Summarization of Social Media “X” Using Latent Dirichlet Allocation: A Case Study of 2024 Indonesian Presidential Election

The rapid growth of social media platforms like X (formerly Twitter) has generated vast data pools, providing valuable insights into public opinion and discourse. This study introduces a novel hybrid model that integrates Latent Dirichlet Allocation (LDA) with a graph-based regularization technique, aiming to analyze and summarize social media discussions surrounding the 2024 Indonesian Presidential Election. Using a dataset of 17,565 tweets collected via the X API, this model enhances traditional topic modeling by applying a graph-based regularization approach, which improves topic coherence and representation. LDA hyperparameters (α=0.5, β=0.01, with 15 topics) were carefully tuned to achieve optimal performance. The model achieved a high coherence score (0.54) along with strong accuracy (0.86), Fl score (0.83), and recall (0.85). This study demonstrates that integrating LDA with graph regularization offers real-time, interpretable insights into political discourse on social media, providing a robust framework for understanding public sentiment during critical political events.