Personalized restaurant recommendation method combining group correlations and customer preferences

The increasing practicality of the group identification approach has led to many studies of restaurant recommendations. The success of group identification depends on how to fully aggregate the customer preferences in a group. However, the aggregation approaches towards customer preferences still pose many challenges to current research. For example, aggregation approaches can cause the group as a whole to report high satisfaction, while the satisfaction reported by individuals is low. Therefore, this paper proposes a novel personalized restaurant recommendation approach that combines group correlations and customer preferences. Our model employs the unsupervised means and probabilistic linguistic term set (PLTS) to conduct the group correlations between customer group and restaurant group. The recommendation list is provided by looking for the most similar group that the target customer belongs to. To validate the model, a case study of TripAdvisor.com is implemented. Our results confirm that the proposed restaurant recommendation approach outperforms the other three benchmark models.