基于深度神经网络的虚拟仿真实验学习效果评估研究

A Study on Learning Effect Evaluation of Virtual Simulation Experiments Based on Deep Neural Network

随着虚拟仿真实验模式越来越多地被应用于实验教学实践,成为传统实验教学的有效补充,其学习效果的评估也日益受到关注。基于虚拟仿真实验活动过程中产生的多维数据开展特征工程,将原始特征划分为学习者特征、行为特征和实验特征,将这些特征作为重要评估参数,建构虚拟仿真实验学习效果评估的深度神经网络模型,并将该模型应用于某校经管类课程教学实践的评估。结果表明,对于经管类虚拟仿真实验教学而言,基于学生基本信息数据、实验数据和实验行为数据构建的概念特征模型具有较好的判别能力;模型采用深度神经网络方法进行评估,较之传统机器学习方法,在评估效果上有较为明显的提升;通过数据驱动的评估方法有利于及时有效地把握实验的学习效果,帮助虚拟仿真实验教学工作者及时且有针对性地把握并改进相应教学环节。

As the virtual simulation experiment mode is more and more applied to the experimental teaching practice,it has become an effective supplement to the traditional experimental teaching,and the evaluation of its learning effect is also increasingly concerned.How to make use of virtual simulation experimental data and models has become a hot topic in the field of education.Feature engineering is carried out based on the multidimensional data generated in the process of virtual simulation experiments,and the original features are divided into learner features,behavioral features,and experimental features,which are important parameters for a deep neural network model constructed to evaluate the learning effect of virtual simulation experiments.Multiple evaluation indexes are used to verify and compare the models,and the model is applied to the evaluation of the teaching practice of economics and management courses in a school.The study found that:for the virtual simulation experiment teaching of economics and management,the conceptual feature model based on students'basic information data,experimental data and experimental behavior data has good discriminative ability;the model is evaluated by the deep neural network method,and the evaluation effect is significantly improved compared with the traditional machine learning method,therefore,the model bears some validity;the data-driven evaluation method is helpful to grasping the experimental learning effect in a timely and effective manner,enabling the virtual simulation experiment teaching staffs grasp and improve the corresponding teaching links more timely and pertinently.