<![CDATA[Social media has emerged as a significant platform for individuals to express opinions and share experiences, extending to discussions on tourist attractions like Surabaya's mangroves. Utilizing sentiment analysis, this study evaluates public perceptions of the mangrove tourist destination. Concurrently, topic modeling extracts prevalent themes from textual data to discern evaluation aspects of Surabaya's mangroves. Employing sentiment lexicons and shadow labeling techniques, the analysis reveals a balanced sentiment distribution but susceptibility to misclassifications due to limited manually labeled data. Conversely, Latent Dirichlet Allocation (LDA) identifies three dominant topics: the attraction's appeal with culinary exploration, concerns over plastic pollution alongside distressing incidents like infant mortality, and issues of plastic waste accumulation and deforestation. Furthermore, after experimenting with various training models including LSTM, B-LSTM, machine learning algorithms (SVM, GNB, LR, LDA, and DTREE), and ensemble learning techniques (max voting, averaging, weighted averaging), the most optimal performance was achieved with the LSTM model, yielding an accuracy of 0.78. This underscores the utility of computational techniques in understanding public sentiment and thematic concerns surrounding Surabaya's mangrove ecosystem for tourism management and environmental conservation stakeholders and policymakers.]]>
Copyrights © 2024
