Privacy Preserving Federated Learning Architectures for Collaborative Tourism Data Analytics

The increasing digitalization of tourism services has given rise to the generation of massive user data that is sensitive to consumers across the distributed stakeholders in the form of hotels, travel platforms, and transportation providers. The traditional centralized analytics solutions may pose a serious risk to privacy and regulatory concerns, which limit the successful exchange of data. A federated learning system preserving privacy that is specifically designed to support collaborative tourism analytics based on machine learning without having access to the raw data is presented in this paper. The methodology is a combination of secure aggregation, differential privacy, and communication-efficient model updates in a decentralized model. In order to quantify the proposed system, a multi-source tourism dataset comprising the user preferences, booking history, and mobility patterns is used. Experimental evidence shows that the federated model has an average prediction accuracy of 92.9% and a 0.93 F1 score, and provides strong data confidentiality. With the addition of the concept of differential privacy, there is a marginal accuracy trade-off of 1.8%, but much more resistance to inference attacks. 27% of overhead communication is minimized with adaptive model compression strategies to enhance the scalability of heterogeneous participants. Moreover, the architecture supports real-time collaborative insights to allow the stakeholders to maximize the recommendation systems, demand forecasting, and personalized services without violating the privacy of the users. The results show that federated learning is a feasible and safer option for distributed tourism data analytics, balancing between performance and privacy needs. Not only is the proposed framework compliant with the data protection standards and requirements worldwide, but it also leads to trust between the collaborating entities. The work helps advance intelligent tourism systems by showing how privacy-conscious machine learning can be successfully deployed in a multi-organizational setting, which prepares the way for safe and data-driven decisionmaking in the tourism industry.

Hali tarjima qilinmagan