MULTI-AGENT REINFORCEMENT LEARNING FOR TOURISM SERVICE RECOMMENDATION SYSTEM IN SMART CITIES DESTINATION

Abstract

The tourism industry is continuously evolving, with travelers seeking personalized and tailored experiences in smart cities destination. To address this challenge this study introduces a novel approach to tourism service recommendation systems by using Multi-Agent Reinforcement Learning (MARL) is a promising approach enables agents to collaborate and compete in complex environments. This paper proposes a MARL-based method for tourism service recommendation which learns from customer feedback to generate personalized recommendations. The proposed method employs a centralized critic and decentralized actor architecture to capture the complex interactions between agents and generate high-quality recommendations. The method's performance is evaluated on a real-world dataset and the results demonstrate that it outperforms existing methods in terms of recommendation accuracy and diversity. Additionally, this paper presents a tourism service recommendation system based on MARL and evaluates it using five algorithms: Real, Random, DQN, DDPG, and MADDPG. The results indicate that the MADDPG algorithm outperforms the other algorithms in terms of MCWT, CFT, and TSF. Finally, the implications of this study to tourism industry and can contribute to the development of more advanced and efficient tourism service intelligent recommendation systems.