The Shopee Application User Reviews Sentiment Analysis Employing Naïve Bayes Algorithm

Abstract

With the significant growth of internet use in Indonesia, there has been a surge in online business activity. The convenience offered by online platforms is increasingly in demand because it allows consumers to shop without being bound by a certain time or location. Before making a purchase, consumers tend to look for information first through various sources such as reviews on blogs, Instagram, TikTok, or reviews on the YouTube platform which is integrated in the application. This research adopted a method that included planning, literature study, data collection, and data processing using a dataset from the Play Store application which was taken using the Python library with an initial amount of 5000 data. After a manual filtering process which involved removing slang words, eliminating duplications, and normalizing punctuation marks, the remaining data was 3946. The application of the Naïve Bayes algorithm in this research uses probability methods to classify and predict 3141 training data and 805 test data, with Python library help. The accuracy calculation results show satisfactory performance, with an accuracy of 86.00%, precision of 80.74%, recall of 78.13%, and f1-score of 79.00% on a dataset of 3946. Analysis from this research shows the dominance of positive sentiment in 2050 data, while sentiment negative amounted to 1199 data. The amount and quality of training data plays an important role in system predictions, where high data quality provides better accuracy in predicting sentiment classes