一种改进的协同过滤图书推荐算法

Improved Collaborative Filtering Algorithm for Book Recommendation

随着图书馆文献资源和评价信息的日益丰富和增长,图书推荐系统已成为目前最受读者欢迎的应用。传统的推荐算法通常采用协同过滤算法基于相似性为用户推荐适合的书单,但评价数据的稀疏特征和推荐中过拟合等因素会影响推荐系统的准确性,导致推荐的质量较差。为此,提出一种基于用户数据处理的增强协同推荐算法,通过剔除不相关用户来降低稀疏性和过拟合问题,并基于图书馆采集的评价信息作为验证数据,将该算法与传统算法进行比较,实验表明,在RSME、准确率和召回率等指标方面有较大的提升。

Along with the enrichment of library literature resources and evaluation information,the book recommendation system has become the most popular application.Traditional recommendation algorithms usually use collaborative filtering algorithm to recommend favorite books to readers based on similarity.However,the data sparse and over fitting affects the accuracy of recommendation system,resulting in poor recommendationquality.Therefore,an enhanced collaborative recommendation algorithm based on user data processing is proposed to resolve the sparsity and over fitting problems by eliminating irrelevant users.Evaluating rating information collected by the library as the verification data,the algorithm is compared with the traditional algorithm,and the experiment shows that it has a great improvement in RSME,accuracy and recall.