Abstract
The rapid growth of online rating information enables firms to learn and monitor consumer preferences over their products. Based on multidimensional rating information systems, we propose a novel class of Evolutive Preference Analysis (EPA) methods to handle the dynamic online ratings with arbitrary rating distribution, which takes all the historical ratings into consideration and delivers a comprehensive ranking evolution. As a new tool to analyze real-time consumer preferences, the EPA class includes the EPA that emphasizes the overall preferences of consumers, and the EPAt that emphasizes the recent preferences that generally provide up-to-date information. Both of them include four indices (the expected priority vector, expected rank, confidence factor, and index of rank probability) to reveal consumer preferences over rated attributes of products along time. The evolution trends of these indices help firms verify whether their advertised products match the preferences of target consumers to improve their products and marketing strategies. Finally, the practical applicability of EPA class is corroborated by several numerical examples and one real-world application on smartphone online ratings, which demonstrates the effectiveness of the proposed EPA class on streaming rating evolution.