Predicting Students' Academic Performance Based On Students' Profile And Course Similarity

"Smart campus" is an inevitable trend of campus development.It is very important to mine effective information from scattered information data to promote the development of campus activities.Student performance is the most important factor to measure students'personal abilities and teachers'teaching level.Predicting student academic performance(SAP)in advance can play a role in early warning and timely correction for students and teachers.SAP is also the most important direction in educational data mining.However,in the existing SAP researches,most of the amount of research data is generally small,and the research direction is not applicable to Chinese education situation.At the same time,the prediction of student performance in traditional education data mining research rarely takes the implicit association between courses into account.In view of the above problems,this paper studies and predicts the students' follow-up academic performance through the relevant data information of the students.The research data comes from the personal statistical data and student transcripts from grades 2014 to 2017 of the City College of Zhejiang University.Based on these data,this paper puts forward two regression models and one ensemble model to predict student performance after research and experiments.Firstly,a SAP model based on the CART regression tree of the students' profile feature library is proposed,and the student performance data and personal information data are used to predict student performance.In the modeling process,this paper using student demographic data in the data set,combined with feature engineering to analyze the students' features,find useful information to establish a "student profile" feature library,and use the CART regression tree model to establish SAP prediction based on student profile.Experiments conducted based on the existing data prove that the constructed student profile feature library can effectively describe student features.Then,a mixed multi-weighted Slope-one SAP model based on course similarity is proposed.Through the semantic similarity of course names,the similarity between courses-courses is found,and combined with the multi-weighted Slope-one collaborative filtering.The effectiveness is verified through experiments that predict the student's subsequent course performance from the student's historical performance when the data is limited.Finally,this paper combines the two models proposed above,taking into account the student's personal characteristics and the correlation between the courses,carries out the weighted average ensemble method,and proposes a student performance prediction model based on the student's profile and course similarity.The performance of the model was evaluated on the above experimental data.The experiment proves that the ensemble model has good performance and can further improve the accuracy of prediction.