Study On Learning Diagnosis And Individualized Item Recommendation Based On Graph Embedding

In recent years,with the rapid development of the "Internet +" era and computer and other information technology,China’s modern education system has undergone fundamental changes,and the traditional way of education has gradually changed into a new intelligent and personalized way of modern education.At the same time,the focus on intelligent and personalized education is also strengthening.However,in traditional education,educators only provide general final test scores to evaluate learners,but cannot provide detailed cognitive structure and knowledge mastery.Excessive and undifferentiated learning resources provided by educators not only lead to cognitive overload of learners,but also fail to enable learners to obtain learning resources suitable for their own needs.Therefore,it is particularly important to make formative diagnosis of learners’ traits and provide adaptive feedback to meet their individual learning needs.As a typical method used for learners’ knowledge assessment in personalized question recommendation,deep knowledge tracking can analyze the sequence data of learners to do the questions,and accurately mine learners’ inner knowledge level,so as to help educators and learners personalized teaching and learning.However,there are still some problems in deep knowledge tracking,such as sparse data,unexpressible hidden relationship between knowledge points and lack of interpretation and accuracy of prediction results.Therefore,it is of great significance to explore the hidden relationship between knowledge points and improve the accuracy of deep knowledge tracking diagnosis and prediction.At the same time,cognitive diagnosis,as another method of modeling learners’ knowledge state,can make up for the shortcoming that deep knowledge tracking model cannot give learners’ specific knowledge mastery.Based on this,the graph convolutional neural network algorithm is firstly used to embed the knowledge graph containing knowledge point relationship into the deep knowledge tracking model,and the traditional deep knowledge tracking model is improved to the graph embedded-based deep knowledge tracking model.Then,combining the improved model with cognitive diagnosis,a learning diagnosis and individualized question recommendation method based on graph embedding is proposed,which further improves the diagnosis of learners’ learning state,and also improves the accuracy of question recommendation for learners.Finally,this study trained and analyzed the data of the seventh grade math midterm test in a middle school in Wuhan.First of all,the validity of the new model improved by this study was verified.Secondly,the cognitive diagnosis model is used to calculate the overall learner,different class group learners and individual learners’ mastery of different knowledge points.Finally,the paper uses the learning diagnosis based on graph embedding and the personalized question recommendation method to recommend questions suitable for learners’ own needs.It also helps educators to better understand the knowledge level of learners,so as to adjust the teaching strategy.At the same time,it also helps learners to understand their own knowledge level,so as to study in a targeted way.