关键词: MOOC, 学习行为分析, 线性回归, 深度神经网络, 成绩预测, 个性化教学

Abstract: MOOC（Massive Open Online Courses） provides learners with high-quality courses, but its low completion rate has become an important factor keeping it out of the learners. With the analysis of open dataset from edX, it can be found that the gradual loss of learners is a factor of low completion rate, and there is a complex correlation between learning behavior and achievement. With predicting the grade with linear regression and DNN（Deep Neural Network）, the experiment shows that the DNN can fit the complex relationship between learning behavior and achievement better, and predict grade more accurate to forewarn the loss of learners. With the iteration of individualized teaching intervention on potential lost learners, the completion rate in MOOC can be improved well.

Key words: MOOC, learning behavior analysis, linear regression, Deep Neural Network（DNN）, performance prediction, personalized learning