Abstract
Planning a tour Itinerary poses a significant challenge for tourists, especially when navigating unfamiliar territories. The computational complexity of tour recommendation further compounds this challenge due to its inherent intricacies. Many existing tour planning systems simplify recommendations by focusing solely on broad Points of Interest (POI) categories, often failing to align with travelers’ preferences and trip-specific constraints. To address this gap, we propose a personalized tour planning recommendation approach leveraging POI-embedding methods for a more nuanced representation of POI types.
Recent advancements in ML and AI algorithms have facilitated the numerical representation of tour data in natural language processing contexts. Coupled with our itinerary prediction algorithm, designed to accommodate temporal constraints, our approach offers a comprehensive solution for generating tour itineraries. The recommendation algorithms proposed to craft a predicted sequence of POIs that strategically balances time constraints, locational preferences, and user-specific inclinations, drawing insights from the past trajectories of similar travelers. Various computational models can be employed to conceptualize tour recommendations. We validate the effectiveness of our algorithms using the Flickr dataset encompassing nine cities, demonstrating their ability to furnish relevant and accurate itinerary suggestions. Performance Evaluation metrics such as F1, recall and precision scores substantiate the efficacy of our approach.
Speaker Bio: Ho Ngai Lam is a PhD candidate at the Information Systems Technology and Design (ISTD) pillar of the Singapore University of Technology and Design (SUTD). He holds a Bachelor of Computing (Hon. I) from the National University of Singapore (NUS). Prior to his doctoral studies, Ho gained substantial experience in software development within the Energy and Banking sectors in Singapore.