Possibilistic Fuzzy C-means Topic Modelling for Twitter Sentiment Analysis

Social media are generating an enormous amount of sentiment data in the form of companies getting their customers’ opinions on their products, political sentiment analysis and movie reviews, etc. In this scenario, twitter sentiment analysis is undertaken for classifying and identifying sentiments or opinions expressed by people in their tweets. Usually, the raw tweets consist of more noises in terms of URLs, stop-words, positive emojis and negative emojis, which are essentially reduced. After pre-processing, an effective topic modelling methodology Latent Dirichlet Allocation (LDA) is implemented for extracting the keywords and identifying the concerned topics. The extracted key words are utilized for twitter sentiment analysis using Possibilistic fuzzy c-means (PFCM) approach. The proposed clustering method finds the optimal clustering heads from the sentimental contents of twitter-sandersapple2 database. The acquired results are obtained in two forms such as positive and negative. Finally, the experimental outcome shows that the proposed approach improved accuracy in twitter sentiment analysis up to 33.5% compared to the existing methods: pattern based approach and ensemble method.