<![CDATA[Emotions are an important indicator in understanding public responses on social media, particularly X, which is the main medium for public expression. This study aims to develop a classification model for the emotions of Indonesian-speaking X users using a Bidirectional Long Short-Term Memory (Bi-LSTM) approach combined with data mining-based feature selection techniques. A dataset of approximately 6,000 tweets was collected through X scraping based on keywords and hashtags representing six main emotions: anger, sadness, fear, happiness, love, and surprise, from January 2024 to March 2025. The obtained data was processed through text normalization, stop word removal, and tokenization stages. Features were extracted using TF-IDF and selected using the Chi-Square method to improve classification performance. Tweets were labeled with emotions manually and semi-automatically. The Bi-LSTM model was trained and tested using accuracy, precision, recall, and F1-score metrics. Initial test results showed an accuracy of 86.3%, with the best performance on the emotions “happy” and “angry.” This study shows that the integration of deep learning and data mining can improve the accuracy of automatic emotion detection in Indonesian text. The main contribution of this study is the integration of Chi-Square feature selection with Bi-LSTM for Indonesian text, which has not been widely explored before.]]>
Copyrights © 2025
