Sentiment Analysis of Product Reviews using Naive Bayes Algorithm: A Case Study

Abstract

In this digital era, social media is used as a means to perform social activity and advertisements by companies. All of companies from small to big online shop created an endorsement to promote their products, and then it can be recognized. As a fast food restaurant, KFC launched the latest product KFC Salted Egg, As known, KFC often release unique products such as ChoChick, chicken sprinkled with chocolate spice. KFC created an endorsement by selecting Raditya Dika as an endorser. By using endorsement, KFC will get a good or bad sentiment. Analysis is needed to gain the sentiment’s effect of the endorsement. In conducting sentiment analysis, data was collected from two social media comments, YouTube, and Twitter. According to research conducted by Statista in 2007, the most widely used social media in Indonesia was YouTube while twitter was seventh. Even so, the development of twitter users time by time was increasing. It indicated that twitter was widely used. Naive Bayes was chosen to perform sentiment analysis because this method has a high accuracy in various studies. The stages of this research are divided into two periods, before and after endorsement. Data has been collected through the process of prepossessing, and then classification is done by using confusion matrix. The result showed that Naive Bayes has an accuracy rate more than 84%. However, negative sentiment rose by 12.51%. Neutral sentiment in this study contains neighbors of social media users who want to try the product, but the result after neutral sentiment endorsement decreased. It can be concluded that 9.77% of the decline has tried the product.