Learner Analysis And Performance Prediction Based On Online Judge System Data

With the rapid development of online education platform,data mining technology has been widely applied in online education platform.Educational data mining has become a new research direction.But nowadays,most of the researches in educational data mining are aimed at large-scale online education platform,and there are few researches on online judge system.With the extensive application of online judge system,the number of users is increasing,the problems and courses are improving constantly,and the amount of data is increasing day by day.In the huge and disorderly data,it is of great significance to excavate valuable information.It can not only help the teaching people to improve the teaching methods,but also help the learners to fully understand their own learning situation.Therefore,according to the data of Shandong University of Science and Technology online program evaluation system,this paper analyzes learners' learning behavior and predicts their achievements.The main research work of this paper is as follows:(1)This paper analyzes the interaction between learners and online evaluation system.Accurately grasping the data characteristics of the system is the primary task for learners to analyze and study.It mainly selects data from a course,including task data,problem data and learners'learning records,and preprocesses them.(2)Using clustering algorithm to analyze the behavior of learners,and the behavior of each learner and the reasons for this behavior are introduced in detail.According to cluster analysis,we can not only find learners' characteristics,but also locate abnormal learners.(3)This paper proposes a way of describing problems,which is to describe subjects by different categories of learners.Through the description of the problem,we can better reflect the relationship between learners and problems.(4)The model of learner performance prediction is established.According to the learners' submission in a course,we predict the final result of the course.Different from previous learners' performance prediction analysis,this paper added the topic factors when extracting learners' characteristics.In this paper,a variety of classification algorithms are used in the model establishment.The design contrast experiment proves that the accuracy of the model can be improved by adding factors to learners' characteristics.