Location Based Real-time Sentiment Analysis of Top Trending Event Using Hybrid Approach

Abstract

Social media has become an integral part of everyone's daily life in today's digital era. Real-time sentiment analysis of this continuously generating data helps to understand people's attitude and behavior on various topics discussed currently. The analysis on the trending topic for particular location helps in identifying and describing the sentiments on a topic such as a product, person, or an event, providing useful insights for taking quick decisions or actions. This paper proposes a model for performing real-time sentiment analysis of top trending event for a given location on social media by developing a hybrid approach. Various combinations of Sentiment lexicon, unigram and bi-gram language model along with Naive Bayes and Support Vector Machine learning algorithm are used for analysis. The results obtained from these combinations are evaluated to check the performance of the model based on various parameters such as data size, feature selection method, training and testing data set and time. Hybrid model developed by combining sentiment lexicon, unigram language model with Support Vector machine algorithm gave maximum accuracy compared to other combinations on location based real-time social media data.