Multimodal machine learning-supported comprehensive evaluation of tourism experience

Tourism experiences are multidimensional as they are influenced by the environmental, cultural and emotional factors. The current tourism analytics methods are based on unimodal data sources (surveys or textual reviews), which creates unimodal bias, cross-modal interaction, and limits the interpretability of the patterns of tourist satisfaction. To overcome these shortcomings, this study suggests a multi-modal machine learning (ML)-based model to evaluate tourism satisfaction comprehensively. It uses Multimodal Tourism Experience Dataset, which consists of 18,300 instances and were gathered on Kaggle; each sample consists of a tourist image, a user-generated review, and a five-point satisfaction rating. The image processing includes Gaussian filters and normalization, and numerical ratings are normalized by the MinMax normalization to guarantee consistency. Sentiment features of texts are obtained by means of Bidirectional Encoder Representations of Transformers (BERT) model, aesthetic and environmental representations are obtained by means of images with the help of a pretrained ResNet model. The extracted modality-specific features are integrated through an early feature-level fusion strategy to construct unified multimodal representation. The proposed Dynamic Differential Annealed Optimized Efficient Categorical Boosting (DDA-ECatBoost) model employs differential annealing to achieve stable convergence, robust hyperparameter tuning, and enhanced predictive capability. Experimental results demonstrate superior performance, achieving an accuracy of 0.9768, precision of 0.9647, recall of 0.9555, and an F1-score of 0.9651. Feature importance analysis reveals that comfort and visual appeal are key determinants of tourist satisfaction. Research demonstrates the effectiveness of multimodal data fusion combined with interpretable ML in enhancing tourism satisfaction analysis. It provides valuable insights for destination planners, cultural heritage managers, and tourism marketers aiming to improve visitor satisfaction.