Research On Knowledge Tracing Model Based On Topological Relationship Of Knowledge Concepts

Online education has gradually become an important supplement to classroom education in recent years,especially during the COVID-19 period,many schools have taken "online courses" as the main teaching form,and various online education systems have also begun to provide services for students.These systems first model the personalized behavior data generated by students,then use the model to provide personalized services for students,and finally achieve "teaching students according to their aptitude" to some extent.For the system,the test answers reflect the difference of knowledge mastery among different students.The knowledge tracking technology can diagnose the current knowledge mastery status of students and predict the answer situation of the next test.Therefore,it is necessary for the online education system to track the knowledge according to the answer situation of different students,so as to improve the service quality of the system.Firstly,this paper expounds the necessity of the research on knowledge tracing model from the research background and significance.Then,the literature in the field of Knowledge tracing both at home and abroad was reviewed.Deep Knowledge Tracing(DKT)and Self-Attentive Knowledge Tracing(SAKT)were introduced in detail.In view of the problems found in model learning,this study has made improvements.The main work is as follows:(1)According to the input data of students’ answers,the knowledge space theory and association rule method are used to extract the topological relationship of knowledge concepts automatically,and the topological relationship is input into the model as additional information in the form of multi-knowledge concepts association coding,so as to increase the information content of the input model.(2)In view of the limitations of the recurrent neural network structure in DKT and the attention mechanism in SAKT,a combination of the recurrent neural network structure in DKT and the attention mechanism in SAKT is proposed,so that the model can extract the implicit information in the local scope,and obtain the attention weight between knowledge points in the global scope.(3)In view of the situation that students forget knowledge in the learning process,the forget layer is proposed to be added into the model,so as to fit the real learning situation at the time level.The experimental results show that the knowledge tracing model based on the topological relationship of knowledge concepts can extract the knowledge concept relationship well,and the prediction accuracy is better than the existing knowledge tracing model.Finally,the existing problems of the model are analyzed and discussed,and suggestions for further improvement are put forward.