An online review-driven two-stage hotel recommendation model considering customers’ risk attitudes and personalized preferences

Hotel recommendation models provide crucial references for customers to select their ideal hotels and help them overcome information overload. However, previous models primarily focus on capturing public preferences, neglecting personalized preferences or different risk attitudes among customers. To address this gap, this paper proposes a novel two-stage hotel recommendation model driven by online reviews, incorporating customers’ risk attitudes and personalized preferences. Firstly, this paper utilizes the Latent Dirichlet Allocation (LDA) topic extraction model and the sentiment analysis tool to extract public and personalized preferences from hotel reviews and customers’ historical reviews respectively. Secondly, in the first stage of hotel recommendation, this paper constructs a hotel filtering mechanism to cater to customers with different risk attitudes, ensuring that the recommended hotels align with customers’ individual risk tolerance. In the second stage of hotel recommendation, this paper introduces the cosine similarity algorithm of probabilistic linguistic term sets, enabling more accurate and tailored recommendations. Finally, to verify the applicability of the proposed model, a case study is conducted using real data from TripAdvisor.com. The results of the comparative analysis indicate that the proposed model outperforms other recommendation models.