An Attention-Based Deep Learning Framework for Multi- Class Tourist Satisfaction Analysis

Abstract

Customer satisfaction is one of the most important factors influencing service quality, market competitiveness, and long-term growth in the tourism and hospitality sector. As online travel websites and review platforms continue to expand, analyzing customer feedback automatically has become essential for understanding visitor experiences and improving services. However, tourist reviews are generally unstructured, large in volume, and contain high-dimensional textual information, which makes accurate analysis difficult for traditional machine learning methods.To overcome these challenges, this research presents an attention-driven deep learning model for multi-class tourist satisfaction classification. The proposed framework employs dense neural layers to learn meaningful feature representations, along with parallel projection layers and an adaptive attention mechanism that highlights the most relevant review patterns. This design enables the model to capture semantic relationships more effectively and improve sentiment classification across multiple categories.The experimental results indicate that the proposed framework performs consistently well and converges stably during training. The model achieves an overall classification accuracy of 84%, while the macro and weighted F1-scores reach 0.85, showing balanced performance across sentiment classes. In addition, the average Area under the Curve (AUC) value of 0.98 demonstrates strong discriminative capability. Precision–Recall evaluation further validates the reliability of the framework, with average precision values ranging from 0.94 to 0.95. Confusion matrix analysis also confirms that the model can effectively distinguish among different satisfaction levels with minimal severe misclassification. Overall, the study highlights the effectiveness of attention mechanisms in improving feature learning and predictive performance, making the framework suitable for practical tourism analytics and intelligent hospitality management applications.