考虑数据稀疏性的图书推荐协同过滤算法仿真
Simulation of Collaborative Filtering Algorithm for Book Recommendation Considering Data Sparsity
贾丽坤 1赵亚丽 2黄晓英 1肖丹1
作者信息
- 1. 河北建筑工程学院,河北 张家口 075000
- 2. 清华大学电子工程系,北京 100084
- 折叠
摘要
图书推荐算法易忽略数据稀疏性问题,导致推荐结果与用户感兴趣内容之间存在较大的偏差.在考虑数据稀疏性的基础上提出一种图书推荐协同过滤算法,对数据预处理,通过对用户和用户之间综合信任度分析,利用分布估计算法对用户兴趣建模;构建用户兴趣簇类集,划分用户兴趣,从中选择出与检索对象最接近的邻居;计算邻近项目得分,按照从大到小的顺序排列,排名靠前的资源项即为图书推荐结果.实验结果表明,所提方法在推荐 500 本图书时,用时在 12s内,且降低了平均绝对误差和均方根误差,实现了最精准的图书推荐.
Abstract
At present,book recommendation algorithms tend to ignore the problem of data sparsity,leading to a large error between recommendation results and content.As a result,a collaborative filtering algorithm for book recom-mendation was proposed based on data sparsity.Firstly,the data was preprocessed.Based on the comprehensive trust analysis between users,the user interest was modeled by the distribution estimation algorithm.Secondly,a set of user interest clusters was constructed to divide user interests.Moreover,the neighbor closest to the retrieved object was se-lected.Furthermore,the scores of adjacent items were calculated and ranked in descending order.The top items were the book recommendation results.Experimental results show that the proposed algorithm takes less than 12s to recom-mend 500 books.Meanwhile,this algorithm reduces the average absolute error and root mean square error,thus achie-ving the most accurate book recommendation.
关键词
数据稀疏性/图书推荐/协同过滤算法/用户兴趣模型/综合信任度Key words
Data sparsity/Book recommendation/Collaborative filtering algorithm/User interest model/Compre-hensive trust引用本文复制引用
基金项目
河北省大中学生科技创新能力培育专项(202151001010544)
河北省教育厅高等学校基本科研业务费项目(2022QNJS12)
出版年
2024
NETL
NSTL
万方数据