Research On The Construction Of Predictive Model Of Academic Performance

With the development of peer assessment in higher education,the reform of education evaluation has been focused by the education and the society,its importance becomes increasingly prominent.Academic performance is an important indicator of educational evaluation,which is closely related to learning behavior.Teachers can effectively predict students’ academic performance by data mining of learners’ online learning behavior laws.Thereby,they can carry out timely intervention to those students who have risk for failing the course assessment.However,most researchers focus on structured data such as explicit learning behaviors,there are few studies present a comprehensive overview of the unstructured data.A large amount of unstructured information generated by peer assessment contain emotional,cognitive and metacognitive information of students.Given that,this study takes the comments and feedback which generated by peer assessment as the research object.The research has analyzed the differences in implicit behavioral patterns of students with high-and low-scores through Quantitative Content Analysis(QCA),Lag Sequence Analysis(LSA)and Social Network Analysis(SNA)from the perspectives of affection,cognition and metacognition to find the key information.The core of the study is to accurate prediction of academic performance on unstructured text data generated by peer assessment via data mining and learning analysis,and provide accurate support for teacher intervention.First,the research elaborates the development of peer assessment and academic performance prediction by combing the exiting literature at home and abroad,and definitize the theoretical basis and related concepts of peer assessment and implicit behavior.Secondly,the study combines the existing learning analysis framework model to design the academic performance framework,and constructs the academic performance prediction model from the perspective of affection,cognition and metacognition.Finally,the study used the network learning space to implement experiment,to track and analyze the differences in implicit behavior patterns(affection,cognition,and metacognition)of students with high-and low-scores in peer assessment by data mining and learning analysis techniques.On the one hand,the results confirm that high-and low-scoring students show obvious differences in implicit behaviors of affection,cognition,and metacognition.Teachers could effectively intervene students;On the other hand,the study can provide a new perspective and reference for academic performance prediction.