Labeling Analysis in the Classification of Product Review Sentiments by using Multinomial Naive Bayes Algorithm

Abstract

Along with the development of technology, e-commerce also experienced a fairly rapid development. The existence of e-commerce becomes another consumer alternative to make it easier for them to fulfill their needs. After buying the goods, consumers are free to assess the products they buy. Product reviews and ratings provided by consumers are one means that can be used to increase sales and can also be used to determine the decision in purchasing a product by reading the product reviews. However, using ratings and reviews alone is not enough to summarize one’s opinion. Therefore, in this Final Project built a system that can classify opinions on product reviews into positive and negative sentiments by utilizing the rating. The dataset used is Grocery and Gourmet Food from Amazon as much as 50,000 which will then be labeled using Labeling Methods Average and Binary. The classification of this opinion uses the approach of Supervised learning Algorithm Multinomial Naïve Bayes. The result of this research shows that labeling using Method Average is suitable for processing Grocery and Gourmet Food Dataset and proves that the best ratio of feature selection usage is 20% succeed to produce 80.48% accuracy.