Sentiment Analysis of Twitter Data

Abstract

Over the past few years, the novel appeal and increasing popularity of social networks as a medium for users to express their opinions and views have created an accumulation of a massive amount of data. This evolving mountain of data is commonly termed Big Data. Accordingly, one area in which the application of new techniques in data mining research has significant potential to achieve more precise classification of hidden knowledge in Big Data is sentiment analysis (aka optimal mining). A hybrid approach using Naïve Bayes and Random Forest on mining Twitter datasets is presented here as an extension of previous work. Briefly, relevant data sets are collected from Twitter using Twitter API; then, use of the hybrid methodology is illustrated and evaluated against one with only Naïve Bayes classifier. Results show better accuracy and efficiency in the sentiment classification for the hybrid approach.