<![CDATA[The development of online loans in Indonesia has led to various public opinions spread across social media, one of which is the X platform. This research aims to analyze public sentiment towards online loans using the Long Short-Term Memory (LSTM) method. The data used consists of 702 Indonesian tweets collected through a crawling process with Tweet Harvest. Of these, 480 tweets were classified as positive sentiment and 222 as negative. The research process includes preprocessing, manual labeling, model training, and evaluation stages. The model was built using Sequential architecture from Keras, consisting of embedding layer, LSTM layer 128 units, 30% dropout, and output layer with softmax activation function. The model was trained using 562 tweets as training data and 140 tweets as validation data with a ratio of 80:20, for 10 epochs and batch size 64. The final evaluation using the entire dataset resulted in 92.59% accuracy, with 79.06% precision, 79.43% recall, and 79.14% F1-score. These results show that LSTM is able to classify sentiment stably and effectively, and has strong potential in sentiment analysis on short text data such as tweets.]]>
Copyrights © 2025
