理论、实操慕课课程用户翘课影响因素研究

Research on the Influencing Factors of User’s Dropout in Theory and Practice Class

摘要: 慕课的高翘课率、低完成率使其难以发挥共享服务经济的优势以促进我国教育模式改革。当前学者对于慕课翘课的预测存在准确率较低和缺乏不同课程针对性分析等问题。选择新的特征变量，借助集成学习中的XGBoost算法对不同类型课程的后台数据进行建模和学员翘课预测。从而通过对比研究出不同类型课程的学员翘课影响因素。实验结果与传统机器学习算法SVM、逻辑回归，集成学习算法AdaBoost进行对比。研究证明了XGBoost算法在慕课翘课预测中的有效性，不同类型课程影响因素的差异性。其结果对平台留存率、资源有效使用率提升具有重要现实意义。

Abstract: Due to the high skipping rate and low completion rate of MOOCs, it is difficult for MOOCs to make full use of the advantages of Shared service economy to promote the reform of China’s education model. Current scholars’ prediction of skipping MOOCs has some problems, such as low accuracy and lack of targeted analysis of different courses. Select new feature variables and use XGBoost algorithm in integrated learning to model and predict the background data of different types of courses. Therefore, the factors influencing students’ skipping classes of different types of courses are studied through comparison. The experimental results are compared with the traditional machine learning algorithm SVM, logistic regression and integrated learning algorithm AdaBoost. The research proves the effectiveness of XGBoost algorithm in the prediction of skipping MOOCs, and the difference of influencing factors of different types of courses. The results are of great practical significance to the improvement of platform retention rate and effective utilization rate of resources.