Publication Details
Abstract
This article examines how artificial intelligence and data analytics can support personalized digital tourism services under realistic conditions of data availability, privacy regulation, and system deployment. Although recent research has advanced tourism recommendation through heterogeneous data integration, context-aware modeling, sequential itinerary prediction, knowledge-enhanced systems, and trustworthy recommender design, the field remains fragmented. Existing studies often prioritize algorithmic accuracy while insufficiently addressing cold-start users, long-tail destination exposure, fairness, privacy preservation, scalability, and operational feasibility. To address this gap, the article proposes a comprehensive research framework based on a multi-source empirical design. The design integrates user–item interactions, points of interest, reviews, mobility traces, booking and search context, weather and temporal variables, and official tourism statistics. Methodologically, it develops a two-stage recommendation architecture consisting of candidate retrieval and learning-to-rank, enhanced where appropriate by sequential and contextual encoders, graph or ontology-based enrichment, and fairness- or privacy-aware re-ranking. Because the study is designed as a methodological framework, the results are defined through a comprehensive evaluation strategy rather than predetermined outcomes. The framework specifies evaluation metrics, ablation studies, statistical testing, and reporting standards needed to assess recommendation quality, robustness, trustworthiness, and deployability. The proposed design implies that effective tourism personalization should not be treated as a single-model optimization task. Instead, it should be developed as a trustworthy, multi-objective, multi-source recommendation stack capable of improving service relevance while supporting user trust, regulatory alignment, equitable exposure, and practical implementation in digital tourism platforms.