Design of a Tourist Attraction Recommendation Model Integrating Differential Privacy

A tourist attraction recommendation model that integrates differential privacy is proposed to address the issue of user privacy leakage in current smart tourism recommendation systems. This model utilizes convolutional neural networks to generate feature vectors for users' historical or browsing records of scenic spots. Then, based on the different levels of user interest, different differential privacy noise is added to the feature vectors of each attraction of the user, and the total potential feature vectors of the user are calculated using a weighted sum method. To ensure the security of user data, a unified Laplacian noise is added to the potential feature vector of the user, and similarity calculation is performed with the feature vector of the candidate scenic spot to be recommended. When the similarity exceeds a certain threshold, the scenic spot is recommended to the user.