Abstract—Recommender systems are well known for their wide spread use in e-commerce, where they utilize information about user’s interests to generate a list of recommendations. To enhance the recommendation quality, the recommendation techniques have sometimes been combined in hybrid recommenders. In this paper, we propose a weighted hybrid recommender system that integrates multiple recommendation algorithms together to improve recommendation performance. In the proposed approach, firstly users are classified by applying clustering technique on ratings data. Subsequently, rule-based reasoning (RBR) and case-based reasoning (CBR) are employed separately to choose classes (neighborhoods) of an active user and then collaborative filtering (CF) is applied on these neighborhoods to produce recommendation lists. These two techniques are respectively called RCF (combination of RBR and CF) and CCF (combination of CBR and CF). The proposed weighted hybrid recommender system (WRCCF) combines RCF and CCF schemes. Experimental results reveal that the proposed WRCCF consistently outperforms Pearson CF (PCF), RCF, and CCF in terms of prediction and classification accuracy.
Index Terms—Hybrid recommender system, collaborative filtering, clustering, case-based reasoning, rule-based reasoning.
Shweta Tyagi is with the School of Computer and Systems Sciences, Jawaharlal Nehru University, New Delhi, India (e-mail: shwetakaushik2006@gmail.com).
Kamal K. Bharadwaj was with the Computer Science Department, BITS, Pilani (Rajasthan), India. He is now with the School of Computer and Systems Sciences, Jawaharlal Nehru University, New Delhi, India (e-mail: kbharadwaj@gmail.com).
[PDF]
Cite:Shweta Tyagi and Kamal K. Bharadwaj, "A Hybrid Recommender System Using Rule-Based and Case-Based Reasoning," International Journal of Information and Electronics Engineering vol. 2, no. 4, pp. 586-590, 2012.
