Deep learning allows the recommender system to discover deep patterns and representations from data, capturing complex interactions between users and objects across multiple domains. Deep learning manages massive amounts of data and produces reliable representations to have effect on the quality of cross-domain recommender system. We utilize a federated learning algorithm to ensure that each domain retains control over its data and contributes to the global improvement of the suggestion, while also avoiding the sharing of sensitive information. To create major scenarios, privacy concerns are restricted to sharing the data. We recommend a novel model for personalizing the recommendation system by embedding the techniques of Federated Learning in tourism promotion. This model enables collaborative training across multiple domains to provide diverse recommendations and secure data.