Research On Universal Recommendation Algorithm In Adaptive Testing System

With the development of psychometric theory and network technology,computerized adaptive testing(CAT)has more and more advantages than traditional testing.In CAT test,the item selection strategy determines the accuracy and rationality of the test.The current strategy of item selection is mostly based on item response theory(IRT)and cognitive diagnosis theory(CDT).Item response theory predicts the probability of a student who the Ability level is ? to correctly answer the next question by three optional parameters: project difficulty parameter,guess parameter,discrimination parameter.This method is suitable for evaluating the overall ability of students,but it can not point out the knowledge state of students.While Cognitive diagnosis can not only evaluate students' ability,but also point out the state of knowledge.Based on the theories and algorithms of cognitive diagnosis and collaborative recommendation,a general test recommendation system is designed in this paper.Firstly,according to the rule space model and attribute hierarchy model of cognitive diagnosis theory,the expected response mode(ERM)of knowledge in a certain field is Available,and we can obtain the master degree of each attribute by analyzing students' test result.in the meanwhile,we can get the the knowledge state of students by calculating.Secondly,after getting the students' knowledge state,we can calculate the test questions which are suitable for the students' ability level by using Bayesian classification method to analyze the students' test records.Finally,according to the students' knowledge state and the level ability,the method of collaborative recommendation is used to recommend the test questions to the student.At the end of this paper,a recommendation system is designed and implemented the algorithm of this paper mentioned,which completes the leap from theory to practice.