Dynamic identification for personalized product recommendations based on fuzzy discrete event systems

In this paper, we propose a new approach of dynamic identification to personalized product recommendations. Using this approach, we can predict the preferences of customers on products and help the customers to find the right products. The approach is a two-step procedure. The first step is to find the preferences of a customer on product features. The second step is to find the relationship between customer´s preferences on product features and customer´s evaluations of these products. For the first step, the preferences of the customer vary over time. In order to deal with the time-varying nature of customers, we use fuzzy discrete event system technology. For the second step, a back-propagation neural network is used to estimate the relationship between customer´s preferences and evaluations. The approach is applied to movie recommendations using Movielens database. The results are significantly better than the traditional approaches.