Research And Application Of Personalized Question Recommendation Methods Based On Cognitive Diagnosis And Collaborative Filtering

In the era of educational informatization,personalized education based on the characteristics of students to achieve targeted teaching is a new trend in current research on smart education and online education.To achieve personalized learning guidance for students,it is necessary to first assess their current learning status.Cognitive diagnosis addresses the shortcomings of traditional educational measurement by providing timely and diagnostic assessment feedback to help students identify knowledge gaps and adjust their learning strategies and methods.However,existing cognitive diagnostic methods still have shortcomings.In addition,after accurately diagnosing a student’s knowledge level,effective test practice is necessary to fill in any gaps in learning.Personalized question recommendation has significant research significance in this process.The integration of information technology and education has resulted in many online learning platforms,where students can access test resources and practice more conveniently to consolidate and deepen their understanding of what they have learned.However,the massive amount of online learning resources has also led to the problem of information overload.In this situation,how to make the most of a student’s historical answering behavior data to recommend test resources that meet their learning needs is a problem worth considering.Therefore,this thesis will focus on research on cognitive diagnosis analysis and personalized question recommendation methods,with the main contributions described as follows:(1)In response to the shortcomings of existing cognitive diagnostic methods,this thesis combines deep learning with cognitive diagnostic methods in educational psychology,and improves the neural cognitive diagnostic model by further introducing the importance factor of knowledge points,guessing factor and slip factor based on students’ historical response records.The proposed model is called the Multi-Factor Neural Cognitive Diagnostic Model(Multi-Factor Neural Cognitive Diagnostic Model,MNCD),which uses deep learning techniques for model training and optimization to achieve accurate diagnosis of a student’s knowledge level.Through the student score prediction experiment,the results show that the MNCD model outperforms other traditional cognitive diagnostic models in terms of Accuracy,AUC,and RMSE.In addition,by setting up a cognitive diagnostic rationality experiment,the results demonstrate that the MNCD model’s diagnosis of a student’s knowledge level is reasonable,ensuring the interpretability of the diagnostic results.(2)In the process of test recommendation,there is a problem that the existing methods for test recommendation cannot simultaneously guarantee interpretability and accuracy of the results.This thesis propose a personalized test recommendation method called MNCD-NMF(Neural Matrix Factorization Based on Multi-Factor Neural Cognitive Diagnostic Model),which combines the neural cognitive diagnosis model with multiple factors proposed in this paper and the neural matrix factorization model in the collaborative filtering algorithm.The MNCD model is used to obtain personalized learning features of students,which are then integrated into the neural matrix factorization model to predict the probability of students answering test questions correctly and recommend suitable test questions within the appropriate difficulty range,ensuring the rationality and interpretability of the recommendation results.Experimental results show that the MNCD-NMF method can effectively improve the effectiveness of test resource recommendation and meet the personalized learning needs of students in the learning process.