摘要
及时有效地对学习结果进行预测,是学习分析的核心议题,也是为学习者提供个性化学习支持服务、保证学习者学习成功的关键。其中,如何寻找兼具预测效力与教学意义的学习行为特征是值得关注的问题。本研究以Cloudbag教育云平台中的108位学生为研究对象,基于特征工程的方法构建学习结果预测框架。基于文献调研和对教师的访谈,本研究将学习结果预测框架分为学生与学生交互、学生与教师交互、学生与内容交互、学生与系统交互四个维度,共包含10个特征变量;结合学生在平台中的学习行为数据,对特征变量进行量化和筛选,通过相关分析、信息增益(率)分析筛选出八个有效的特征变量,构成最终的特征集合;使用八种机器学习算法对学习结果进行预测,结果表明:随机森林算法对学习结果的预测效果优于另外七种算法,其特征集合对学习结果的预测准确率可以达到73.15%。本研究最后从有效学习行为指标和有效学习行为特征等方面对研究结果进行总结和反思,期望能够为混合式学习环境下学习分析和评估提供研究支持。
As the core issue of learning analytics,timely and effective prediction of learning outcomes is the key to ensuring learners’success and an essential means to provide learners with personalized learning support services.How to find learning features that have both predictive effectiveness and teaching significance has become a problem worthy of attention.This study built a learning outcome prediction framework by feature engineering method,based on 108 students in the CLOUDBAG Platform.Firstly,through literature review and teacher interviews,the framework was divided into four dimensions:student-student interaction,student-teacher interaction,student-content interaction,and student-interface interaction,which contained a total of ten learning features.Secondly,using students’learning behavior data in the platform,ten features were quantified and filtered,and eight effective features were selected through correlation analysis and information gain(rate)analysis to form the final feature set.Finally,learning outcomes were predicted by eight machine learning algorithms.The results showed that the random forest algorithm was better than the other seven algorithms,and the prediction accuracy can reach 73.15%.At the end of the paper,research results were summarized and reflected from the aspects of effective learning behavior indicators and effective learning features.It was expected to provide research support for learning analytics and evaluation in a blended learning environment.
作者
武法提
田浩
WU Fati;TIAN Hao(School of Educational Technology,Beijing Normal University,Beijing 100875,China)
出处
《开放教育研究》
CSSCI
北大核心
2019年第6期75-82,共8页
Open Education Research
基金
北京师范大学教育学部学科建设综合专项科研培育项目“场景驱动的个性化学习服务模型及其应用研究”(2019KYPY005)
关键词
学习分析
学习行为特征
学习结果预测
学习结果
feature engineering
learning analytics
learning features
learning outcome prediction
random forest