Intelligent Social Media Text Mining for Political Influence Analysis: A Case Study on Makassar Mayor Election

Jufri (2024) Intelligent Social Media Text Mining for Political Influence Analysis: A Case Study on Makassar Mayor Election. Doctoral thesis, Asia e University.

	Text Thesis Jufri.pdf Restricted to Repository staff only Download (5MB)
	Text Thesis Jufri-1-24.pdf Download (718kB)

Abstract

Social media has become a common means of communication in society and plays a big part in many community activities in areas such as politics, economy, education, and information sharing. Important platforms such as Instagram, Twitter, Facebook and others can drive increased participation in elections. In this context, this research uses the Twitter application to influence political conversations through social media platforms. In the context of political communication during the Makassar mayor election, this study examines the application of a hybrid text mining approach that combines Support Vector Machine (SVM), Naïve Bayes, and K-means clustering for sentiment analysis. Traditional methods of sentiment analysis often fail to capture the nuanced sentiments of the electorate because of the complex nature of political discourse. This study aims to address these limitations by leveraging the use of SVM, Naïve Bayes, and K-Means. This research conducts data preparation involves cleansing and organizing textual data, segmenting it using K-means clustering, classifying it into sentiment classes using Naïve Bayes classifier, and enhancing classifications with the SVM. The results demonstrate that the hybrid model has superior performance compared to traditional methods, attaining an accuracy rate of 85.43% in contrast to the 64.96% accuracy rate achieved by traditional approaches. The hybrid approach demonstrates superior performance in sentiment accuracy and thematic analysis compared to traditional methods, highlighting its potential to extract meaningful insights from complex textual data. The findings reveal significant sentiment trends and discourse themes that influenced public opinion during the election. Furthermore, the research showcases the adaptability of the hybrid approach to diverse data sources and its applicability to other domains requiring detailed sentiment and thematic analysis. These findings constitute a valuable contribution to the fields of political science and computational linguistics by presenting a novel framework for sentiment analysis. This framework improves the analytical abilities of political analysts, campaign strategists, and policymakers.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Social media, twitter, political communication, K-means, Naive Bayes and SVM
Depositing User:	Muhamad Aizat Nazmi Mohd Nor Hamin
Date Deposited:	14 Jan 2025 03:49
Last Modified:	14 Jan 2025 03:49
URI:	http://ur.aeu.edu.my/id/eprint/1310

Actions (login required)

View Item