.png){kind=icon}
返回
Modeling interaction behavior and preference decline for live stream recommendation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.dss.2023.114146.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
In recent years, live streaming has experienced rapid growth and become an influential way to engage people online. How to recommend live streams to viewers to improve user experience is the core business problem of live streaming platforms. On such platforms, viewers frequently change live streams to watch in each session for the enjoyment of the watching process. Live interaction along with streaming, which enables broadcasters and viewers to better connect with each other, is a core feature of live streaming and plays important roles in viewers' watching behaviors. How to model the changing process of viewer preference during each watching behavior and capture the influence of live interaction is key for understanding viewer behaviors and making effective recommendations of live streams but has not been well studied in the literature. To address this research gap, we incorporate user behavior theory into data driven modeling method and propose an interaction-aware predictive framework for live stream recommendation. Specifically, we develop a novel threshold-based modeling framework to systematically model viewers' watching behaviors. Guided by the theory of hedonic decline, we further model the process of viewers' preference decline during watching each live stream with consideration of live interactions, shedding light on the roles of different kinds of live interactions on viewers' watching behaviors. Comprehensive experiments conducted on a real-world data set demonstrate that our proposed framework can enhance both watching duration prediction and live stream recommendation. Besides, our proposed framework sheds light on understanding and explanation of viewers' watching behaviors.
Keyword:
User interaction
Live streaming
Recommendation system
Preference decline
User behavior analysis
Explainable artificial intelligence
期刊
IF:
6.8
论文数:
3.8K
被引数:
1.5W
机构
引用论文
.png){kind=icon}
分享收藏Extrinsic versus Intrinsic Rewards for Contributing Reviews in an Online Platform在线平台上贡献评论的外在奖励与内在奖励
PREAI
分享收藏Spillover Effects of Financial Incentives on Non-Incentivized User Engagement: Evidence from an Online Knowledge Exchange Platform财务激励对非激励用户参与度的溢出效应: 来自在线知识交流平台的证据
PREAI
分享收藏
查看更多内容