基于复合神经网络的学习风格动态分析与研究

Research on dynamic learning style identification based on hybrid Neural Networks

近年随着慕课(MOOC)等新兴教育教学手段的快速发展,大量的学习者学习行为可以被系统所记录和分析,从而为个性化教学奠定了重要基础。在Felder-Silverman学习风格模型的理论基础上,通过引入智能分析算法动态地分析和识别学习者学习风格,构建了一套融合了卷积神经网络和循环神经网络的"识别-推理"复合模型,通过学习者的线上学习行为、社区交互行为、学习内容浏览行为、点击拖动行为等学习过程识别其学习行为特征,并使用基于门控循环单元(Gated Recurrent Unit,GRU)的循环神经网络处理和预测其可能的学习风格及对学习内容形式的偏好,以更高效地为学习者提供适应于其学习风格的学习内容和路径,优化学习体验,为大规模、个性化和高质量的下一代学习平台提供技术支撑。

The emergence of Massive Open Online Courses（MOOCs）in 2012 has been boomed in recent years, learners＇ learning behaviors can be recorded and analyzed by the system. According Felder-Silverman learning style model, this paper applies a hybrid Neural Networks（NN）which combines a Convolutional Neural Networks（CNN）and connects with a Gated Recurrent Unit（GRU）based Recurrent Neural Networks（RNN）, to detecte learning style dynamically. The model is applied in analyzing the learners＇ learning behavior, including online-community interaction, learning content browsing log, click and drag the behavior, identifying learning style dynamically, and the using GRU based RNN to process and predict the possible learning styles and provides learning content and path for optimizing learning process efficiency, improving learning experience, supports large-scale, personalized and high quality education.