基于行为序列的学习过程分析与学习效果预测

Learning Process Analysis and Learning Achievement Prediction with Behavioral Sequences

学习过程数据反映了学习者在学习过程中的状态。当前众多对学习者学习过程数据进行挖掘与分析的研究,大多基于学习者在某一学习行为上投入的精力和时间来开展。这些粗粒度数据并不能细致地反映学习者的认知投入水平,且部分学习行为数据对学习效果的预测正确率不高。与学习者参与度相比,学习过程中的学习行为序列,更能反映学习者学习行为轨迹、意愿与认知过程。利用滞后序列分析法对DEEDS平台上的学习过程数据的分析发现:滞后序列分析法可以清晰地揭示若干重要的学习行为序列;相较于支持向量机、逻辑斯蒂回归以及决策树等数据挖掘方法,朴素贝叶斯方法具有良好的预测性能,平均正确率超过70%。研究结果证明,学习者的学习行为序列可以为教师呈现更全面的在线学习图景,帮助教师发现学习者的学习习惯、偏好以及认知过程,辅助教师对教学过程进行反思。同时,通过行为序列数据可以较准确地预测出学习者的学习成就,继而对预测模型中关键属性进行分析,为教师在后续教学过程中采取有针对性的干预措施提供建议,达到提高教育教学绩效的目的。

The learning process data reflects the state of students at the period of learning.At present,most researches on data mining and analyzing of students’learning process had been carried out based on the students’engaged effort and time of some learning behaviors.However,on the one hand,these coarse-grained data could not reflect the detailed level of students’cognitive engagement;on the other hand,some learning behavior data could not get a high predictive accuracy of learning achievement.To contrast with the engagement of students,their learning behavioral sequences are more reflective of their learning trace,willingness and cognitive process.The students’learning process data on DEEDS platform had been analyzed via lag sequential analysis method.It shows that a number of important learning behavioral sequences can be revealed clearly by using lag sequential analysis method;compared with the predictive results of such methods of data mining as support vector machines,logistic regression and decision tree,the naive Bayes has a better predictive accuracy of learning achievement,which has the average rate of over 70%.The results obtained verify that the behavioral sequences not only show the students’learning scene graph to the teacher,but also help to find students’learning habit,preference and cognitive process and even to help rethink the teaching process.Meanwhile,it can predict students’learning achievement accurately and provide more specifically interventional suggestions to the future teaching and learning process in order to promote the performance of teaching and learning.