With the advancement of Internet technology and the increasing demand for effi-cient and flexible learning methods,online education has become an important trend in the global education field.However,with the development of online education,the problems of information overload and individual needs have emerged.How to real-ize personalized learning according to the characteristics and needs of each student has become an urgent problem to be solved.Exercises are an important part of learning re-sources.They can not only help students consolidate their knowledge,but also test their learning effects.However,most of the online test question websites currently provide a relatively simple exercise resource library,screening methods and structures,which are difficult to meet the individual needs of different learners.Therefore,it is neces-sary to study an efficient and reliable exercise recommendation algorithm,integrate the knowledge structure of the subject,learning objectives,and students’ knowledge level into the recommendation method,make recommendations to students more accurately,and improve the accuracy and pertinence of recommendations.In view of the above problems,the paper studies in detail how to use knowledge graph to represent knowledge structure and use knowledge tracing model to model stu-dents’ knowledge level,and proposes a personalized exercise recommendation strategy based on knowledge graph and knowledge tracing.The main contributions of the paper are as follows:(1)Construct a knowledge graph of junior high school mathematics subjects.Se-lect junior high school mathematics as the subject background,and construct the math-ematics subject knowledge graph through the steps of knowledge point induction,data acquisition and improvement,information extraction,and information storage.The use of knowledge graphs to organize and represent subject knowledge enables in-depth mining and analysis of knowledge,providing strong support for subsequent knowledge tracing models and personalized exercise recommendations.(2)Propose a Feature and Relation Enhanced Knowledge tracing Model(FRKT).Unlike most knowledge tracing models,which only study at the level of knowledge points,FRKT mines semantic information and difficulty information from exercise texts and learning records,and tracks students’ learning process from a more comprehensive perspective.In the stage of knowledge status update,FRKT utilizes one-way and two-way influence propagation modes,considers multiple relationships between knowledge points,and improves the performance of the model in both spatial and temporal dimen-sions.Model comparison experiments on two real data sets,Junyi and Eedi,show that FRKT outperforms other models in terms of ACC and AUC.(3)Propose a personalized exercise recommendation strategy based on knowledge graph and knowledge tracing(KGFRKT-ER).This strategy improves the accuracy and effectiveness of exercise recommendation by virtue of the ability of the knowledge graph to represent the knowledge structure and the ability of the knowledge tracing model to track and evaluate the student’s knowledge state.It also uses the simulated an-nealing algorithm to increase the diversity of recommended exercises.Through model comparison experiments on the Junyi and Eedi data sets,it can be obtained that the KGFRKT-ER proposed in the paper is superior to other comparison models in three in-dicators: recommendation accuracy,exercise diversity,and ability improvement index. |