| Abstract |
As smart cities and tourism industries continue to evolve, personalized recommendation systems have become essential for meeting the diverse and dynamic preferences of users. This study proposes a hybrid tourism recommendation framework that integrates optical image recognition with collaborative filtering (CF), addressing challenges such as data sparsity, cold-start issues, and the lack of contextual understanding. By leveraging optical technologies, such as drone-captured aerial imagery and optical character recognition (OCR), the framework analyzes real-time scenic area information, including crowd density and weather conditions, while enhancing user scene comprehension through the recognition of maps, road signs, and menus. Additionally, the framework constructs a domain-specific tourism knowledge graph to model destinations, activities, and user preferences as multi-relational data. Advanced embedding techniques, such as TransE and RotatE, are applied to extract meaningful semantic relationships, which are incorporated into the CF model to enrich the user-item interaction matrix with contextual information. Extensive experiments conducted on real-world datasets demonstrate that the proposed framework significantly outperforms traditional CF and existing hybrid models, achieving notable improvements in Precision@k, Recall@k, F1-score, and diversity. These results highlight the ability of the framework to generate highly personalized and context-aware recommendations, providing an innovative solution for advancing smart tourism systems and enhancing user satisfaction through accurate and diverse travel suggestions. © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only. |
