摘要
为加强师生交流,在慕课(MOOC)学习平台讨论区允许学生通过发帖的形式针对课程内容进行讨论。与此同时,学生发帖内容书写的随意性和数据量巨大的特点,给及时回帖带来了挑战。为识别可能存在的无用帖,提出了基于无监督学习的无用帖识别方法。首先对发帖内容和发帖学生行为特征进行融合优化,建立无用帖识别模型;然后采用无监督学习的方法对学生发布的帖子进行识别,以判定该贴是否为无用帖。最后,在真实数据集上的实验证明了方法的有效性。
In order to enhance the communication between teachers and students, students were allowed to discuss the course content in discussion area of MOOC by publishing posts. However, the randomness of content and the huge data volume brought new challenges to timely reply. For detecting possible useless posts, a useless post recognition model based on unsupervised learning algorithm was proposed. In the model, post content and behavior features of students who published posts were optimized by integration.Furthermore, unsupervised learning algorithm was adopted to recognize useless posts. At last, the experiment on a collection of real MOOC learning data showed the effectiveness of the proposed model.
引文
[1]蒋卓轩,张岩,李晓明.基于MOOC数据的学习行为分析与预测[J].计算机研究与发展,2015,52(3):614-628.
[2]CHANG J W.Exploring engaging gamification mechanics in massive online open courses[J].Journal of Educational Technology&Society,2016,19(2):177-203.
[3]李亚员.国内慕课(MOOC)研究现状述评:热点与趋势:基于2009—2014年CNKI所刊文献关键词的共词可视化分析[J].电化教育研究,2015(7):55-60.
[4]黄如花,刘龙.MOOC在线交流存在的问题及对策[J].图书与情报,2014(6):18-22.
[5]刘艳春,张庆普,李占奎.基于扎根理论的MOOC在线深度互动影响因素[J].开放教育研究,2017,23(5):64-73.
[6]陈德华,周蒙,孙延青,等.MR-GSpar:一种基于Map Reduce的大图稀疏化算法[J].计算机科学,2013,40(10):190-193.
[7]林煜明.面向网络评论的观点分析技术[D].上海:华东师范大学,2013.
[8]潘荣华,陈秀宏,曹翔.一种新的有监督的局部保持典型相关分析算法[J].计算机工程与科学,2015,37(6):1175-1182.