Bayesian network and sentiment analysis for evaluating online hotel reviews under multiple criteria

Online travel agency platforms provide extensive hotel reviews that reflect customer perceptions on multiple criteria. A novel multi–criterion decision–making approach is introduced that integrates Bayesian networks and the graph theory matrix approach to rank hotels based on online customer reviews. A sentiment analysis algorithm is developed to extract sentiment orientations from textual reviews. A Bayesian network is trained using both numerical ratings and sentiment scores to capture probabilistic dependencies among criteria and generate relative importance weights. The derived weights are embedded into the graph theory matrix approach ranking process. A case study on ski hotels in Turkey demonstrate that the Bayesian network graph theory matrix approach integration reflects customer preferences more effectively than conventional multi–criteria decision–making approaches that assume criterion independence. The results indicate that price–performance is a dominant factor in recommending ski hotels. Service quality and food quality are also important criteria that directly affect recommendation decisions and indirectly influence them through price-performance.