数据与计算发展前沿 ›› 2021, Vol. 3 ›› Issue (3): 148-155.

doi: 10.11871/jfdc.issn.2096-742X.2021.03.013

• 技术与应用 • 上一篇    

基于自编码器与属性信息的混合推荐模型

陈子健1,2(),李俊1,*(),岳兆娟1(),赵泽方1,2()   

  1. 1.中国科学院计算机网络信息中心,北京 100190
    2.中国科学院大学,北京 100049
  • 收稿日期:2021-02-02 出版日期:2021-06-20 发布日期:2021-07-09
  • 通讯作者: 李俊
  • 作者简介:陈子健,中国科学院计算机网络信息中心,在读硕士研究生,主要研究领域为推荐系统、机器学习等。
    本文中负责模型设计、实验设计与文献撰写。
    CHEN Zijian is a graduate student in Computer Network Information Center of Chinese Academy of Sciences. His main research areas are recommender system and machine learning.
    In this paper, he is responsible for the model design, experi-ments design and paper writing.
    E-mail: chenzijian@cnic.cn|李俊,中国科学院计算机网络信息中心,研究员,博士生导师,中国科学院特聘研究员,主要研究领域为人工智能和大数据应用、互联网体系结构等。
    本文中负责研究指导,论文结构组织。
    LI Jun is a research fellow and PhD supervisor at Computer Network Information Center of Chinese Academy of Sciences, specially appointed researcher of Chinese Academy of Sciences. His main research interests are artificial intelligence and big data technical applications and future Internet architecture.
    In this paper, he is responsible for the research guidance and paper structure organization.
    E-mail: jlee@cstnet.cn|岳兆娟,中国科学院计算机网络信息中心,高级工程师,主要研究领域为空间信息处理、大数据处理等。
    本文中负责研究指导。
    YUE Zhaojuan is a senior engineer at Computer Network Information Center of Chinese Academy of Sciences. Her main research interests are spatial information processing and big data processing.
    In this paper, she is responsible for the research guidance.
    E-mail: yuezhaojuan@cnic.cn|赵泽方,中国科学院计算机网络信息中心,在读硕士研究生,主要研究领域为自然语言处理、情感分析等。
    本文中负责实验设计。
    ZHAO Zefang is a graduate student in Computer Network Information Center of Chinese Academy of Sciences. His main research areas are natural language processing and deep learning.
    In this paper, he is responsible for the experiments design.
    E-mail: zhaozefang@cnic.cn
  • 基金资助:
    面向科学大数据传输的全球科研创新平台GRP关键技术研究(241711KYSB20180002);国家自然科学基金面上项目“命名数据网络多源多路径传输控制机制研究”(61672490)

Hybrid Recommendation Model Based on Autoencoder and Attribute Information

CHEN Zijian1,2(),LI Jun1,*(),YUE Zhaojuan1(),ZHAO Zefang1,2()   

  1. 1. Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2021-02-02 Online:2021-06-20 Published:2021-07-09
  • Contact: LI Jun

RichHTML

32

PDF

649

摘要：

【目的】 传统的协同过滤推荐模型无法提取到用户与项目之间复杂的交互关系,这对于最终的推荐结果会造成一定的不良影响。 【方法】 针对这一问题,本文提出了一种混合推荐模型DAAI（Denoising Autoencoder with Attribute Information）,采用降噪自编码器提取评分矩阵中的深层次非线性特征,在此基础上,使用DNN、CNN等方式提取属性信息中隐藏的特征,最后通过多层感知机融合多种特征得到最终的预测评分。 【结论】 将该模型在电影数据集MovieLens上进行实验,与奇异矩阵分解（SVD）、概率矩阵分解（PMF）、AutoRec等传统推荐算法进行比较,实验结果表明DAAI模型具有更好的推荐效果。 【局限】 神经网络结构较为复杂,所以本文的模型相较于传统的推荐模型训练时间有所增加。

关键词: 自编码器, 卷积神经网络, 深度学习, 推荐模型

Abstract:

[Objective] The traditional collaborative filtering recommendation models cannot extract the complex interactive relationship between users and projects, which causes a certain adverse impact on the final recommendation results. [Methods] To solve this problem, a hybrid recommendation model DAAI (Denoising Autoencoder with Attribute Information) is proposed in this paper. The denoising autoencoder is used to extract the deep nonlinear features of the scoring matrix. On this basis, DNN, CNN, and other methods are used to extract the hidden features in the attribute information. Finally, a multi-layer perceptron is adopted to generate the final prediction score by aggregating various features.[Conclusions] The proposed model is tested on the MovieLens dataset and compared with the traditional recommendation algorithms such as SVD, PMF, and AutoRec. The experiment results show that the DAAI model can achieve better recommendation results. [Limitations] Due to the complex neural network structure, the training cost of our model increases slightly compared with traditional models.

Key words: autoencoder, convolutional neural network, deep learning, recommendation model