Sentiment Analysis in Product Reviews with Maximum Entropy and Naïve Bayes Using N-gram Method

Online shoppers now primarily rely on product reviews and user experiences shared on social media platforms before making an online purchase. The opportunity to capture product sentiments on social media has drawn interest in data mining analysis as it provides exciting ideas to support business growth and success. The two biggest online retailers, Amazon and Shein product reviews will be analyzed with the Maximum Entropy and Naive Bayes using the n-grams approach based on precision, recall, f-1 score, and accuracy. This study will also evaluate their time complexities depending on hardware and software requirements. Findings reveal that Maximum Entropy with n-grams generated improved accuracy performance, particularly with unigram, with an average accuracy rating of 91.5% compared to Naive Bayes with n-grams. However, the Naïve Bayes performs better when tested on different hardware and software specifications. Thus, the Naive Bayes algorithm performs better in execution time, and the Maximum Entropy using the n-gram approach resulted in improved accuracy.