Personalized Recommender Agent for E-Commerce Products Based on Data Mining Techniques

Abstract

In this article, a recommender agent is designed to meet the increasing demand of consumers for the diversity of products offered by big e-commerce Web sites. Turning visitors of Web sites into customers and clicks made by them into purchases is a challenging task which is achieved through accurate product recommendation. The recommendation algorithm designed in the present work initially design a user-context click matrix that predicts each users’ preference level based on deals offered by the company. The users are clustered on the basis of these preferences and neighborhood formation task is completed using collaborative filtering technique that is based on user-item category matrix. The matrix shows users’ preference for the type of item user is interested in. After the neighborhood formation task, like-minded users are found using various similarity measures. Finally, the products that are clicked by similar users are marked to find association level among these products using association rule mining to generate user-product preference matrix. The proposed work is flexible and can be applied to Web sites that keep track of users’ click-stream behavior. The experimental results clearly justify that the proposed work outperforms the conventional ones.