文章导读
总览 评价 陈妍 1, , 马杰良 2,* , 周真争 2, ( 1、 南京信息工程大学电子与信息工程学院; 2、 南京信息工程大学电子与通信工程学院; ) 摘要: 个性化推荐系统中最经典、最常用的是协同过滤算法,但协同过滤算法存在数据稀疏性和精准性问题,导致了
陈妍1,, 马杰良2,*, 周真争2,
(
1、南京信息工程大学电子与信息工程学院; 2、南京信息工程大学电子与通信工程学院; )
摘要:
个性化推荐系统中最经典、最常用的是协同过滤算法,但协同过滤算法存在数据稀疏性和精准性问题,导致了推荐质量的降低。针对这些问题,文章提出了一种基于属性的综合推荐算法,通过根据其变化的频率将属性划分为静态对象属性和动态对象属性,对两种属性分别预估用户对项目的喜好度,且结合线上和线下用户行为数据综合寻找最近邻居,提高用户间相似度计算的精准度。实验结果表明,应用该算法不但可以解决上述问题,与传统的推荐技术相比,还能有效的提高推荐质量。
关键词:
综合推荐,属性,协同过滤,用户喜好
CHEN Yan1,, MA Jieliang2,*, ZHOU Zhenzheng2,
(
1、Nanjing University of Information Science & Technology Electronics and information engineering institute; 2、Nanjing University of Information Science & Technology Electronic and communication engineering institute; )
Abstract:
Personalized recommendation system is the most classic, the most commonly used in collaborative filtering algorithms, but the data sparseness existence collaborative filtering algorithm and precision problems, led to a lower quality recommendations. To solve these problems, the article puts forward a comprehensive recommendation algorithm based on attribute, through according to the change of frequency will be divided into the static object attribute and dynamic object properties, of the two attributes respectively forecast user preference for the project, and the combination of online and offline user behavior data comprehensive find nearest neighbors, improving the user similarity calculation precision. The experimental results show that this algorithm can not only solve the above problem, compared with the traditional recommendation technology, can effectively improve the quality of recommendation.
Tag:
点此返回栏目查看更多>>>参考论文