摘要
为提高用户兴趣挖掘的准确性,实现更加精准的用户个性化推荐,提出一种融合标签和长短期兴趣的矩阵分解推荐算法。利用用户使用各标签的次数和生命周期挖掘用户的长短期兴趣,计算用户标签偏好值;利用用户标签偏好值比较用户间的兴趣,获得更加精准的用户间兴趣相似度;将用户间兴趣相似度引入矩阵分解模型,预测项目评分并进行推荐。实验结果表明,该算法挖掘出的用户兴趣比其它推荐算法准确。
To improve the accuracy of users’interest mining and achieve more accurate users’personalized recommendations,a matrix decomposition recommendation algorithm fusing tag and long-term and short-term preference was presented.The number of times and life cycles that users use each tag were applied to mine their long-term and short-term preference,and user-tag pre-ference values were calculated.Comparing the interest between users based on the user-tag preference values,more accurate users’interest similarities were obtained.The users’interest similarities were introduced into the matrix factorization model to predict the items’ratings and make recommendations.Experimental results show that the users’interest mined using this algorithm is more accurate than that using other recommendation algorithms.
作者
姬璐
于万钧
陈颖
JI Lu;YU Wan-jun;CHEN Ying(School of Computer Science and Information Engineering,Shanghai Institute of Technology,Shanghai 201418,China)
出处
《计算机工程与设计》
北大核心
2023年第3期777-783,共7页
Computer Engineering and Design
基金
国家自然科学基金项目(61976140)。
关键词
用户个性化推荐
协同过滤推荐算法
矩阵分解
标签信息
长短期兴趣
用户标签偏好值
兴趣相似度
user personalized recommendation
collaborative filtering recommendation algorithm
matrix decomposition
tag information
short-term and long-term preference
user-tag preference value
interest similarity