Taste Transitivity for Collaborative Filtering: A Stochastic Network Dynamics Approach
构建随机演员网络模型分析电影评论数据，发现品味传递性支持协同过滤推荐，并探讨电影流行度、类型、用户属性对品味网络演化的影响，为改进推荐系统提供见解。
ABSTRACT We develop a stochastic actor‐based network model for online movie reviews and analyze social network dynamics to study the user‐movie taste networks constructed from real movie review datasets. Examining such taste networks provides useful insights into factors underlying the performance of collaborative filtering‐based recommenders. Our results show that similar taste transitivity effect does exist that support collaborative filtering methods to recommend movies based on user taste similarity. We also investigate the role of movie popularity, genre, user gender, age, and geographic location in the taste network evolution. The findings provide insights to improve current movie recommendation systems. The two‐ mode network analyses approach taken in this article will also be broadly useful in obtaining better understanding of factors that drive user appreciations for varied products.