引用本文:张敏军, 华庆一, 贾伟, 陈锐, 姬翔.基于深度神经网络的个性化推荐系统研究[J].西南大学学报(自然科学版),2019,41(11):104~109
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 35次   下载 34 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于深度神经网络的个性化推荐系统研究
张敏军, 华庆一, 贾伟, 陈锐, 姬翔1,2
1. 西北大学 信息科学与技术学院, 西安 710127;2. 宁夏大学 新华学院, 银川 750021
摘要:
为了有效地提升海量文献信息检索过程中的用户个性化满足程度,该文设计了一个全新的个性化推荐系统.在这个系统中,核心算法是基于深度神经网络的个性化推荐方法.此方法构建的深度神经网络,包含了嵌入层、编码层、个性化特征融合层、解码层4个层次,从而准确地反应用户的个性化需求并完成查询推荐.以基于DNN网络的文献检索方法、基于Segnet网络的文献检索方法、基于Seq2Seq网络的文献检索方法为对比算法,针对计算机、通信、机械、电气、建筑、历史、政治、经济、数学、英语10类文献数据进行查询推荐实验,比较4种方法检索结果与用户需求的吻合程度.实验结果表明:该文提出的基于深度神经网络的检索方法,其检索结果的用户个性化需求吻合度高于其他3种方法近10个百分点,对于英语类文献检索结果的个性化需求吻合度,甚至达到了90.2%,这充分说明了该文提出的检索方法和构建的个性化推荐系统有效.
关键词:  个性化推荐  深度神经网络  个性化融合  推荐查询
DOI:10.13718/j.cnki.xdzk.2019.11.013
分类号:TP393
基金项目:国家自然科学基金项目(61272286);高等学校博士学科点专项科研基金项目(20126101110006).
A Personalized Recommendation System Based on Deep Neural Network
ZHANG Min-jun, HUA Qing-yi, JIA Wei, CHEN Rui, JI Xiang1,2
1. College of Information Science and technology, Xibei University, Xi'an 710127, China;2. Xinhua College, Ningxia University, Yinchuan 750021, China
Abstract:
In order to effectively improve the personalized satisfaction of users in the process of massive document information retrieval, this paper designs a new personalized recommendation system, whose core algorithm is the personalized recommendation method based on deep neural network. The deep neural network constructed with this method includes four layers:an embedding layer, a coding layer, a personalized feature fusion layer and a decoding layer, and can accurately reflect the user's personalized needs and complete query recommendation. In an experiment of query and recommendation with the literature data of computer, communication, machinery, electricity, architecture, history, politics, economy, mathematics and English, this method is compared with another three literature retrieval methods based on DNN, Segnet or Seq2Seq networks for the similarity between their retrieval results and users' needs. The results show that the matching degree of users' individualized needs of the proposed retrieval method is about 10% higher than that of the other three methods, its matching degree of individualized needs of English literature retrieval results being as high as 90.2%, which convincingly indicates that the retrieval method and the personalized recommendation system described in this paper is effective.
Key words:  personalized recommendation  deep neural network  personalized fusion  recommendation query
手机扫一扫看