MOOC学习结果预测指标探索与学习群体特征分析

发布时间：2018-04-03 12:33

本文选题：MOOC　切入点：学习行为数据　出处：《现代远程教育研究》2017年03期

【摘要】：高辍学率与低参与度是MOOC面临的一个主要问题。根据学习结果预测,及时开展有效的教学干预是改善此问题的途径之一。当前基于MOOC学习行为数据进行结果预测主要以次数分析为主,较少探索其他行为指标;在预测算法上以回归分析为主,缺少不同预测算法效果的比较分析。以ed X平台上一门MOOC课程的学习行为数据为研究对象进行的探索研究发现:学习结果预测的主要参照行为指标组合为视频学习次数、文本学习次数、评价参与时长、评价参与次数和论坛主题发起数;学习次数的预测效果要好于学习时长,并与学习时长和学习次数结合后的预测效果接近;BP神经网络预测准确率要优于决策树和朴素贝叶斯网络,且预测准确率与样本数量呈正相关;而在课程学习模块的预测比较上,评价模块和文本模块的学习行为数据预测率较高,互动模块预测率最低。研究还发现,MOOC学习群体包含三类,分别是以视频学习和学习评价为主、以互动交流为辅的学习群体;以视频学习和文本学习为主、以评价参与为辅的学习群体,以及以文本学习和学习评价为主、以互动交流为辅的学习群体。
[Abstract]:High dropout rate and low participation are a major problem for MOOC.According to the prediction of learning results, timely and effective teaching intervention is one of the ways to improve this problem.At present, the prediction of results based on MOOC learning behavior data is mainly based on the frequency analysis, less exploration of other behavioral indicators, and the prediction algorithm based on regression analysis, the lack of comparative analysis of the results of different prediction algorithms.Taking the learning behavior data of a MOOC course on ed X platform as the research object, it is found that the main reference behavior indexes of learning result prediction are video learning times, text learning times, and the time of evaluation participation.The number of evaluations of the number of participants and the number of forum themes initiated; the prediction of the number of learning times was better than the length of the learning period,The prediction effect of the combination of learning time and learning times is close to that of decision tree and naive Bayesian network, and the prediction accuracy is positively correlated with the number of samples, and the prediction accuracy of course learning module is compared with that of course learning module, and the prediction accuracy of BP neural network is better than that of decision tree and naive Bayesian network.The prediction rate of learning behavior data is higher in evaluation module and text module, and the lowest in interactive module.The study also found that there are three kinds of learning groups: video learning and learning evaluation, interactive communication, video learning and text learning, evaluation and participation.And text learning and learning evaluation as the main, interactive communication as a supplementary learning groups.
【作者单位】：江南大学教育信息化研究中心;北京师范大学教育技术学院;
【基金】：2014年全国教育科学“十二五”规划教育部重点课题“基于教育大数据的学习分析工具设计与应用研究”(DCA140230) 中央高校基本科研业务费专项资金资助课题:“‘互联网+’环境下的理解性学习与认知研究”(2017JDZD07)
【分类号】：G434

【参考文献】