580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can dynamic and personalized data integration improve thematic travel recommender systems?
This research area focuses on developing travel recommender systems that integrate heterogeneous, dynamic data sources (such as images, reviews, climate, social media) to generate personalized, thematic travel recommendations that go beyond standard location-based approaches. It matters because conventional destination-based recommendation methods often lack sufficient customization and adaptiveness, failing to capture traveler-specific constraints or evolving preferences, especially in the context of large, unstructured tourism data.
Key finding: Proposes a travel recommender system prototype leveraging augmented big data analytics from multiple dynamic sources (images, reviews, climate, social media) combined with collaborative filtering and crowdsourcing,... Read more
Key finding: Introduces THOR, a hybrid contextual preference model using classification and clustering (KNN, SVC, decision trees, Random Forest, Logistic Regression, K-means, DBSCAN) to build personalized user models from historical... Read more
Key finding: Develops a data-mining recommender system that constructs knowledge bases from historical travel data, applying association rule mining to model and update user preferences over time. The system scores and ranks travel offers... Read more
Key finding: Provides a comprehensive review identifying hybrid tourist recommender systems that integrate content-based filtering, collaborative filtering, and demographic or decision-tree based models as dominant approaches to mitigate... Read more
2. What are effective algorithmic and AI techniques for personalized and adaptive travel planning and trip itinerary generation?
This theme centers on methodologies for generating complete travel plans and itineraries tailored to individual or group constraints and preferences, using AI techniques like genetic algorithms, reinforcement learning, conversational agents, and scheduling heuristics. It is critical for overcoming user effort barriers in travel planning and enabling dynamic, adaptive itinerary construction that respects practical constraints such as time, opening hours, and routing.
Key finding: Proposes a genetic algorithm-based trip planner that generates full itineraries from minimal user constraints, considering place categories, opening hours, durations, and routing. The system offers a web app demonstrating... Read more
Key finding: Develops a conversational travel recommender system employing reinforcement learning to autonomously learn optimal interaction strategies that maximize user satisfaction and efficiency during travel planning dialogues. Unlike... Read more
Key finding: Introduces a heuristic solution to the Tourist Itinerary Design Problem (TIDP), considering personalized user preferences, time, POI opening times, and travel durations, producing near-optimal daily sightseeing routes. The... Read more
Key finding: Presents MobyRek, a dialogue-based mobile travel recommender that interactively elicits user preferences through critiques on recommended products, adjusting its preference model dynamically. The system blends initial... Read more
Key finding: Describes mITR, an on-tour mobile recommender integrated with a pre-travel planner NutKing, which initiates recommendations based on user historical preference data and location context. It employs a critique-based... Read more
3. How do group dynamics and multi-profile classification impact urban and cultural tourism recommender system effectiveness?
This research theme investigates recommender system approaches that address the complexity of recommending for groups of travelers or diverse visitor profiles, incorporating user attribute-based clustering and preference uncertainty modeling. It matters for designing systems that can effectively reconcile differing user opinions, manage heterogeneous preferences, and offer personalized recommendations even in group or cultural visitation contexts.
Key finding: Proposes a two-stage group recommender: first clustering urban tourists using a modified k-means with a custom distance metric based on user attributes (age, gender, expenditure, job) to form homogeneous groups; second,... Read more
Key finding: Reviews tourism recommender system approaches integrating user profile representation methods, including demographic, semantic, and uncertainty-based models, with a focus on collaborative filtering incorporating geospatial... Read more