Recurrent Session approach to Association Rule based Recommendation

This article introduces an association rule-based recommendation system (RS) using a recurrent neural network approach that is applied when a user browses items in a browsing session. It is proposed to overcome the limitations of the traditional query-based session method which is not adaptive to input, thus is not generative in generating recommendations. Our contribution lies in the training set which is formed from a series of rules and fed to the model, so it can predict the next-items from the input itemID series. The proposed method can adaptively and generatively produce recommendations from a series of items that a user sees in a session. Compared to traditional method, our method can generate recommendations for 100% of item series browsed by user, which traditional methods are also capable of, including those that traditional methods are unable to produce.