CUSTOMER PREFERENCE ANALYSIS FROM ONLINE REVIEWS BY A 2-ADDITIVE CHOQUET INTEGRAL-BASED PREFERENCE DISAGGREGATION MODEL

Online reviews have become an important data source for analyzing consumers’ preferences. Consumer preference analysis assists product managers to understand consumers’ propensity for different product attributes and make consumer-oriented market strategies. Existing studies on consumer preference analysis used simple additive algorithms to represent the relationship between overall ratings and attribute ratings, but ignored the interactions between attributes. In addition, not all attribute ratings were given by consumers when calculating the overall ratings of a product. To fill these gaps, a preference model based on the extended 2-additive Choquet integral is constructed. The 2-additive Choquet integral can reflect the importance of attributes and the interactions between pairs of attributes when integrating attribute ratings. In cases where consumers choose only a subset of product attributes to rate a product, we introduce the scale parameter into the 2-additive Choquet integral to characterize the relationship between different attribute subsets. Afterwards, a preference disaggregation paradigm based on nonlinear programming is provided to solve the preference model. Finally, the proposed method is validated by experimental analysis using the dataset collected from TripAdvisor.com. Experimental outcomes indicate that our approach can deduce consumers’ preferences and approximate the evaluation behavior of consumers efficiently.