摘要
构建文章推荐系统需要把文章向量化,然后组建一个推荐矩阵,矩阵里的元数据(数值)会影响推荐效果,如何使文章推荐矩阵元数据更好地与用户行为关联起来,这里提出了基于TFIDF算法关联到用户行为的表示和更新机制,根据用户行为的特点,给予不同的权重,最终会影响到元数据的取值,进而能量化用户行为标签的兴趣值,另一方面也能产生更好的推荐效果。
Static Building the article recommendation system needs to vectorize the article, and then form a recommendation matrix.The metadata(value) in the matrix will affect the recommendation effect. How to make the article recommendation matrix metadata better correlated with user behavior, here is based on TFIDF, The algorithm is associated with the representation and update mechanism of the user's behavior. According to the characteristics of the user's behavior, different weights are given, which ultimately affects the value of the metadata, thereby enlightening the interest value of the user's behavior tag, and on the other hand, it can produce better recommended effect.
引文
[1]高凤丽,孙连山.个性化推荐系统概述[J].技术与市场,2015,22(2):78-79.
[2]Resinick P,Varian H R.Recommender systems[J].Communications of the ACM,1997,40(3):56-58.
[3]王国霞,刘贺平.个性化推荐系统综述[J].计算机工程与应用,201248(7):66-76.
[4]GB/T 7714曹小娜.一种适用于礼品推荐领域的基于内容推荐算法[J].中国科技论文在线,2010.
[5]施聪莺,徐朝军,杨晓江.TFIDF算法研究综述[J].计算机应用,2009,29(b06):167-170.
[6]陈大力,沈岩涛,谢槟竹,等.基于余弦相似度模型的最佳教练遴选算法[J].东北大学学报(自然科学版),2014,35(12):1697-1700.