Techniques for Dynamic and Diversified Relaxation in Constraint-Based Recommender Systems

Constraint-based recommenders support users in the identification of interesting items from large and potentially complex assortments. In cases where no recommendation could be found for a given set of requirements, such systems propose recommendations that indicate minimal sets of faulty requirements. Unfortunately, such recommendations are not diversified and do not include repair proposals which triggers a low degree of satisfaction and frequent repairs of recommendation sessions. This paper presents a dynamic and diversity relaxation approach that integrates the calculation of similarity requirements of customers and recommended products resulted in each atom of customer query and the cost of each atom of customer query. The Algorithm DDnRelax returning `at-least-n´ products can be efficiently solved diversified preference of customer.