Analisis Sentimen Pengguna E-Commerce untuk Review Produk Komponen dan Aksesoris Komputer Menggunakan Bidirectional Encoders Representations from Transformers

Abstract

The need for computer component and accessory product has increased, one of the website that provides such products and reviews are the Tokopedia website. This reviews can help consumers decide which products they need and can also helps producers and distributors to get feedback from the consumers. However, an analysis is need to determine the polarity of such products review. In this study, a method that will be used to determine the polarity of those product reviews which divided into 3 categories, which is positive, neutral, and negative. This study was conducted using the TextBlob library and use the Bidirectional Encoders Representations from Transformers algorithm to conduct the sentiment analysis. After the implementation of the algorithm, an evaluation will be made to determine the accuracy of the sentiment analysis result using the K-Fold Cross Validation and Confusion Matrix methods. Based on the study, the train accuracy obtained is 92,27% and the test acccuracy us 91,26% for immediate result. For the evaluation, the train accuracy is 92,29% and the test accuracy result is 90.27% using the K-Fold Cross Validation method with a total of 5 folds.