<![CDATA[This study proposes an itinerary recommendation model based on Case-Based Reasoning (CBR), enhanced with an auto-revise mechanism and multi-cluster modeling using the DBSCAN algorithm. The model is developed from four primary data sources: historical travel cases, visit statistics, social media reviews, and contextual data. The auto-revise mechanism is activated when case similarity falls below 0.95, allowing solution adjustments based on six feature subsets: spatial, categorical, attraction, destination type, popularity, and visitor segmentation. Evaluation was conducted through 5-fold cross-validation and new-case testing, yielding F1-scores of 92.60% and 90.29%, respectively, while ranking performance remained consistently high across both evaluation scenarios. The model also demonstrated improvements in recommendation quality metrics, including novelty, diversity, and serendipity, alongside a reduction in average response latency from 25.53 ms to 20.09 ms. These results indicate that the proposed integrative CBR auto-revise approach, supported by contextual data and multi-cluster structuring, provides an adaptive and efficient itinerary recommendation framework suitable for real-time decision-support scenarios.]]>
Copyrights © 2026
