ŞENYÜREK, Edip Turgut Özal Üniversitesi - Ankara Meslek Yüksekokulu, Turkey , POLAT, Hüseyin Anadolu Üniversitesi - Mühendislik Fakültesi - Bilgisayar Mühendisliği Bölümü, Turkey

EFFECTS OF BINARY SIMILARITY MEASURES ON TOP-N RECOMMENDATIONS

34098

Shopping over the Internet through several e-commerce sites is receiving increasing attention. Customers want to purchase those products that they might like without wasting time and/or money. To help their customers, many online companies provide top-N recommendations by means of recommender systems. Similarity measures used to find out the most similar entities might affect the overall performance of top-N predictions. Although there are various binary ratings-based similarity metrics, their effects on accuracy and online efficiency of top-N recommendations have not been deeply studied. In this study, we investigate seven well-known binary ratings-based similarity metrics in terms of both preciseness and efficiency while providing top-N recommendations. To compare them with respect to accuracy and competence, we perform several experiments based on two well-known real data sets. We modify top-N recommendation algorithm in such a way so that the most similar users’ data are involved in recommendation process. We also study how varying controlling parameters affect overall performance with different similarity metrics. We analyze our empirical results and provide some suggestions.

55

Similarity metric , Top , N recommendation , Accuracy , Online efficiency.

Anadolu University Journal of Science and Technology. A : Applied Sciences and Engineering

65