在线学习行为投入分析框架与测量指标研究 ———基于LMS 数据的学习分析

Mining LMS Data for Behavioral Engagement Indicators in Online Learning Environments

大量研究表明，学生的学习行为投入是影响学习绩效的重要因素。基于此，本文从学习行为投入 的角度对基于学习管理系统的学习行为分析和测量进行了理论与实践的探索。文章首先在行为投入定义、分 类、评测相关研究基础上提出了在线学习行为投入分析框架，包括参与、交互、坚持、专注、学术挑战、自我监控六 个维度。之后，文章结合在线案例课程2268 名学生基于Moodle 平台的数据对在线学习行为投入测量指标进行 了统计分析，通过缺失值分析、成绩相关分析、因子分析确定21 个测量指标，以及主动交互、平均参与度、绩效努 力、学术挑战、自我监控五个投入因子。其中，四个指标对课程成绩有显著预测作用，能够预测成绩26. 9% 的变 异。文章最后对在线学习行为投入框架和测量指标、有效投入以及与相关研究结果的比较进行了讨论与反思。 本研究能够为以促进有效学习投入为目的LMS 数据分析与学习支持提供依据。

Lots of literature documented the link between behavioral engagement and positive learning outcomes. Therefore， timely measurement of learning behavioral engagement and appropriate intervenes would contribute to the improvement of learning performances. This paper presented an empirical study on indicators for measuring behavioral engagement in online learning environments through LMS data mining. Firstly， the paper proposed a behavioral engagement analysis framework of LMS based learning for identifying indicators by reviewing definitions， theories and instruments of behavioral engagement in literature. The framework has six constructs including participation， interaction， persistence， concentration， academic challenge and self-regulated learning. Then，according to the framework the paper explored the indicators of behavioral engagement based on the Moodle data of 2268 students from an online course in practice. 18 indicators with six constructs were finally identified using missing value analysis， correlation analysis with student grades， and exploratory factor analysis on the Moodle data of the student sample. Six constructs identified underlying above 18 indicators are effort on learning course content， active interaction， performance effort， academic challenge， average participation， average time spent on learning task. Thirdly， a predictive model for the case course was generated using regression analysis which incorporated four indicators relating to performance effort and active interaction and which explains 26. 5% of the variations in student grades. The paper finally conducted discussions and made reflections on the behavioral engagement indicators， effective engagement behaviors， and comparisons of relevant research results. It is expected that the current paper would provide a reference to LMS data mining and student support for the purpose of improving learning engagement in online courses.