Sentiment Classification Using Supervised and Unsupervised Approach

Abstract

In past few years, the data available on internet has multiplied at an alarming rate. Tweets, reviews, blogs and comments on social media have been a huge factor which has resulted in such a huge amount of increase in the available data. Because of this datasets being highly unstructured and of high dimensionality, sentiment classification becomes a very tiresome task. Sentiment Analysis is used to estimate the user opinion on various issues. It consequently mines states of mind and perspectives of clients on particular issues. It�s a multistep preparation where choosing and extracting elements is an indispensable stride that controls execution of sentiment classifier. In this paper we have used three supervised techniques namely SVM, Decision Tree and Nave Bays Algorithm and three unsupervised techniques called DE, PSO and K-Means The results are validated using different three benchmark labeled datasets data sets and on the different feature sets We have also performed feature selection using genetic algorithm and validated results using the features selected by the GA Experimental results shows that supervised techniques have outperformed supervised techniques on one dataset while for the two datasets supervised techniques have outperformed unsupervised techniques