Research on Hybrid Hotel Recommendation Algorithm Based on Probabilistic Linguistic Term Sets

This paper takes hotels as a typical tourism product and addresses the personalized and accurate recommendation needs and cold start problems in hotel recommendations. It introduces a hybrid hotel recommendation algorithm that incorporates probabilistic language term sets and online reviews, functioning as a switching-type hybrid approach. For returning users, an improved collaborative filtering recommendation algorithm is used, which fully integrates the advantages of probabilistic language term sets in processing fuzzy semantics. Firstly, the algorithm is based on the extraction and expression of the featuress of the hotel, and then the maximum deviation method is applied to determine the weights of the hotel properties. Lastly, it calculates the fused similarity matrix for hotels, combining adjusted cosine similarity with user ratings. The recommendation results are generated by combining the user's historical data. For new users, a user preference-based recommendation algorithm is used, which fully utilizes the online review text data of returning users and the historical search data of new users. The algorithm first collects and expresses hotel feature information from online reviews using probabilistic language term sets, constructs a hotel evaluation matrix, and then builds a user preference matrix derived from the user's past interaction data. The similarity matrix is calculated by cosinesimilarity, and the hotel is recommended based on similarity and hotel recommendations are provided according to the resulting similarity scores. This study performs a case analysis involving ten hotels in Macau, China, to assess the performance of the proposed hybrid algorithm. Different recommendation strategies are adopted for different users to meet their needs. This algorithm enriches the research on hotel recommendation algorithms to some extent.