基于聚类与预测填充的协同过滤图书推荐算法研究  

Research on Collaborative Filtering Book Recommendation Algorithm Based on Clustering and Prediction Fill

作　　者：涂铁 刘斌[1] TU Tie;LIU Bin(School of Information and Artificial Intelligence,Anhui Business College,Wuhu,Anhui241002,China)

机构地区：[1]安徽商贸职业技术学院信息与人工智能学院,安徽芜湖241002

出　　处：《贵州工程应用技术学院学报》2024年第3期100-105,共6页Journal of Guizhou University Of Engineering Science

基　　金：安徽商贸职业技术学院校级项目“基于个性化推荐技术的图书推荐系统研究”,项目编号:2022KYZ03;安徽省教育厅自然科学研究重点项目“AR及Core ML技术在古诗App‘绘传承’中的应用研究”,项目编号:KJ2020A1082。

摘　　要：推荐技术作为一种能够提高用户满意度的个性化技术,在注重客户需求的场景下应用十分广泛。本文针对学校图书借阅场景提出了一种组合推荐算法,该算法在常用的协同过滤算法基础上前置了Mini Batch K-means聚类算法,并在划分后的读者簇中应用了基于图书项目特征的预测评分数据填充技术,从而降低了读者—图书矩阵的稀疏度,提高了推荐效率。实验表明,该算法相对于传统的协同过滤算法,其推荐质量得到了一定程度的提高,同时其预测评分机制也可以改善新入库图书的冷启动问题。As a personalized technology that can improve users’satisfaction,recommendation technology is widely used in customer-oriented situations.In this paper,a combined recommendation algorithm is proposed for book rental service in school libraries.The algorithm presets the Mini Batch K-means clustering algorithm on the basis of the common collaborative filtering algorithm,and applies the predicted score data filling technology in the divided reader cluster based on the characteristics of the book items,so as to reduce the sparsity of the reader-book matrix.Therefore,the recommendation efficiency is largely improved.Experiments has revealed that compared with the traditional collaborative filtering algorithm,the recommendation quality of this algorithm has been improved to some extent,and the predictive scoring mechanism can also improve the cold start problem of new books.

关 键 词：图书推荐 协同过滤 聚类 数据填充 矩阵稀疏度 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]