摘要
协同过滤是电子商务推荐系统中应用最广泛的算法,传统协同过滤算法在面临数据稀疏性问题时,其相似度计算不够准确,影响了推荐系统的准确度。鉴于此,提出了一种基于项目属性聚类及相似度优化的协同过滤算法。首先,利用杰卡德系数衡量项目间属性距离,利用改进后的K-means算法对项目聚类;然后,计算目标项目与每个类中心的属性距离,设定阈值并筛选出小于阈值的类,将搜索范围缩小到对应的类中;最后,引入属性权重概念,将评分与属性权重相结合,优化相似度计算。在MovieLens数据集上的实验结果表明:改进后的算法能够显著提高推荐的准确度、查准率和覆盖率。
Collaborative filtering is the most widely used algorithm in e-commerce recommendation system.Traditional collaborative filtering algorithm faces the problem of data sparsity,which leads to inaccurate similarity calculation and affects the accuracy of recommendation system.In view of this,a collaborative filtering algorithm was proposed based on item attribute clustering and similarity optimization.Firstly,the attribute distance between items was measured by Jaccard coefficient,and the items were clustered by improved K-means algorithm.Then the classes that were less than the thre-shold was filtered out to narrow the search scope to the corresponding classes,after calculating the attribute distance between the target item and the center of each class and setting the threshold.Finally,the concept of attribute weight was introduced to optimize the similarity calculation by combining the score and attribute similarity.Experimental results on MovieLens dataset show that the improved algorithm can significantly improve the accuracy,precision and coverage of recommendation.
作者
苏凯
张萱
付静
SU Kai;ZHANG Xuan;FU Jing(Dept. of Management Engineering and Equipment Economics, Naval Univ. of Engineering, Wuhan 430033, China;Teaching and Research Support Center, Naval Univ. of Engineering, Wuhan 430033, China)
出处
《海军工程大学学报》
CAS
北大核心
2022年第2期20-26,共7页
Journal of Naval University of Engineering
基金
国家自然科学基金资助项目(61802425)。
关键词
推荐系统
协同过滤
杰卡德系数
项目属性聚类
recommended system
collaborative filtering
Jaccard index
items attribute clustering
