Analyzing Abstention Discourse in Presidential Elections: Knowledge Discovery in X Using ML, LDA and SNA

As a social media platform, X (formerly Twitter) has become a massive source of real-time, unstructured data, providing valuable insights into people's opinions on various issues. One crucial social phenomenon that has attracted attention on social media is the discourse around abstention (commonly known as “golput” in Indonesia) in the context of the presidential election. Abstention refers to the deliberate act of refusing to vote. Understanding the patterns, preferences, and topics associated with abstention discourse can provide valuable knowledge for political analysis. This study aims to discover knowledge based on patterns, sentiment polarization, and issues from unstructured X data to understand the discourse surrounding abstention in the 2024 Indonesian presidential election. The methodology involves collecting data from the X API, conducting Social Network Analysis (SNA) to analyze the social structure, preprocessing the data, and searching for the best sentiment analysis model through hyperparameter tuning on six Machine Learning (ML) models. Then, Latent Dirichlet Allocation (LDA) is employed with coherence score evaluation to identify topics related to the issue. The results indicate that 2,489 tweets discussing abstention were collected during the study period, exhibiting varied daily trends. SNA analysis reveals the formation of clusters within the dataset, alongside the identification of influential actors through three different centrality calculations. The sentiment analysis results show that the Logistic Regression (LR) model with count vectorizer is the best-performing model, with a predominance of positive sentiment polarity over negative sentiment. Evaluation of LDA using coherence scores indicates the presence of five topics related to abstention. This research contributes to knowledge discovery on the X platform by providing valuable insights into the discourse surrounding abstention in the Indonesian presidential election. These findings offer a deeper understanding of public opinion, political engagement, and election dynamics.