基于主题热度调权的用户兴趣建模研究  

Research on User Interest Model Constructing Based on Topic Heat

作　　者：曾子明[1] 周知[1] ZENG Zi-ming, ZHO U Zhi(School of lnformation Management, Wuhan Universit）,, Wuhan 430072,Chin)

机构地区：[1]武汉大学信息管理学院,湖北武汉430072

出　　处：《情报科学》2018年第4期150-154,170,共6页Information Science

摘　　要：【目的/意义】针对不同主题下资源数量的差异对用户兴趣建模存在影响的问题,提出一种基于主题热度的兴趣建模策略,提升模型的预测能力与推荐系统的推荐效果。【方法/过程】以主题下不同资源的数量代表该主题的热度,以此对用户兴趣特征进行调权处理,并在此基础上利用向量空间模型进行兴趣表示。以抓取的"豆瓣电影"675351位用户的观影数据进行推荐实验,验证本文策略的效果。【结果/结论】实验结果显示,基于主题热度调权的兴趣建模方法的推荐准确率明显高于传统基于绝对频次的兴趣建模方法,该策略可以提升用户兴趣建模效果。[ Purpose/significance] Aiming at the problem that user interest modeling are affected by the number difference of resources under different themes,it put forward a modeling of interest based on the theme of heat optimization strategy, in order to improve the prediction ability of the model and the effect of recmnmendation system. [Method/process] The number of dift^rent resources under a theme represents the heat of the subject, which is used to adjust the user＇s interest characteristics, and based on this, the interest is expressed by the vector space model. The recommend experiments are conducted to verify the effectiveness of this strategy with the viewing data of 675351 ＂Douban movie＂ users. [Results/ conclusions ] experimental results show that the recommendation accuracy rate of interest modeling method based on the theme heat transfer is significantly higher than that of traditional interest modeling method based on absolute frequency, the strategy （＇,an optimize the effect of user interest modeling.

关 键 词：兴趣建模 用户日志 主题热度 

分 类 号：G252.0[文化科学—图书馆学]