Application Research On Cognitive Diagnosis Of Objective Questions Based On Knowledge Attributes

Against the backdrop of the pandemic and the internet,"online education" is thriving.However,there is a problem with "online education" where educational participants(mainly teachers and students)cannot clearly grasp the mastery of students’ knowledge points.Only by understanding the mastery status of students’ knowledge points can teachers provide personalized guidance and students engage in personalized learning.Therefore,based on real data,this article combines the classic cognitive inputs,noise and gate model(DINA)with the recommendation model(Factorization Machine based neural network,Deep FM)to establish a new model,and verifies the impact of hierarchical relationships between attributes on the effectiveness of cognitive diagnosis through experiments,in order to more accurately grasp the status of students’ mastery of knowledge points.This article collects records of students in a high school in Henan Province doing physics tests,conducts Cognitive Diagnosis(CD)on objective physics questions in the dataset,and designs and implements three experiments.As the test questions are the source of experimental data,the rationality of the test questions is first studied through multiple evaluation indicators such as difficulty and differentiation,providing reliable support for the next step of the experiment.The research results indicate that all evaluation indicators are within a reasonable range,indicating that the preparation of test questions is relatively reasonable and the experimental data is reliable.Then,this article applies the new model to multiple choice questions in the dataset and proposes a cognitive diagnosis model for multiple choice questions based on knowledge attributes,namely the Deep FM-CD model.The diagnostic process involves inputting multiple-choice question data into the DINA model for preliminary cognitive diagnosis;Then preprocess the diagnosis results,that is,the status of students’ knowledge mastery and student score information,the incidence matrix of topics and attributes,and input the processed data into the recommendation model.Utilize the Factorization Machine(FM)section and the Deep neural network(DNN)section to learn the low-order and high-order interaction relationships between knowledge attributes,respectively,in order to improve the accuracy of diagnosis.Through Experiment 2,the impact of hierarchical relationships between attributes on the cognitive diagnostic performance of multiple-choice questions was verified,that is,the error of the Deep FM-CD model is smaller than that of the DINA model.Finally,for the first time,a cognitive diagnostic model based on knowledge attributes,namely the MDeep FM-CD model,was proposed for multiple topic selection in high school physics.This model mainly verifies the impact of the interaction relationship between attributes on the effectiveness of multi topic cognitive diagnosis.Due to the diversity of scoring methods for multiple choice questions,this article transforms multiple choice questions into multiple-choice questions for research.Both retaining the original data and considering the impact of interference options on cognitive diagnostic performance.The experimental results indicate that the interaction between attributes has a certain impact on the effectiveness of multi topic cognitive diagnosis,that is,the diagnostic error of the MDeep FM-CD model is smaller than that of the DINA model.