Data-driven fuzzy preference analysis from an optimization perspective

How to leverage massive online data to understand consumer preferences over products and services has accumulated significant attention in business. In this paper, we analyze consumer preferences by modeling it as a decision-making problem of ranking alternatives with consumers' online ratings. We propose a data-driven fuzzy preference analysis (D-FPA) method to obtain the priorities of alternatives. We show that the D-FPA is tractable and with high computation efficiency. In addition, we propose a natural indicator to measure the reliability of the derived ranking results and suggest thresholds of this indicator for better control of the method. A real-world application about online film rating is provided to illustrate the D-FPA, demonstrating that the derived ranking results converge rapidly and remain stable with the observed empirical data. Finally, we show how to build up an effective recommendation system with empirical data from MovieLens.