Educational Context-aware Cognitive Portrait Methods For Students

With the rapid development of intelligent education,various of online educational systems come out.As the basic of many intelligent education applications,the technology of student cognitive portrait like cognitive diagnosis has been widely studied.The existing student cognitive diagnosis researches usually mine the cognitive state of students from students’ response records and the related exercise information.But the rich educational context information in student learning process is ignored in cognitive diagnosis field,which has attracted great attention in traditional education domain.Therefore,diagnosing the student’s cognitive state with educational context is a study with great significance and application prospects.However,there are still some fundamental challenges in this open topic.For example,lack of exact definition about the relationship between educational context and student’s cognitive state,the difficulty of modeling complex influence from educational context and lack of common dataset including the educational context of students in cognitive diagnosis.To this end,this thesis studied these challenges in deep and proposed appropriate solution.The contents and contributions of this research can be summarized as following:Firstly,this paper analysed the Programme for International Student Assessment(PISA)and its questionnaire framework in details.And a new dataset for cognitive diagnosis was built with the public data from PISA 2015 including more than 70 countries or regions.Secondly,this paper concluded the cognitive function of existing cognitive diagnosis methods and proposed an Educational context-aware Cognitive Diagnosis(ECD)framework.The ECD framework could be expanded to specific model based the existing cognitive diagnosis method.Thirdly,this paper analysed the educational context in dataset and designed an educational context modeling network to learning the influence from educational context.Finally,this paper completemented three ECD models(i.e.,ECD-IRT,ECD-MIRT and ECD-NCD)based on three classical cognitive diagnosis methods(i.e.,IRT,MIRT and NeuralCD)and conducted extensive experiments.The results in real-world dataset demonstrate the effectiveness as well as interpretability of ECD framework.Besides,this paper discussed the difference of results in some regions,which provided a new perspective for educational context analysis.