基于SVD填充和用户特征属性聚类的混合推荐算法  

Hybrid Recommendation Algorithm Based on SVD Filling and User Feature Attribute Clustering

作　　者：秦灿 QIN Can(College of Computer Science and Engineering,Wuhan Institute of Technology,Wuhan 430205,China)

机构地区：[1]武汉工程大学计算机科学与工程学院,湖北武汉430205

出　　处：《电脑知识与技术》2020年第16期11-15,共5页Computer Knowledge and Technology

摘　　要：面对评分矩阵的数据量不断增加,解决数据稀疏问题并提高推荐准确率是关键,因此,本文提出基于SVD填充和用户特征属性聚类的混合推荐算法。首先利用SVD技术对评分矩阵拆分,并使用随机梯度下降法对空缺值填充;然后对用户特征属性聚类,以此缩小邻居节点的搜索范围;接着利用遗忘曲线思想改进用户的相似度公式,结合Jaccard系数和流行度思想改进项目的相似度公式;再将用户偏好和项目特征的维度加权融合;最后,将本文的SK-HCF算法和其他同类算法进行对比实验,并证明该算法的推荐准确率有明显提升。Facing the increasing amount of data in the scoring matrix,it is the key to solve the problem of sparse data and improve the accuracy of recommendation.Therefore,this paper proposes a hybrid recommendation algorithm based on SVD filling and user feature attribute clustering.Firstly,SVD technology is used to split the scoring matrix,and random gradient descent method is used to fill in the vacancy value;then cluster user feature attributes to narrow the search range of neighbor nodes;then use the forgetting curve idea to improve the user's similarity formula.Combine the Jaccard coefficient and popularity ideas to improve the similarity formula of the project;Afterwards the dimension of user preference and project characteristics is weighted;finally,compare the SKHCF algorithm of this paper with other similar algorithms and prove that the recommendation of the algorithm is accurate The rate has improved significantly.

关 键 词：推荐算法 协同过滤 奇异值分解 K均值聚类 遗忘曲线 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]