AN OPINION MINING APPROACH TO INVESTIGATE CROWDSOURCED WEB DATA by Ramandeep Sharma
The rapid proliferation of Web 2.0 has made it possible for people to express their opinions over the Internet. Users express their views, thoughts or feelings on Social media or on review sites about products or movies in a very convenient way. Sentiment classification plays a vital role in the area of data mining. It defines the attitude, emotions or thoughts expressed by the users about the particular topic or subject. This research uses the movie reviews to classifying the sentiments posted by the users about the movies over the internet. The classifier is trained with movie review dataset and classify the reviews into positive and negative. In this research, Na¨ıve Bayes is hybridized with Particle Swarm Optimization to improve its accuracy. The accuracy improves from 75.58% (using Naive Bayes) to 78.60% (using NB-PSO) by using unigrams as feature and Term Frequency-Inverse Document Frequency as a feature weighting scheme. The various contemporary techniques are also
compared with the proposed technique.