深度学习支持下多模态学习行为可解释性分析研究

A study on Interpretable Analysis of Multimodal Learning Behavior Supported by Deep Learning Learning

当前,学习行为分析已成为研究热点。基于大数据技术挖掘学生的学习行为特征,能为学习过程的改善、学习评价的优化提供重要依据。然而,现有研究却存在分析数据类型单一、实时性不强、结果准确度不高、缺乏可解释性等问题。文章从利用深度学习算法进行多模态学习分析入手,采用HDRBM(Hybrid Deep Restricted Boltzmann Machine,深度混合判别受限玻尔兹曼机)神经网络模型,建立多模态学习分析模型,为教育技术领域中利用多模态大数据挖掘学习者行为特征提供了新范式;接着从可解释性分析的角度,阐述利用深度学习算法进行多模态学习行为分析的算法设计与实现过程;通过实验表明,研究中所采用的方法与技术路线对提高学习行为分析的可解释性有较好成效。

Today,learning behavior analysis has become a research hotspot.Based on big data technology,mining students'learning behavior characteristics can provide an important basis for the improvement of learning process and optimization of learning evaluation.However,there are still some problems in the existing research,such as single type of analysis data,weak real-time performance,low accuracy of results and lack of interpretability.This paper uses deep learning algorithm to analyze multimodal learning,adopts the HDRBM(Hybrid Deep Restricted Boltzmann Machine)neural network model to establish a multimodal learning analysis model,which provides a new paradigm for using multimodal big data to mine learners'behavior characteristics in the field of educational technology.Then,from the perspective of interpretable analysis,this paper expounds the algorithm design and implementation process of multimodal learning behavior analysis using deep learning algorithm.The experimental results show that the methods and technical routes adopted in this study are effective in improving the interpretability of learning behavior analysis.