Adaptive Personalization for Tourism Recommendations: A Hybrid CBR-AHP-TOPSIS Framework with Context-Aware Mode Adjustment

Background: Conventional case-based reasoning (CBR) recommender systems employ fixed similarity weights, failing to capture dynamic user preferences across travel contexts.Objective: This paper proposes an adaptive hybrid framework integrating CBR with multi-criteria decision making (AHP and TOPSIS) for context-aware personalization in tourism recommendations.Methods: Weights of criteria are derived from user surveys using the analytic hierarchy process (AHP) and fine-tuned using transaction-informed mode-specific adjustment factors calculated from historical rating patterns.The framework is evaluated on a real-world tourism dataset from Yogyakarta, Indonesia (15,870 records). Results:The proposed method achieves a precision of 0.85, recall of 0.88, F1-score of 0.86 and RMSE of 0.27.Compared to feature-enhanced deep learning baselines (LightGCN-feature, SASRec-feature), it achieves 4.9% higher precision and 3.6% higher F1-score.Statistical tests (Friedman, Wilcoxon) confirm medium-to-large effect sizes (r = 0.63-0.92).Cross-regional validation (Bali, Malang), cold-start evaluation and bootstrap confidence intervals support the robustness of the framework within the tested regional subsets, particularly in sparse-data scenarios. Conclusion:The proposed framework offers an interpretable, generalizable and efficient approach to context-aware personalization, with documented practical implications for destination management organizations.