Improving the accuracy of collaborative filtering recommendations using clustering and association rules mining on implicit data

• Main Authors: Najafabadi, M. K., Mahrin, M. N., Chuprat, S., Sarkan, H. M.
• Format: Article
• Published: Elsevier Ltd 2017
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/75957/; https://www.scopus.com/inward/record.uri?eid=2-s2.0-84995610183&doi=10.1016%2fj.chb.2016.11.010&partnerID=40&md5=1724c1f18704634e5c43c28b74a18040
• Institution: Universiti Teknologi Malaysia

The recommender systems are recently becoming more significant in the age of rapid development of the Internet technology due to their ability in making a decision to users on appropriate choices. Collaborative filtering (CF) is the most successful and most applied technique in the design of recommender systems where items to an active user will be recommended based on the past rating records from like-minded users. Unfortunately, CF may lead to the poor recommendation when user ratings on items are very sparse in comparison with the huge number of users and items in user-item matrix. To overcome this problem, this research applies the users’ implicit interaction records with items to efficiently process massive data by employing association rules mining. It captures the multiple purchases per transaction in association rules, rather than just counting total purchases made. To do this, a modified preprocessing is implemented to discover similar interest patterns among users based on multiple purchases done. In addition, the clustering technique has been employed in our technique to reduce the size of data and dimensionality of the item space as the performance of association rules mining. Then, similarities between items based on their features were computed to make recommendations. The experiments were conducted and the results were compared with basic CF and other extended version of CF techniques including K-Means clustering, hybrid representation, and probabilistic learning by using public dataset, namely, Million Song dataset. The experimental results demonstrated that our technique achieves the better performance when compared to the basic CF and other extended version of CF techniques in terms of Precision, Recall metrics, even when the data is very sparse.