郭文锋1 樊 超2
（1.山西农业大学 基础部，山西 晋中 030801；2.成都理工大学 管理科学学院，四川 成都 610059）
【摘 要】MOOCs的较低完成率极大地制约了其持续发展。尽早预测学习者是否退课将有助于改善其学习效果，提高课程的完成率。本研究以学堂在线平台的37门计算机类课程中学习者产生的行为数据为研究对象，整合资源访问度和资源访问规律两类因素，构建了28个学习行为指标并分类，采用三种经典的机器学习算法支持向量机（SVM）、逻辑回归（LR）和朴素贝叶斯（NB）,探索并分析有效预测退课的指标。研究发现：①在不同类型的学习行为指标的预测上，资源访问规律指标预测效果最好且超过了指标全集的预测结果；②在预测退课的最优指标组合上，会话数、活跃天数、访问间隔天数等七个行为指标构成预测退课的最优组合。通过对影响退课的行为指标分析，可以为MOOCs平台识别退课者并实施对应的教学干预措施提供判断参考。
【关键词】MOOCs；退课预测；学习行为指标；机器学习算法
Learning Behavior Indicators Analysis for Affecting Dropout Based on MOOCs Data
GUO Wenfeng1 and FAN Chao2
(1.Basic Courses Department, Shanxi Agricultural University, Jinzhong 030801, China; 2.College of Management Science, Chengdu University of Technology, Chengdu 610059, China)
Abstract：Low completion rate of massive open online courses (MOOCs) greatly restrains the continuous development of MOOCs. Predicting learners’willingness to drop out as early as possible is helpful to improve learning outcomes and completion rate of courses. Taking learning behavior data generated by learners of 37 computer courses on the online platform of the XuetangX as object, this study firstly extracts 28 indicators of learning behavior and categorizes them by integrating two types of factors: resource accesses and resource access rules, and then explores indicators that can effectively predict dropping out behavior by applying three classic machine learning algorithms, namely Support Vector Machine(SVM), Logistic Regression(LR) and Naive Bayes(NB). It was found that, firstly, among different types of indicators, resource access rules have the best prediction effect and overwhelm the effect of the whole set of indicators. Secondly, in predicting dropout, the optimal combination of effective predictors includes seven indicators, i.e., number of sessions, active days, interval days and so on. Through the analysis of indicators of behaviour for dropping out, it is expected to provide a basis to identify dropouts and to implement teaching interventions for MOOCs platform.
Keywords：MOOCs; dropout prediction; indicators of learning behaviour; machine learning algorithm
下载：
影响MOOCs退课的学习行为指标分析.pdf
