<![CDATA[Understanding public sentiment toward online transportation services through social media analysis has gained increasing importance. This study provides a comparison between the effectiveness of Deep Neural Network (DNN), and Long Short-Term Memory (LSTM) models in analyzing user sentiment toward online transportation services in Indonesia using Twitter data. The dataset consists of 10,000 tweets related to Gojek, Grab, Maxim, and InDrive, collected from January to December 2023. Data preprocessing includes noise removal, case folding, tokenization, and stemming. Sentiment labeling was conducted using IndoBERTweet and manually validated. Using K-Fold Cross-Validation, both DNN and LSTM models were trained, and assessed using performance metrics such as accuration, precision, recall, and F1-score, training time, and Mean Absolute Error (MAE). The LSTM model demonstrated superior performances with accuration of 82,15%, precision of 82,21%, recall of 82,15%, specificity of 90,74%, F1-score of 82,10%, and MAE of 23,15%, compared to the DNN model which achieved an accuracy of 81,22%, precision of 81,20%, recall of 81,22%, specificity of 90,18%, F1-score of 81,12%, and MAE of 24,46%. However, DNN outperformed LSTM in training time efficiency (50,435 seconds vs. 148,765 seconds). LSTM shows significant advantages in understanding context and word relationships in sentiment analysis, while DNN offers better computational efficiency. The findings of this study can be utilized by online transportation services providers to improve service quality based on user feedback from social media.]]>
Copyrights © 2025
