用户行为序列个性化推荐研究综述

Research of Personalized Recommendation Based on User Behavior Sequence

作为缓解信息过载的一种重要方式,推荐系统可以帮助用户从海量信息中快速找到有价值的信息,其应用也越来越广泛.用户行为序列个性化推荐,又简称为序列推荐,主要根据用户与物品交互行为对用户特征进行建模,进而使用不同方法捕捉用户的长期偏好和短期偏好,向用户推荐其可能感兴趣的物品.本文从用户偏好会随时间变化的视角出发,分为用户长期偏好、短期偏好和长短期偏好3个方面,重点探讨了现有用户行为序列个性化推荐研究取得的主要进展;分析了当前用户行为序列个性化推荐研究中存在的冷启动、数据稀疏和噪声干扰等主要问题,并进一步展望了该领域未来主要研究方向.

As an important way to alleviate information overload,recommendation system can help users quickly find valuable information from massive information.Its application is becoming more and more widespread.User behavior sequence personalized recommendation,also referred to as sequence recommendation,mainly models user characteristicsbasedon the sequential behavior of user interaction with items.Then uses different methods to capture long-term and short-term preferences of users and recommend items that may be of interest to users.From the perspective that user preferences change over time,this paper discusses the main progress of personalized recommendation research based on user behavior sequence from the perspective of user long-term preference,short-term preference and long and short-term preference.The main problems existing in the current research on personalized recommendation based on user behavior sequence,such as cold start,data sparseness and noise interference,are analyzed,and the main research directions in this field are prospected.