论文部分内容阅读
【目的】通过北京大学在Coursera平台上运行的课程数据,对学生的辍学行为进行研究,以期预测学生的辍学点和辍学行为,改建教学慕课质量和方法。【方法】在课程数据基础上,提取19个特征,使用机器学习算法构建滑动窗口模型,动态预测学习者辍学率。【结果】模型预测准确率高,普遍在90%以上,效果稳定,支持向量机(SVM)和长短记忆网络(LSTM)方法建模效果更好。【局限】课程数据选课人数偏多,没有考虑其他课程数据稀疏问题,模型的可移植性仍需要进一步考虑。【结论】使用滑动窗口模型建模,能够帮助MOOC课程教师和设计者动态地追踪课程学习者辍学行为,准确率高,可以帮助教师通过快速的反馈来调整课程,降低辍学率。
【Objective】 Through the course data run by Peking University on Coursera platform, this paper studies the dropout behavior of students, in order to predict the dropout and dropout behavior of students, and to rebuild the quality and method of teaching MOOC. 【Method】 Based on the course data, 19 features were extracted and the sliding window model was constructed by machine learning algorithm to dynamically predict the dropout rate of learners. 【Result】 The results show that the accuracy of model prediction is high, generally over 90%, and the effect is stable. SVM and LSTM methods are more effective. [Limitations] Course data Course enrollment is too large, did not consider the sparseness of other course data problems, model portability still needs further consideration. 【Conclusion】 Modeling with sliding window model can help MOOC teachers and designers to dynamically track the dropout behavior of curriculum learners with high accuracy and can help teachers adjust the curriculum through rapid feedback to reduce the dropout rate.