Geo-Aware Real-Time Sentiment Analysis on Social Media Using Ensemble Machine Learning Techniques

Abstract

Real-time sentiment analysis has grown to be a significant factor in knowing public moods with crucial developments in tracking trends and immediacy in decision-making as a result of a rapid increase in the use of social networking sites for personal and professional purposes. This paper proposes a geo-aware sentiment-analysis framework that uses ensemble machine learning techniques to analyze social media data in real time. The key objectives include creating a predictive model for detecting sentiment trends within given geographic locations and tracking their propagation, as well as adding a threat visualization layer using geographic heat maps. The proposed system employs ensemble algorithms (Random Forest, Gradient Boosting, and Voting Classifier) that function together to provide accounts from many base learners to improve prediction accuracy and robustness. Sentiment data will be extracted from social platforms such as Twitter, pre-processed, and classified as positive, negative, or neutral. Mapped Sentiment-intensity, along with geo-taged metadata, will aid in identifying the emotional-spread across different regions. Experimental evaluation on benchmark datasets has shown that the Voting Classifier ensemble attained maximum accuracy of 89.3%, surpassing other individual models in both precision and recall. Through dynamic heat maps, the system shows effectively visualized high-risk areas, which are very important to applications in public safety, disaster responses, and policy planning. Thus, the research promises such integration of sentiment classification with geographic intelligence as a scalable, effective paradigm for social media analytics. This ensemble learning very well improves the reliability of sentiment detection and enables real-time monitoring of the general public opinion along with early identification of threats.