<![CDATA[Aspect-based sentiment analysis is becoming increasingly important for understanding user opinions across various digital platforms; however, differences in text characteristics such as length, formality, and writing style often affect algorithm performance. This study aims to evaluate the effectiveness of a hybrid model that integrates Latent Dirichlet Allocation (LDA) with three classification algorithms Multinomial Naïve Bayes, XGBoost, and AdaBoost across three different platforms: Twitter, Google Play, and Female Daily. The evaluation was conducted using the Precision, Recall, F1-Score, and Accuracy metrics, with F1-Score serving as the primary indicator due to the imbalanced data distribution across aspects. The results show that Multinomial Naïve Bayes is more suitable for short and emotional texts on Twitter, XGBoost excels at technical reviews on Google Play, while AdaBoost is quite adaptive to narrative reviews on Female Daily. Key findings confirm that XGBoost is the most stable algorithm with the highest average F1-Score, while Multinomial Naïve Bayes remains competitive for informal text. This study contributes to the development of aspect-based sentiment analysis methodologies by providing practical guidelines for algorithm selection based on platform characteristics, and opens avenues for further research through dataset expansion, microtext normalization, and the application of deep learning to improve generalization]]>
Copyrights © 2026
