Multi-objective contextual bandits in recommendation systems for smart tourism

Area of Science:

• Artificial Intelligence
• Computer Science
• Tourism Management

Background:

• Recommender systems are crucial for personalized smart tourism, but context-aware systems face challenges with real-time adaptation and multi-objective balancing.
• Information overload complicates tourist decision-making, necessitating advanced recommendation strategies.

Purpose of the Study:

• To develop a novel multi-objective contextual multi-armed bandit (MOC-MAB) recommender system for smart tourism.
• To address limitations of existing systems in real-time adaptation and balancing multiple objectives like relevance and fairness.

Main Methods:

• The proposed system integrates contextual bandit algorithms with multi-objective optimization.
• It dynamically learns from user feedback to optimize recommendations based on relevance and fairness.
• Experiments were conducted on simulated and real-world datasets (TripAdvisor).

Main Results:

• The MOC-MAB approach demonstrated superior performance compared to baseline methods.
• Key performance indicators included higher cumulative reward, improved click-through rates, and minimized regret.
• The system showed practical applicability in smart tourism scenarios, such as in Marrakesh.

Conclusions:

• The MOC-MAB system effectively enhances personalized recommendations in smart tourism.
• It offers a robust solution for balancing multiple objectives, improving user experience and decision-making.
• The study highlights the potential of advanced recommender systems in smart city tourism development.