面向个性化服务的用户兴趣偏移检测及处理方法

Personalized Service Oriented User Interest Shift Detection and Processing

个性化服务系统的目标是根据不同用户的兴趣喜好为不同用户提供针对性服务,其核心是建立关于用户兴趣的描述,即用户兴趣建模。然而,现实生活中用户兴趣常常发生不可预测的变化,兴趣偏移问题一直困扰着建模技术,阻碍个性化服务系统性能的进一步提高。为了寻找切实可行的方法解决兴趣偏移问题,本文针对用户兴趣建模的兴趣偏移问题进行系统的研究,着重分析了兴趣偏移的检测方法和处理机制,对时间窗口、遗忘模型、长短期模型等隐式调整方法以及主要显式检测方法和技术进行了系统评述,并在此基础上提出了针对兴趣偏移问题的进一步研究方向。

The aim of personalized service system is to provide personalized services to different users with various interests and loves, and the core of which is to establish the description of user＇s interests, i.e. modeling the user＇s interests. However the user＇s interests suffer from the problem of shift in the practical life, and this problem has troubled the modeling technique and blocked the further improvement of personalized service system. In order to find practical ways to solve the interest shift problem, on the basis of systematic study on it, this paper focuses on an analysis of the interest shift detection methods and treatment mechanism, and systematically reviews some implicit methods including time window, forgetting model, long-short term model and some main explicit methods and techniques. On this basis we indicate some ways for further research on interest shift problem.