Sentiment Analysis using Naive Bayes and Complement Naive Bayes Classifier Algorithms on Hadoop Framework

Sentiment analysis is a popular topic of scientific and market research areas in the last years. Sentiments can be in form of attitudes, emotions and opinions. Sentiment analysis focuses on texts such as reviews and attitudes about a product, a person, an event or an idea. In general, texts are classified into two groups such as positive-negative, good-bad, like-dislike etc. On the other hand, more classes can be added to these groups. Sentiments can be classified using machine learning methods, lexicon-based methods and hybrid which is combination of machine learning techniques and lexicon-based technique. In this study, sentiment analysis was conducted using machine learning techniques such as Naive Bayes and Complement Naive Bayes Algorithms using Hadoop software framework. Experiments were carried out using varying sizes of training datasets and about 8 million of reviews were classified as positive, negative and neutral. Performance of the algorithms were compared according to accuracy, precision, recall, and F-measure performance evaluation criterions.