摘要
针对传统协同过滤算法过分依赖用户历史评分数据及评分数据存在严重稀疏性问题的情况,提出一种基于关联规则的协同过滤改进算法。该算法设置相似度阈值,计算近邻用户与目标用户之间的相似度,选取相似度最高的近邻用户组成邻居集,若邻居集中的所有近邻用户与目标用户的相似度都高于阈值则按照传统协同过滤进行评分预测,否则引入关联规则的算法对目标用户进行评分预测。首先,对利用Apriori算法输出的关联规则进行拆分,得到一对一、多对一两种形式的规则;其次,基于支持度和置信度构建推荐度计算方法;再次,形成引入关联规则的算法;最后,根据阈值选择相应的算法进行评分预测,将评分高的项目推荐给用户。实验结果表明:所提出的算法与传统协同过滤算法、基于用户平均值填充的协同过滤算法相比,在MAE、RMSE上都有明显下降,可以在一定程度上提高推荐质量。
Aiming at the problem that traditional collaborative filtering algorithms rely heavily on user history rating data and severe data sparsity,an improved collaborative filtering algorithm based on association rules was proposed. The algorithm sets the similarity threshold,calculates the similarity between the neighboring user and the target user,and selects the nearest neighbors with the highest similarity. If the similarities between the neighbors and the target user all are higher than the threshold,the traditional collaborative filtering is adopted. If the similarities between the neighbors and the target user all are less than the threshold the target user is scored using an algorithm that introduces an association rule. Firstly,the association rules output by Apriori algorithm are divided to obtain two-form rules of one-to-one,many-to-one. Secondly,the recommendation degree calculation method is constructed based on the support degree and confidence degree;then,the algorithm to introduce association rules is formed;Finally,according to the threshold,the corresponding algorithm is selected to perform rating prediction,and the user with high rating is recommended. Compared with the traditional collaborative filtering algorithm and the collaborative filtering algorithm based on user average filling,the experimental results have significantly decreased on MAE and RMSE. It is proved that the proposed improved collaborative filtering algorithm based on association rules can improve the recommendation quality to a certain extent.
作者
张小川
周泽红
向南
桑瑞婷
ZHANG Xiaochuan;ZHOU Zehong;XIANG Nan;SANG Ruiting(College of Computer Science and Engineering,Chongqing University of Technology,Chongqing 400054,China;Liangjiang International College,Chongqing University of Technology,Chongqing 401135,China)
出处
《重庆理工大学学报(自然科学)》
CAS
北大核心
2019年第3期161-168,共8页
Journal of Chongqing University of Technology:Natural Science
基金
国家自然科学基金资助项目(60443004
61702063)
关键词
推荐系统
协同过滤
关联规则
相似度
推荐算法
recommender system
collaborative filtering
association rules
similarity
recommended algorithm