Research On Fuzzy Cognitive Diagnosis And Test Question Recommendation Methods Oriented To Personalized Learning

In recent years,with the rapid development of online teaching,the limitations of offline teaching in time and space are alleviated to a certain extent.Test question training is an important part of the tutorial.This makes personalized test question recommendation a great significance.Test question recommendation based on cognitive diagnosis is the mainstream direction of current research.However,not only are the current methods of student cognitive diagnosis inadequate,but the accuracy of the existing test question recommendation algorithm needs to be further improved.Based on student test questions answering data,this paper studies the fuzzy cognitive diagnosis student modeling for personalized learning and its application in test question recommendation.The main contributions of this thesis include:(1)Aiming at the problems of the probability of guessing right in objective questions and the mastery in subjective questions,this paper proposes an improved fuzzy cognitive diagnosis model(I-Fuzzy CDF),realizing the effective diagnosis of students’ potential traits.In the objective questions,the students’ average mastery of all knowledge points in the test questions is used as the correction factor to redefine the guessing rate of the original students’ test questions.This makes the probability of guessing right increase with the mastery of knowledge points.In subjective questions,building on the idea of knowledge points compensation,the geometric average is used to calculate the student’s mastery of the test questions,reflecting students’ mastery of test questions objectively.Results show that the I-Fuzzy CDF model can further improve the accuracy of model diagnosis.(2)Based on the I-Fuzzy CDF model,a personalized test question collaborative recommendation method—Probabilistic Matrix Factorization Based on Fuzzy Cognitive Diagnosis(PMF-FCD)is proposed.Firstly,to solve the problem of existing recommendation methods that didn’t fully consider the students’ cognitive level,the I-Fuzzy CDF model is used to diagnose students’ mastery of subjective and objective questions;secondly,a method for calculating the similarity of students based on the score characteristics of knowledge points is designed,achieving more reasonable results;finally,in consideration of the connection between students and test questions,student similarity and test question similarity are integrated into the probability matrix factorization model,increasing the accuracy of model prediction.Results show that the PMF-FCD method is more effective than the existing method in recommendation of objective questions and subjective questions.