Abstract

The era has faced with explosive growth in data generation. Data generation has undergone a renaissance change. This availability of data has led a paradigm shift in the E-commerce sector; data is no longer a by-product of business activities, but are the asset to a company it helps in providing insights which are required in satisfying customers’ needs. This paper provides an overview of sentiment analysis of product reviews based on different algorithms and its efficiency in determining positive from negative reviews based on N-gram, Bigram with the application of Count-Vectorizer and (Term Frequency-Inverse Document Frequency) (TFIDF) Matrix. Different classification models have been employed to check the prediction accuracy of the unlabeled text. Based on the above  classification and tool has been developed which predicts the incoming reviews and classify its sentiment polarity.

Keywords: Text mining, sentiments, K-Nearest Neighbor (KNN), Random forest, Multinomial Naïve Bayes, TFIDF, Count-Vectorizer