融合情境感知的移动图书馆个性化推荐技术研究  

Research on Personalized Recommendation Technology of Mobile Library Based on Context Awareness Integration

作　　者：刘晓艳[1] 张潇璐[1] 赵学敏[2] 孔云[3] Liu Xiaoyan;Zhang Xiaolu;Zhao Xuemin;Kong Yun(Yunnan University Library,Kunming Yunnan 650091;Yunnan University Archives,Kunming Yunnan 650091;Kunming University of Science and Technology Library,Kunming Yunnan 650093)

机构地区：[1]云南大学图书馆,云南昆明650091 [2]云南大学档案馆,云南昆明650091 [3]昆明理工大学图书馆,云南昆明650093

出　　处：《情报探索》2021年第9期46-52,共7页Information Research

基　　金：云南省哲学社会科学规划项目“高校图书馆数字资源多目标协同优化配置研究”(项目编号:YB2019068);云南省哲学社会科学规划项目“双一流建设背景下智慧图书馆情境化知识资源推荐系统研究”(项目编号:QN2019057);云南大学人文社会科学研究基金项目“双一流背景下云南高校仪器设备采购与信息化建设研究--以云南大学为例”(项目编号:18YNUHSS003)成果。

摘　　要：[目的/意义]建立基于情境感知的移动图书馆个性化推荐模型,可以有效提高推荐结果的准确度。[方法/过程]针对移动图书馆研究了从情境信息中充分挖掘读者偏好关系的方法,进而提出一种融合情境感知的矩阵分解推荐算法,将情境感知因素引入矩阵分解模型,并将矩阵分解结果应用于协同过滤推荐,形成最终预测评分。[结果/结论]在真实数据集Book-Crossing上该模型预测准确度相比于传统的矩阵分解方法提高了36.7%,而且当数据密度从20%增加到80%,模型的预测性能得到了进一步提升。[Purpose/significance]Construction of personalized recommendation of mobile library based on context awareness could improve the accuracy of recommendation results effectively.[Method/process]The paper studies the method of fully exploiting readers’preference relationship from context information in mobile library proposes a matrix factorization recommendation algorithm integrating context awareness introduces context awareness factors into matrix decomposition algorithm and applies matrix decomposition results in collaborative filtering recommendation to get the final result.[Result/conclusion]On the Book-Crossing data set the prediction accuracy of the model is improved by 36.7%compared with the traditional matrix factorization method and the prediction performance of the model is further improved when the data density is increased from 20%to 80%.

关 键 词：情境感知 矩阵分解 移动图书馆 协同过滤推荐 

分 类 号：G250[文化科学—图书馆学]