and effectiveness of the proposed approaches using real-life graphs.
Metrics
24 File views/ downloads
21 Record Views
Details
- Title
- Attributed graph analysis with usability
- Creators
- Qi Song
- Contributors
- Yinghui Wu (Advisor)Diane Cook (Committee Member)Lawrence Holder (Committee Member)Assefaw Gebremedhin (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Number of pages
- 235
- Identifiers
- 99900581704001842
- Language
- English
- Resource Type
- Dissertation