摘要
电子商务推荐系统通过捕获用户的兴趣和偏好为用户提供个性化的选择。利用已有的主题建模技术对用户在电子商务网站上对某品牌手机的评论进行了潜在主题识别,并对主题建模技术进行了对比;运用无监督学习技术主题建模并识别文档(产品的所有评论)隐藏主题;通过一致性分数衡量主题建模技术优劣程度,反映判断的质量。实验结果表明:潜在语义分析对数据集的处理效果较好,一致性分数高,语义上话题连贯。为电商用户个性化推荐系统设计提供参考。
E-commerce recommendation system provides users with personalized choices by capturing users’interests and preferences.This paper uses the existing topic modeling technology to identify the potential topic of the user’s comments on a brand mobile phone on the E-commerce website,and compares the topic modeling technology.Using unsupervised learning technology to model and identify hidden topics in documents(all comments on products).The consistency score is used to measure the pros and cons of topic modeling technology to reflect the quality of judgment.The experiment results show that latent semantic analysis has a good effect on data set processing,high consistency score and topic coherence in semantics,which provides a reference for the design of personalized recommendation system for E-commerce users.
作者
潘海兰
杨晓环
PAN Hai-lan;YANG Xiao-huan(Research Center of Resource Recycling Science and Engineering,Shanghai Polytechnic University,Shanghai 201209,China;Shanghai Polytechnic University School of Economics and Management,Shanghai 201209,China;Cosco Shipping Technology Co.,Ltd.,Shanghai 201209,China)
出处
《信息技术》
2021年第7期48-53,共6页
Information Technology
基金
上海第二工业大学管理科学与工程课题(XXKPY-1606)。
关键词
一致性分数
协同过滤
特征提取
推荐系统
consistency score
collaborative filtering
feature extraction
recommendation system