A Personalized Recommendation Model based on Collaborative Filtering and Federated Learning for Cultural Tourism Attractions in Fujian-Taiwan

In the era of big data, the explosive growth of cultural tourism information should provide personalized and privacy requirements for people’s travel. However, the existing travel recommendation algorithms have some obvious challenges, such as weak personalization, significant cold-start problem, and poor user privacy protection, etc. We propose a personalized intelligent recommendation model and method, which constructs a new collaborative filtering recommendation algorithm based on user characteristics portrait and historical behaviors, and designs an user-attraction data correction model based on federal learning to implement user privacy protection. Furthermore, We fully combine user age, gender, hobbies, and explicit and implicit data feedback to improve the accuracy of the proposed recommendation model. A sufficient experiments have been done on the real Fujian-Taiwan cultural and tourist attractions dataset, and the results show that, the proposed model protects the user’s private information and ensures a higher recommendation accuracy. Compared with the traditional collaborative filtering recommendation, it is concluded that the recommended attractions can accurately hit 50%-90% of users’ clear travel destinations, and the satisfaction rate of users can reach 77% among the attractions recommended to users outside the expectation, which shows good performance and can be effectively extended to other recommendation systems for cultural tourism attractions from the experimental results.