Sentiment Analysis and Classification on Amazon Products using Improved Support Vector Machine for Multiclass Classification

Abstract

There is a huge increase in number of peoples who have been accessing many social networking sites especially user post or reviews for a specific product, company, brand, individual, forums and movies etc. These reviews are helpful in judging customer perception on certain thing. The development of algorithms that could automate the categorization of distinct comments based on feedback from consumers became an analyst project, and this automated classification process is known as sentiment analysis. This research main goal is to analyze Amazon product reviews using an approach to Machine Learning (ML) built around TF-IDF and then employ the Support Vector Machine (SVM) algorithm to categorize the sentiment scores and sentences. SVM can handle binomial classification but the customer reviews is mostly classified into positive, negative and neutral and in some applications, it is fine grained into star ratings such 1-5 or sometimes 1-10. Also, in some applications features or attributes are high in number in which some are irrelevant. Hence, this work applies feature subset algorithm and improves the existing SVM to handle multiclass classification. The Sentiment analysis, Rapidminer tool is considered for classification and the results are visualized, assessed with suitable classification metrics.