<![CDATA[This study presents an optimized approach to sentiment classification of hotel reviews using a hybrid deep learning architecture. The model proposed combines Bidirectional Long Short-Term Memory (BiLSTM) with LSTM networks, enhanced by pre-trained GloVe word embeddings and SMOTE-ENN for handling data imbalance. The architecture incorporates a BiLSTM layer with 64 units and an LSTM layer with 32 units, complemented by dense layers and dropout regularization for optimal performance. Experimental results demonstrate the effectiveness of our approach, achieving an accuracy of 89.47% and an AUC score of 0.9607. The model shows robust performance across positive and negative sentiments, with precision scores of 0.96 and 0.82, respectively. Integrating SMOTE-ENN for data balancing and GloVe embeddings significantly enhanced the model's ability to capture semantic relationships in text data. Our findings indicate that this hybrid approach effectively addresses the challenges of sentiment analysis in the hospitality domain, particularly in processing nuanced customer feedback. The high AUC score suggests strong discriminative capability, while the balanced precision-recall trade-off demonstrates the model's practical applicability for real-world hotel review analysis.]]>
Copyrights © 2025
