Intelligent Online Store: User Behavior Analysis based Recommender System

Abstract

Recommender systems provide personalised recommendations to users, helping them find their ideal items, also play a key role in encouraging users to make their purchases through websites thus leading to the success of online stores. The collaborative filtering method is one of the most successful techniques utilized in these systems facilitating the provision of recommendations close to that of the customer's taste and need. However the proliferation of both customers and products on offer, the technique faces some issues such as "cold start" and scalability. As such in this paper a new method has been introduced in which user-based collaborative filtering is used at a base method along with a weighted clustering of users based upon demographics in order to improve the results obtained from the system. The implementation of the results of the algorithms demonstrate that the presented approach has a lower RMSE, which means that the system offers improved performance and accuracy and that the resulting recommendations are closer to the taste and preferences of the users.