文献类型：期刊文献

英文题名：Research on Constructing Online Learning Performance Prediction Model Combining Feature Selection and Neural Network

作者：Mi, Huichao[1];Gao, Zhanghao[1];Zhang, Qiaorong[1];Zheng, Yafeng[1]

第一作者：米慧超

通讯作者：Zheng, YF[1]

机构：[1]Henan Univ Econ & Law, Coll Comp & Informat Engn, Zhengzhou, Peoples R China

第一机构：河南财经政法大学计算机与信息工程学院

通讯机构：[1]corresponding author), Henan Univ Econ & Law, Coll Comp & Informat Engn, Zhengzhou, Peoples R China.|[1048412]河南财经政法大学计算机与信息工程学院;[10484]河南财经政法大学;

年份：2022

卷号：17

期号：7

起止页码：94-111

外文期刊名：INTERNATIONAL JOURNAL OF EMERGING TECHNOLOGIES IN LEARNING

收录：EI(收录号：20221712041503);Scopus(收录号：2-s2.0-85128828785);WOS:【ESCI(收录号:WOS:000788914700007)】；

基金：This material is based upon work supported by Industry-university cooperative education project (202101045002) and the Science and Technology Research Project of Henan Province ("Research on text classification model and application for massive open online courses").

语种：英文

外文关键词：learning performance prediction; machine learning; deep neural networks; multiple linear regression; feature selection

摘要：Learning performance prediction can help teachers find students who tend to fail as early as possible so as to give them timely help, which is of great significance for online education. With the availability of online data and the continuous development of machine learning technology, learning performance prediction in large-scale online education is gaining new momentum. Traditional prediction methods include statistical methods, machine learning and neural networks. Among them, statistical methods and machine learning have low prediction efficiency. Although neural network can improve prediction efficiency, it ignores the impact of artificial feature filtering on model performance, and cannot find key factors for performance prediction, making predictions uninterpretable. Therefore, this paper proposes an online academic performance prediction model that integrates feature selection and neural network. Multiple linear regression analysis is used for feature extraction to obtain key influence features, and then deep neural networks is used for prediction. The results show that the F1 score of our model on large-scale data set is 99.25%, which is 1.25% higher than that of other related models.