A Sentiment Driven Approach to Detect an Offline Event on Social Media Platform

A sentiment based approach to detect an event and predict a trend on social media platforms is proposed in the presented work. The trends are analyzed and events are detected against Twitter dataset on Sabarimala dispute. The dataset is collected over three months using Twitter API and then preprocessed and cleaned data is used for further analysis. The proposed approach uses numerical attributes and textual attributes with a blend of clustering and topic modelling techniques for analyzing trends and detecting an event. The proposed model successfully detects an event up to 80% of accuracy. The comparison of a detected event and real time event statistics drawn from social media and news channels highlights the performance and accuracy of the proposed model.