Managing uncertain preferences of consumers in product ranking by probabilistic linguistic preference relations

Product recommendation helps reduce the difficulty for consumers to make purchases from overloaded products on e-commerce platforms. However, “ramp-up” problems are common as many consumers have little information on purchases and ratings, making it difficult to predict their preferences. In this study, an interactive process is implemented between consumers and the system to capture consumer preferences. To make it easier for consumers to express their preferences for products with respect to different attributes, we design the interface in terms of linguistic terms and probability that are closest to expression habits, and structure the expressions through probabilistic linguistic preference relations. A consistency checking method and an inconsistency repairing method are presented to make the preference information provided by consumers meet the consistency requirements. We propose a partial ranking method based on attribute aggregation to rank products with preference information from the consumer. To rank products for which the consumer does not provide preference information, we develop a global ranking method by estimating the priorities of different kinds of attribute performance. Two examples of hotel recommendations prove that the proposed product ranking methods are suitable for consumers with clear purchasing demands.