Research on Hybrid Collaborative Filtering Hotel Recommendation Algorithm Based on Online Reviews and Ratings

In addressing the challenges of information overload and the filter bubble in online hotel reservations, this paper proposes a hybrid collaborative filtering recommendation algorithm that leverages online hotel reviews, likes, and rating data. Firstly, the algorithm utilizes the advantage of probabilistic linguistic term sets to handle fuzzy semantics, aggregating and expressing hotel feature information from online reviews. This information is further refined by adjusting with the number of likes on the reviews. Then, the maximum deviation method is employed to determine the weights of the hotel attributes. By combining cosine similarity and adjusted cosine similarity to integrate pessimistic and optimistic results, a fused similarity measure is obtained. Hotel rating similarities are calculated based on user ratings. The hybrid similarity is derived by obtaining a weighted sum of the fused similarity and rating similarity. Subsequently, this hybrid similarity is employed to construct the hybrid similarity matrix. Finally, the recommendation results are generated, incorporating user historical data. An empirical analysis was conducted using ten hotels in Shenzhen to validate the effectiveness of the hybrid algorithm. The findings demonstrate that the proposed algorithm enriches the research on hotel recommendation systems to some extent.