Naive Bayes Classifier and Word2Vec for Sentiment Analysis on Bahasa Indonesia Cosmetic Product Reviews
Cindy Charella Putri Hapsari, Widi Astuti, Mahendra Dwifebri Purbolaksono · 2021
Cosmetic products are products that are widely sold on e-commerce. A product, including a cosmetic product can generate mixed sentiments in the form of customer reviews. Therefore, customer reviews are one of the most important to be paid attention to. This is because from the customer reviews, it can be known the level of customer satisfaction about the product that has been purchased. Sentiment analysis is a solution that can be used to measure customer satisfaction. Sentiment analysis is a text-based research field that is suitable to discuss the problem of customer satisfaction about the product. The analysis used is based on several aspects of cosmetic products, namely aroma, packaging, price, and product. In this study, the problem was solved by analyzing sentiment using the Naive Bayes and Word2Vec methods. Based on experiment results, Naïve Bayes and Word2Vec could be used to classify the sentiment. The best model of this research produces an accuracy of 68.17 % with an accuracy of 56.36 % for product aspects, 70.96 % for price aspects, 68.79 % for packaging aspects, and 76.57 % for aroma aspects. This result is obtained when the translated data is used and validated using 2-fold cross validation. The parameters for Word2Vec are window = 7 and size = 300.