Application Of Diversity In Recommendation System Based On Association Rules

Information overloading has become an increasing problem nowadays due to thepopularization of both Internet. This is particularly true in e-business domains whereusers find it difficult to identify products they need from the vast amount ofinformation provided to them. Recommendation systems are therefore developed forcustomer guidance.Association rule mining is widely used in most e-business recommendationsystems. However, it still has difficulties in unveiling meaningful rules and providingcustomers suggestions in a personalized manner. In addition, there is a concern fordiversity which suggests that customers would like to see recommendations fromcomplementary categories rather than purely similar products. This paper proposes amodel which can make a reasonable tradeoff between diversity and pertinence ofrecommendations. This is based on association rule mining on a conceptual hierarchymodel.A prototype system has been developed to evaluate the performance on theproposed model. This is based on an online toggery shop. Preliminary experimentalresults demonstrate improvement in the recommendation hit rate.