Perbandingan Analisis Sentimen Menggunakan Long Short Term Memory (Bi-Lstm) Pada Ulasan Pengguna Twitter Setelah Adanya Rebranding Menjadi 'X'

Abstract

Twitter is one of the most iconic social media platforms in digital communication. After being acquired by Elon Musk, it has rebranded the application, logo, and even the name Twitter itself to 'X'. The rebranding of Twitter is certainly a major concern on social media. Twitter users and the general public continue to provide various comments about this change. Twitter users not only share their comments or opinions on Twitter, but also express their feelings through reviews and comments on Google Play (Play Store). This study aims to conduct sentiment analysis related to user reviews of the Twitter application on Google Play after the rebranding and change to 'X' by comparing classification methods using Long Short Term Memory (LSTM) and Bidirectional Long Short Term Memory (Bi- LSTM). These two methods will be compared based on the classification results and accuracy obtained. Reviews from users will be classified based on positive and negative sentiments. In this study, the Word2Vec word embedding method with CBOW architecture model is used. The results show that the performance of the Bi- LSTM method is better than the LSTM method. The Bi-LSTM method produces an accuracy of 91.51%, FI-Score of 91.85%, precision of 90.25%, and recall of 93.51%. Meanwhile, the LSTM method with the best model produces an accuracy of 87.30%, F1-score of 87.11%, precision of 90.64%, and recall of 83.83%.