A parallel and constraint induced approach to modeling user preference from rating data
作者：
Highlights：
• Use a latent variable to describe user preference for personalized services.
• Establish a Bayesian network with a latent variable for preference estimation and rating prediction, called UPBN.
• Present constraints and initial structure and parameters for efficient UPBN construction.
• Propose Spark based and constraint induced algorithms for parallel learning of UPBN.
• Outperform some state-of-the-art models on efficiency and effectiveness by experimental results.
摘要
•Use a latent variable to describe user preference for personalized services.•Establish a Bayesian network with a latent variable for preference estimation and rating prediction, called UPBN.•Present constraints and initial structure and parameters for efficient UPBN construction.•Propose Spark based and constraint induced algorithms for parallel learning of UPBN.•Outperform some state-of-the-art models on efficiency and effectiveness by experimental results.
论文关键词：Rating data,User preference,Latent variable,Bayesian network,Expectation maximization,Spark
论文评审过程：Received 23 April 2020, Revised 26 June 2020, Accepted 29 June 2020, Available online 2 July 2020, Version of Record 4 July 2020.
论文官网地址：https://doi.org/10.1016/j.knosys.2020.106206
