Mukti, Dhea Sila (2026) KLASIFIKASI TOPIK DI TWITTER MENGGUNAKAN MODEL GRAPH NEURAL NETWORK (GNN) BERBASIS MULTI-VIEW. S1 thesis, Universitas Malikussaleh.

	Text cover.pdf Download (275kB)
	Text Abstrak.pdf Download (511kB)
	Text Bab 1.pdf Download (643kB)
	Text Daftar Pustaka.pdf Download (456kB)
	Text full_text.pdf Restricted to Registered users only Download (3MB)

Abstract

Twitter has become a primary platform for political discourse in Indonesia, generating large volumes of unstructured text data that are difficult to classify manually. This study aimed to construct Twitter data representations as three graph types, develop a Multi-View Graph Neural Network (GNN) model for political topic classification, and analyze propagation patterns and user relationships based on classification results. Tweet data were represented as a Semantic Content Graph using IndoBERT embeddings and cosine similarity, a User Interaction Graph based on mention, reply, and retweet relationships, and a Temporal Propagation Graph based on chronological propagation sequences. All three graphs shared 15,131 tweet nodes with 155,428, 184,751, and 103,055 edges respectively. A Multi-View Graph Attention Network Fusion model was developed to integrate the three graphs using a fully learned attention mechanism, trained under a semi-supervised setting with 1,677 labeled tweets out of 15,131 total nodes. The model achieved an Accuracy of 84.23%, Macro Precision of 82.52%, Macro Recall of 83.76%, and Macro F1-Score of 83.04%. Attention analysis revealed that the Semantic Content Graph contributed most at 54.53%, followed by the Temporal Propagation Graph (26.99%) and User Interaction Graph (18.48%). Of 15,131 analyzed tweets, 7,996 (52.84%) were classified as political content, while the remainder covered diverse non-political topics including a significant volume of commercial activities, reflecting the heterogeneity of discourse in the Indonesian Twitter ecosystem. Keywords: Graph Neural Network, political topic classification, multi-view learning, IndoBERT, Twitter

Item Type:	Thesis (S1)
Subjects:	Q Science > Q Science (General) Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software T Technology > T Technology (General)
Divisions:	Fakultas Teknik > 55201 - Jurusan Teknik Informatika
Depositing User:	Dhea Sila Mukti dhea
Date Deposited:	23 Sep 2026 03:54
Last Modified:	23 Sep 2026 03:54
URI:	https://rama.unimal.ac.id/id/eprint/23158

Actions (login required)

View Item