摘要
传统算法对多维信任数据进行推荐时,均衡性较差,导致用户信任度不高,为此,提出一种基于全卷积神经网络的多维信任数据协同推荐算法。构建多维信任数据的存储和服务结构模型,运用该模型提取多维信任数据的关联规则特征量;采用隐含层节点抽取机制,将提取的特征数据输入到卷积神经网络中进行自适应学习,保留重要隐含层中的关键信息;采用协同滤波方法训练隐含层和输出层的连接权值,从而实现多维信任数据协同推荐。实验结果表明,该算法的稀疏学习性能较好,降低了推荐网络模型的复杂性,提高了置信度水平。
When the traditional algorithm recommends multi-dimensional trust data,the balance is poor,which leads to the low trust degree of users.Therefore,we propose a multi-dimensional trust data collaborative recommendation algorithm based on full convolution neural network.The storage and service structure model of multi-dimensional trust data was constructed,and feature quantity of association rules of multi-dimensional trust data was extracted by using this model;we used the hidden layer node extraction mechanism to input the extracted feature data into the convolutional neural network for adaptive learning,and retained the key information in the important hidden layer;the cooperative filtering method was used to train the connection weights of the hidden layer and output layer,so as to realize the collaborative recommendation of multi-dimensional trust data.The experimental results show that the sparse learning performance of the proposed algorithm is better,which reduces the complexity of the recommendation network model and improves the confidence level of the recommendation.
作者
李晓峰
王妍玮
王建华
Li Xiaofeng;Wang Yanwei;Wang Jianhua(Department of Information Engineering,Heilongjiang International University,Harbin 150025,Heilongjiang,China;Department of Mechanical Engineering,Purdue University,West lafayette,Indianan IN47906,US;College of Computer Science and Information Engineering,Harbin Normal University,Harbin 150025,Heilongjiang,China)
出处
《计算机应用与软件》
北大核心
2020年第8期233-238,255,共7页
Computer Applications and Software
基金
国家自然科学基金项目(61803117,41071262)
教育部科技发展中心产学研创新基金项目(2018A01002)
国家科技部创新方法专项(2017IM010500)。
关键词
全卷积神经网络
多维信任数据
协同滤波
推荐算法
Full convolution neural network
Multi-dimensional trust data
Collaborative filtering
Recommendation algorithm
