Power coefficient as a similarity measure for memory-based collaborative recommender systems
作者：
Highlights：
• Power coefficient is a more general coefficient for asymmetric binary variables.
• Simple asymmetric coefficients (only positive matches) are not suitable for CRS.
• Our approaches assume two asymmetric binary variables for the user profile.
• Power-based CRSs elect better sets of like-minded users thus outperform other CRSs.
• Priority-based prediction keeps the predicted rating within the system scale range.
摘要
•Power coefficient is a more general coefficient for asymmetric binary variables.•Simple asymmetric coefficients (only positive matches) are not suitable for CRS.•Our approaches assume two asymmetric binary variables for the user profile.•Power-based CRSs elect better sets of like-minded users thus outperform other CRSs.•Priority-based prediction keeps the predicted rating within the system scale range.
论文关键词：Web personalization,Collaborative recommender system,Similarity measure,Jaccard coefficient,Dice coefficient,Power coefficient
论文评审过程：Available online 26 March 2014.
论文官网地址：https://doi.org/10.1016/j.eswa.2014.03.025
