摘要
因子分解机是一种基于矩阵分解的机器学习方法,通过在矩阵中引入上下文信息,构建含有上下文信息的矩阵,能够很好地进行用户评分预测。隐马尔可夫模型是一种含有隐含未知参数的统计模型,使用隐藏状态能够更好地符合实际情况。传统的推荐算法在进行推荐时通常并没有引入上下文信息,这通常会影响推荐算法的效果。鉴于上下文感知推荐算法通常能有效提高推荐精度,文中通过对推荐系统引入上下文信息并为用户添加用户隐藏兴趣状态,能够更精确地对用户进行推荐。为此,提出了一种结合因子分解机和隐马尔可夫模型的方法。在公开数据集上的验证结果表明,该方法相较于一些传统的推荐算法能够有效地提升推荐精度,并且在数据量增加的情况下也有较高的推荐精度。
Factorization machine is a machine learning method based on matrix decomposition. By introducing context information into the matrix and constructing the matrix containing context information,it can predict the user rating well. Hidden Markov model is a statistical model with hidden parameters implied,using a hidden state can better meet the actual situation. Traditional recommendation algorithms usually do not introduce context information when making recommendations,which usually affects the effect of recommendation. In view of the fact that the context-aware recommender algorithm can effectively improve the precision,the users can be recommended more accurately by introducing context information into the recommendation system and adding users' hidden interest state. Therefore,we present a method of binding factorization machines and hidden Markov models. The verification results on the public data set show that the proposed method can effectively improve the recommendation accuracy compared with some traditional recommendation algorithms,and it also has higher recommendation accuracy when the amount of data increases.
作者
王晓耘
李贤
袁媛
WANG Xiao-yun;LI Xian;YUAN Yuan(School of Management,Hangzhou Dianzi University,Hangzhou 310018,China;School of Computer Science and Technology,Hangzhou Dianzi University,Hangzhou 310018,China)
出处
《计算机技术与发展》
2019年第6期85-89,共5页
Computer Technology and Development
基金
浙江省高校人文社科研究项目(GK140203204004/02)
