基于SOM神经网络的教学认知诊断模型研究

Research on Cognitive Diagnostic Model for Teaching Based on SOM Neural Network

认知诊断模型在评估知识点依赖关系松散型学科知识状态时,因其生成的可达矩阵较为稀疏,易导致认知诊断的效率和准确度降低。而自组织映射(Self-Organizing Maps,SOM)神经网络可在不考虑知识点依赖关系的情况下,通过自动发现样本中的内在规律和本质属性,实现对学生知识状态的分类与评估。基于此,文章构建了基于SOM神经网络的教学认知诊断模型,在“大学英语视听说”课程中开展了此诊断模型的教学应用,并从诊断结果、学习成效、满意度三个方面进行了应用效果分析,结果表明:诊断模型用于知识点依赖关系松散型学科时,能够通过自组织聚类消除学生答题过程中的误差,精准、高效地分类评估学生的知识状态;诊断模型有助于教师实施有针对性的补救教学,在提高学生的学习效率和避免成绩的两极化方面效果显著;学生对诊断模型的整体满意度也较高。文章的研究丰富了教学认知诊断理论,可为个性化学习、精准补救教学提供诊断方法和实践经验。

When cognitive diagnostic model is used to evaluate the knowledge state of subjects with loose dependency relationship of knowledge points,its generative reachable matrix is sparse,which tends to reduce the efficiency and accuracy of cognitive diagnosis.However,the self-organizing maps(SOM)neural network can realize the classification and evaluation of students’knowledge state by automatically discovering the inherent laws and essential attributes in samples without considering the dependency relationship of knowledge points.Based on this,the paper constructed the cognitive diagnostic model for teaching based on SOM neural network,carried out the teaching application of this diagnostic model in the“College English Audiovisual Speaking”course,and analyzed the application effect from three aspects of diagnostic result,learning effect and satisfaction degree.The results showed that when the diagnostic model was applied to the subject with loose dependency relationship of knowledge points,it can eliminate the errors in students’answering process by self-organizing clustering,and accurately and efficiently evaluating students’knowledge state by category.The diagnostic model was helpful for teachers to implement targeted remedial teaching,improve students’learning effectiveness and avoid the polarization of grades.Students’overall satisfaction with the diagnostic model was also high.The research of this paper enriched the teaching cognitive diagnostic theory and provided diagnostic methods and practical experience for personalized learning and precision remedial teaching.