Attribute Correlation and Hotel Recommendation Based on Online Reviews

Online travel website reviews contain abundant hotel attribute information. Nevertheless, evaluations of such hotel attributes are not completely objective, as they are affected by subjective perceptions from different types of travelers. This study explores the correlations of attribute perceptions across different traveler groups and optimizes hotel recommendation models from the perspective of attribute correlation. Hotel attributes are analyzed based on online reviews collected from TripAdvisor, and the TOPSIS method is used to identify the core attributes. The SentiWordNet sentiment lexicon and PolarityRank algorithm are utilized to calculate attribute sentiment values, which are then applied to OLS regression analysis to explore the relationships between attributes. Finally, a hotel recommendation model based on the collaborative filtering algorithm is constructed, incorporating traveler types and attribute correlation. The research results show core attributes for different types of travelers, verify the existence of attribute perception correlation, and demonstrate that a model considering attribute correlation enhances recommendation accuracy, with an accuracy improvement of 3.16 percentage points.