Improving collaborative filtering recommendations by estimating user preferences from clickstream data
作者：
Highlights：
• We propose a method for creating a user-item rating matrix of high quality.
• Our method uses item-choice probabilities estimated from clickstream data.
• We test two collaborative filtering algorithms: user-based one and NMF-based one.
• Our method substantially improves recommender performance of the two algorithms.
• We discuss and detect useful features reecting user preferences for items.
摘要
•We propose a method for creating a user-item rating matrix of high quality.•Our method uses item-choice probabilities estimated from clickstream data.•We test two collaborative filtering algorithms: user-based one and NMF-based one.•Our method substantially improves recommender performance of the two algorithms.•We discuss and detect useful features reecting user preferences for items.
论文关键词：Collaborative filtering,User preference,Rating matrix,Clickstream data,E-commerce,Recommender system
论文评审过程：Received 24 October 2018, Revised 12 July 2019, Accepted 13 July 2019, Available online 16 July 2019, Version of Record 19 August 2019.
论文官网地址：https://doi.org/10.1016/j.elerap.2019.100877
