Consumer Rating And Sentiment Analysis Using Weighted Support Vector Machine

Abstract

Sentiment analysis is the analytical study of people's reviews, opinions, feelings, and attitudes. This significant problem is increasingly important in industry and culture. It presents many challenging research situations but ensures a relevant insight for everyone interested in opinion evaluation and social media analysis. This paper's main objective is to detect sentiment polarity such as positive, negative, and emoji representation using customer opinions on various products. Opinion mining from e-commerce websites plays an important role in making purchase decisions and creators to increase their product and marketing plans. However, it becomes very difficult for the customers to understand and evaluate the product's actual opinion manually. For this reason, we need an automatic way. Most of the researchers used machine learning algorithms to perform automatic representation of word embedding. One of the popular techniques in machine learning was used the support vector machine (SVM). The weighted support vector machine (WSVM) is the enhanced version for the standard SVM to increase the outlier sensitivity issue. This mainly focuses on word level and sentence level classification for sentiment analysis (SA). In this paper, the word2Vec model is used to extract the features from the customer reviews in WSVM based sentiment analysis of product reviews in E-commerce sites.  The experiment result shows that the proposed WSVM can works better on the sentiment classification job doing any model applied.