[1]任静霞,武志峰.动态信任衰减和信息匹配的混合推荐算法[J].计算机技术与发展,2021,31(10):30-37.[doi:10. 3969 / j. issn. 1673-629X. 2021. 10. 006]
　REN Jing-xia,WU Zhi-feng.Hybrid Recommendation Algorithm Based on Dynamic Trust Decay and Information Matching[J].,2021,31(10):30-37.[doi:10. 3969 / j. issn. 1673-629X. 2021. 10. 006]
点击复制

动态信任衰减和信息匹配的混合推荐算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
31
期数:
2021年10期
页码:
30-37
栏目:
大数据分析与挖掘
出版日期:
2021-10-10

文章信息/Info

Title:
Hybrid Recommendation Algorithm Based on Dynamic Trust Decay and Information Matching
文章编号:
1673-629X(2021)10-0030-08
作者:
任静霞; 武志峰
天津职业技术师范大学 信息技术工程学院,天津 300222
Author(s):
REN Jing-xia; WU Zhi-feng
School of Information Technology Engineering,Tianjin University of Technology and Education,Tianjin 300222,China
关键词:
兴趣漂移; 信任衰减; 协同过滤算法; 矩阵分解; 推荐系统
Keywords:
interest drift; trust decay; collaborative filtering algorithm; matrix decomposition; recommendation system
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2021. 10. 006
摘要:
针对推荐系统中的传统协同过滤算法刻画邻居差异不够深入、忽略用户兴趣漂移、冷启动问题仍然存在、矩阵分解算法可解释性差且单一推荐算法适应性不强的问题,提出了一种动态信任衰减和信息匹配的协同过滤混合推荐算法(DTDIM-CF)。 该算法考察用户间公共评分个数和时间信息调整用户或物品邻居选择机制、引入时间因子造成的信任衰减概念重新定义近邻影响以改进协同过滤算法,提出一种相似度计算的主观评分规范化方法。 根据算法初步推荐和历史信息的不匹配度,将改进后的两种动态信任协同过滤算法与矩阵分解算法按照特定规则进行混合,采用二次矩阵分解解决算法的冷启动问题。 在真实 MovieLens 电影数据集上的实验表明,改进后的算法能有效降低推荐误差、提高推荐精度,混合推荐使得算法的可解释性、可扩展性均有了很大改善。
Abstract:
Aiming at the problem that the traditional collaborative filtering algorithm in the recommendation system does not describe the difference between neighbors deeply enough,ignores user interest drift,cold start problems still exist,the matrix factorization algorithm is poor in interpret ability and the single recommendation algorithm is not adaptable, a collaborative filtering hybrid recommendation algorithm ( DTDIM-CF) with dynamic trust decay and information matching is proposed. The algorithm examines the number of public ratings among users and time information to adjust the neighbor selection mechanism of users or items,introduces the concept of trust decay caused by the time factor and redefines the influence? of neighbors to improve the collaborative filtering algorithm. A standardized method of subjective scores for similarity calculation is presented. According to the initial recommendation of the algorithms and them is match of historical information, the two improved dynamic trust collaborative filtering algorithms and the matrix factorization algorithm are mixed according to specific rules, and the second matrix factorization is used to solve the cold start problem of the algorithm. The experiment in real Movie Lens data set shows that the improved algorithm can effectively reduce the recommendation error and improve the recommendation accuracy. The mixed recommendation makes the algorithm’s interpret ability and scalability have been greatly improved.
更新日期/Last Update: 2021-10-10