中文摘要：

文章以国防科学技术大学梦课平台选课人数最多的8门MOOC课程数据为基础,抽取了课程因素等三个维度、共计40余项学习数据开展辍课预测问题研究。首先,文章分析了各门课程中对预测辍课最有帮助的行为数据;其次,文章根据分析结果,选取11项行为数据训练多元线性回归和神经网络两种预测模型,实验结果表明针对不同课程建立不同的预测模型对学习者进行辍课预测的准确率平均达到90%以上。这一结果对预警学习者辍课从而实施教师干预,最终提高MOOC课程中完成课程的学习者比例带来帮助。

英文摘要：

Three dimensions of 40 learning data observed from the eight MOOC courses with maximum registration presented on the MengKe platform of National University of Defense Technology were extracted to conduct the research on the dropout prediction. Firstly, the most helpful behavior data for the dropout prediction was analyzed. Secondly, two prediction models including the multiple linear regression and neural network algorithm were built based on the 11 select behavior data. The results showed that the precision rate of the corresponding dropout prediction modes for different courses could reached up to over 90 %. The results in this paper were very helpful for teachers to prevent students＇ dropout in an early time, finally improving the passing rate of the MOOC courses.