Abstract:

In this paper, we develop a hybrid recommender system that combines two recommendation paradigms: collaborative recommendation and content-based recommendation. In this recommender system, we use two main techniques of Data Mining. First, we apply associations mining [Agrawal et al (1994)] to customer purchase data in order to derive relationships between product classes and subclasses. Second, we use clustering [Michaud (1997)] to assign customers into groups with similar interests, based on their prior purchase patterns. The proposed system suggests new products to a customer and predicts his preference for each product contained in his personalized list of recommendation. These predictions are computed using evaluation data provided by each customer concerning his previous purchased products.