<![CDATA[The development of generative artificial intelligence, particularly ChatGPT, has elicited various responses from users on the social media platform Twitter (X), including support, criticism, concerns, and neutral information. This research aims to analyze Twitter (X) users' sentiments toward ChatGPT and compare the performance of the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) models in classifying negative, neutral, and positive sentiments. The research uses a quantitative approach based on deep learning. Data were collected using the keywords "ChatGPT" and "Chat GPT." From the 11,210 tweets obtained, 10,586 were used after the cleansing process, which involved removing URLs, mentions, hashtags, emojis, symbols, punctuation, special characters, and excessive spaces. The data distribution consists of 6,281 negative tweets, 3,484 neutral, and 821 positive. The data was divided with an 80:20 ratio into 8,468 training data and 2,118 test data. Evaluation used accuracy, precision, recall, and F1-score. CNN achieved an accuracy of 69.55%, a macro F1-score of 57.17%, and a weighted F1-score of 69.36%, while LSTM achieved an accuracy of 72.29%, a macro F1-score of 62.07%, and a weighted F1-score of 72.78%. LSTM produced 1,531 correct predictions, while CNN produced 1,473. Based on these results, LSTM shows better performance in classifying sentiment toward ChatGPT.]]>
Copyrights © 2026
