摘要
为了有效提升课程推荐覆盖度及精准度,采用深度学习算法用于学习用户和在线课程的多维度特征训练,并借助并行计算优化推荐效率。首先,获得在线课程及学习用户行为记录,并借助主成分分析进行关键特征提取;然后根据关键特征构建课程评分函数,并引入正则项提高课程推荐覆盖度;接着采用循环神经网络(Recurrent Neural Network,RNN)算法用于学习者—课程特征训练,通过历史时间序列的隐藏层循环叠加,融入历史时间序列样本对当前时刻的作用影响,提高个性化课程推荐的准确性。通过差异化设置RNN规模及参与运算的特征量,对五个不同在线学习平台的个性化课程推荐仿真,结果表明:RNN个性化课程推荐模型对五个在线学习平台的TOP10推荐覆盖度及准确度均达到了80%以上,且稳定性高。
The former,as the mainstream service mode of large-scale data,had become the focus of personalized course recommendation research.In order to effectively improve the coverage and accuracy of course recommendation,the deep learning algorithm was used to train the multi-dimensional features of learning users and online courses,and the recommendation efficiency was optimized by parallel computing.Firstly,the behavior records of online courses and learning users were obtained,and the key features were extracted by PCA algorithm;Then,according to the key features,the course scoring function was constructed,and regular items were introduced to improve the coverage of course recommendation;Then,the recurrent neural network(RNN)algorithm was used to train learner-curriculum characteristics.Through the cyclic superposition of hidden layers of historical time series,the influence of historical time series samples on the current moment was integrated to improve the accuracy of personalized curriculum recommendation.By setting the RNN scale and the feature quantity involved in the operation differently,the personalized course recommendation simulation of five different online learning platforms was carried out.The results showed that the coverage and accuracy of the RNN personalized course recommendation model for the TOP-10 recommendation of five online learning platforms were above 80%,and the stability was high.
作者
宋艳艳
宋艳
SONG Yan-yan;SONG Yan(Changzhou Vocational Institute of Light Industry,Jiangsu,Changzhou 213164;Faculty of Materials Science&Engineering,Changzhou University,Jiangsu,Changzhou 213164)
出处
《贵阳学院学报（自然科学版）》
2024年第1期36-41,共6页
Journal of Guiyang University：Natural Sciences
基金
江苏省教育科学“十四五”规划2021年度课题“以学习为中心”的高职教育教学范式改革研究”(项目编号:B/2021/03/13)
2020年度常州大学高等职业教育研究院课题“高职教育线上教学平台用户持续使用行为优化策略研究”(项目编号:CDGZ2020042)
中华职教社黄炎培职业教育思想研究规划课题“黄炎培职教思想当代启示与创新发展(项目编号:ZJS2022YB211)。
关键词
课程推荐
深度学习
数据驱动
循环神经网络
Course recommendation
Deep learning
Data driven
Recurrent neural network
