摘要
大数据具有数据量大、数据类型众多的特点,由于用户行为的复杂性,导致大数据推荐的精度降低。为了提升大数据推荐结果的准确性,提出一种基于多访问并行特征提取的大数据推荐算法。通过爬虫技术在访问网站和分类目录网站获取网站分类标签库,识别上网终端搭载的操作系统。分析历史访问行为和兴趣网站之间的关系,根据经验提取用户访问行为特征,采用RNN提取序列特征。通过BiasSVD模型简化矩阵维度,获取目标用户预测评分,利用聚类用户最近邻的真实评分和预测评分之间的平均差值调整目标预测评分,完成大数据推荐。实验结果表明,所提算法在推荐列表为12时,其覆盖率为75%~98%之间,且提高了加速比,全面验证了所提算法的优越性。
Due to the complexity of user behavior,the precision of big data recommendations is not high.In order to improve the accuracy of big data recommendation,this article proposed an algorithm for big data recommendation based on multi-access parallel feature extraction.Firstly,crawler technology was used to obtain the classification tag library on the target website and directory website and identify the operating system on the Internet terminal.Moreo-ver,the relationship between historical access behavior and interest websites was analyzed.Based on experience,user access behavior features were extracted.Furthermore,RNN was used to extract sequence features.Meanwhile,the Bias SVD model was used to simplify matrix dimensions,thus obtaining the predicted score of target users.Final-ly,the target prediction score was adjusted by using the average difference between the real score of the nearest neighbor of clustered users and the predicted score,thus completing the big data recommendation.Experimental re-sults show that the coverage rate of the proposed algorithm is 75%~98%when there are twelve recommended lists,and the speedup ratio is also improved,which fully verifies the advantages of the algorithm.
作者
李斌
许朝阳
王尚鹏
LI Bin;XU Chao-yang;WANG Shang-peng(Concord University College,Fujian Normal University,Fuzhou Fujian 350117,China;School of Information Engineering,Putian University,Putian Fujian 351100,China;College of Mathematics and Statistics,Fujian Normal University,Fuzhou Fujian 350117,China)
出处
《计算机仿真》
北大核心
2023年第7期486-490,共5页
Computer Simulation
基金
福建省自然基金面上项目(2020J01922)
福建省教育厅中青年教师教育科研项目(JOPX20046)。
关键词
多访问并行特征
提取
大数据推荐
Multi-access parallel feature
Extraction
Big data recommendation
