NORMA eResearch @NCI Library

A study of Graph Neural Networks and Graph Attention Networks for Node Classification

Bhargava, Yash (2024) A study of Graph Neural Networks and Graph Attention Networks for Node Classification. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
Preview
PDF (Master of Science)
Download (1MB) | Preview
[thumbnail of Configuration Manual]
Preview
PDF (Configuration Manual)
Download (2MB) | Preview

Abstract

Graph representations have received considerable attention as they can capture complex data relationships represented as nodes and edges, such as molecular structures, social network analysis, healthcare, and citation networks, etc. Due to this diversity of application areas, finding a suitable network representation to effectively handle complex structures, such as high dimensional term document matrices where the relationship between the nodes and edges is complex is difficult. In this study, the authors have focused on two prominent neural network representations, namely Graph Neural Networks (GNNs) and Graph Attention Networks. These two well-known neural graph architecture are designed to improve the performance of baseline GNN architectures. A range of models are proposed, and were trained and tested on the CORA dataset which consisted of research papers and citations. While the two type of neural models can handle graph based representations, it is not known a priori which one is the most suitable one for node classification. It was found that Graph Attention Networks based on attention mechanisms outperformed all proposed architectures with an accuracy of 73.8%. An accuracy that is even better than the model found in Keras. The major findings of this study show that the graph neural network models based on attention mechanisms were better than the simple GNN node classifiers and also set new targets for effectively handling complex graph-structured data in various applications such as social network analysis and citation networks.

Item Type: Thesis (Masters)
Supervisors:
Name
Email
Estrada, Giovani
UNSPECIFIED
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Ciara O'Brien
Date Deposited: 07 Aug 2025 13:58
Last Modified: 07 Aug 2025 13:58
URI: https://norma.ncirl.ie/id/eprint/8466

Actions (login required)

View Item View Item