基于社会化标签的用户标注行为和时间因素的个性化推荐方法研究  

Personalized Recommendation Based on User Tagging Behavior and Time Factors in Social Tags

作　　者：郭雪梅[1] GUO Xue-mei(Library of Tianjin Medical University,Tianjin 300070,China)

机构地区：[1]天津医科大学图书馆,天津300070

出　　处：《情报科学》2020年第2期68-74,共7页Information Science

基　　金：2018年度天津市哲学社会科学规划项目“基于人工智能的公众医学信息服务平台构建”(TJTQ18-010).

摘　　要：【目的/意义】为了提高信息服务的质量,文章融合"用户-标签-资源项"关系模型以及时间因素对于用户标注资源的影响,提出了一种个性化推荐方法。【方法/过程】首先建立起"用户-标签-资源项"三者之间的关系模型,分别计算用户对标签的偏好程度以及资源与标签的相关程度,以此为基础进行用户相似性和资源项相似性的度量;然后,考虑标签使用的时间因素对用户兴趣偏好的影响结合基于用户标注行为的用户相似性以及资源项相似性度量方法提出了改进的个性化推荐方法。【结果/结论】提出了一种综合标签和时间因素的推荐算法,该方法利用标签使用频率描述用户偏好,并结合标签使用的时间因素动态更新用户偏好,提高推荐精度。该方法应用于医学信息服务应用场景之中,并收集实验数据,最后将提出的方法与其他基于标签信息的协同过滤推荐方法在实验数据集进行比较,实验结果发现该方法在推荐效果上优于对比方法。【Purpose/significance】In order to improve the quality of information service,this paper proposes a personalized recommendation method,which integrates the relationship model of"User-Tag-Item"and the Effect of time factor on user tagging resources.【Method/process】Firstly,the relationship model of"user-tag-resource item"is established to calculate the user’s preference for tags and the correlation between resources and tags.Then,the user similarity and resource item similarity are measured based on the model.Then,considering the influence of the time factor of tag usage on user’s interest preference,the user similarity based on user’s tagging behavior is combined.An improved personalized recommendation method is proposed based on the similarity measurement of resource items.【Result/conclusion】This paper proposes a recommendation algorithm which integrates user tags and time factors.This method describes user preferences by tagging behavior frequency,and dynamically updates user preferences by tagging behavior time factors in order to improve recommendation accuracy.This method is applied to medical information service application scenarios,and experimental data are collected.Finally,the proposed method is compared with other collaborative filtering recommendation methods based on tag information in experimental data sets.The experimental results show that the proposed method is superior to the comparison method in recommendation effect.

关 键 词：社会化标签 用户标注行为 个性化推荐 相似性度量 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]