Research And Application Of Knowledge Tracing Based On Clustering Neural Network

With the development of online education systems,online learning platforms have generated a large amount of educational data,promoting the development of personalized learning.Knowledge tracing(KT)technology can provide students with more accurate and personalized services by tracing their knowledge status.Nowadays,KT models based on deep neural networks are widely studied to enhance personalized learning.However,to ensure the practical deployment of DNN based KT models,prediction accuracy,training efficiency and interpretability should be improved in deep.In this article,we observe that clustering the features of students and problems can improve the prediction accuracy of the KT model.Based on this observation,a distributed KT scheme is proposed,which includes dataset generation,clustering algorithm selection,KT theory and KT application.The research and innovation in this article are as follows:(1)In response to the problem of missing question information and knowledge point relationships in open source datasets,this article provides two types of datasets:1)This article contributes a set of real datasets to the research of KT,which are obtained from our real school collection.2)This article proposes a data generation technique that combines Rule Space Model(RSM)and Diagnostic Classification Model(RPa LLM).Through experimental verification,the average knowledge point coverage rate of the dataset in this article is as high as 27.92%,which is 24.05% higher than the average of the open source dataset; the average noise rate of the dataset in this article is 1.21%,which is 4.44% lower than the average of the open source dataset.(2)By comprehensively analyzing the model structure and algorithm process of clustering algorithms,this article conducts research on clustering algorithms from three aspects: data distribution,clustering stability and algorithm efficiency,and proposes the optimal clustering selection model based on neural networks.Using a deep neural network model to train the mapping relationship between the data set and the optimal clustering,and infer the optimal clustering algorithm for the data set in this article.Experiments have shown that the model has extremely high accuracy,with an AUC of92%.(3)A KT model based on clustering neural networks is proposed to address the problems of low operational efficiency and low prediction accuracy in KT.This provides a feasible solution for distributed deployment of KT,thereby improving the training efficiency of KT.At the same time,clustering algorithms are used to aggregate similar features and improve the prediction accuracy of KT.This article collects real educational data for experimentation.The results show that compared with the baseline method,the prediction accuracy and training efficiency of this system have been improved by 4.08%and 67.28%,respectively.In response to the problem of poor interpretability in deep learning,this article adds a student knowledge state matrix to the knowledge tracking output layer to simulate the student’s knowledge state and improve the interpretability of the knowledge tracking model.By monitoring the changes in students’ knowledge states over multiple time periods,growth monitoring and early warning are performed.