AI-Driven Framework for Location-Aware Sentiment Analysis and Topic Classification of Public Social Media Data in West Malaysia

Faisal, Amna
Jhanjhi, NZ
Ashfaq, Farzeen
M. Ahmed, Husham
Khan, Azeem

International Journal of Engineering, Science and Information Technology
Vol 5, No 4 (2025)

While social media has facilitated communication, it has also amplified collective attitudes, often leading to polarized opinions and negative emotional expressions that can disrupt social harmony. Consequently, monitoring public sentiments on social media, and identifying thematic trends across regions has become crucial for understanding collective emotions and opinions. Despite advancements in sentiment analysis and topic classification, very little research has been done to integrate geospatial analysis with these techniques, limiting their ability to provide location-aware insights into public sentiments and discussion trends. This study develops an AI-driven framework that leverages social media data to analyze public sentiments and classify discussions into relevant topics. Specifically, this research focuses on understanding the emotions and conversations of Peninsular Malaysia citizens using a self-collected dataset of public Facebook posts, analyzed at the state level to provide location-aware insights. Using VADER for sentiment analysis and zero-shot transformer for topic classification, this study categorizes posts into five predefined topics: politics, religion, tragedy, tourism, and food. The proposed architecture achieves a sentiment classification accuracy of 97% and a topic classification accuracy of 89%. Findings reveal that the Peninsular Malaysian population generally maintains a positive online environment, though some states showed a dominant negative sentiment. Patterns of dissatisfaction were largely related to political issues and local incidents, while positive emotions were primarily associated with tourism, religious festivities, and food-related news. This research not only identifies areas with dissatisfied publics but also explores the topics contributing to this sentiment. By emphasizing location-aware sentiment and topic trends, this framework offers insights to help policymakers and sociologists address region-specific issues, potentially reducing dissatisfaction and fostering a more harmonious society.