摘要
在信息推荐系统中,传统的方法是通过对内容、行为去预测用户的兴趣点来实现信息推送。国内外研究实验结果表明,这种模型推荐性能较为显著,说明用户行为和内容是相关的。根据相关性的对称原理,文章提出了基于用户行为的Word2Vec协同推荐算法,通过神经网络模型来隐式地抽取商品和用户的相互关系并进行向量化表示,能够更好地计算商品和用户间的相似性,以达到提升模型的推荐效果和泛化能力。
In the information push system,the traditional method is to predict the user's interest points through content and behavior to achieve information push.The domestic and foreign research results show that the recommendation performance of this model is still remarkable,which also shows that user behavior and content are related.According to the symmetry principle of relativity,this paper proposes a Word2Vec collaborative recommendation algorithm based on user behavior,which implicitly extracts and vectorizes the relationship between goods and users through the neural network model,so that we can better calculate the similarity between goods and users,and improve the recommendation effect and generality of the model.
作者
张华伟
Zhang Huawei(Jiangxi University of Finance and Economics,Network and Information Administrtion Center,JiangxiNanchang 330013)
出处
《网络空间安全》
2019年第6期25-28,共4页
Cyberspace Security
基金
江西省科技计划项目(项目编号:2013ZBBE50009)
关键词
Word2Vec
词向量
协同推荐
卷积神经网络
Word2Vec
word vector
collaborative recommendation
convolution neural network