Font Size: a A A

Abnormal Students Prediction Based On Activity Trajectory

Posted on:2018-10-15Degree:MasterType:Thesis
Country:ChinaCandidate:S Y HanFull Text:PDF
Abstract/Summary:PDF Full Text Request
Academic abnormality students are those whose grades are unsatisfactory.The influence factors usually related to the relegation,demotion and dropout etc.Causes of academic difficulties including the addiction to the game,health problems,family-related problems and psychological problems.If abnormal students can be found early,then communicating and counseling with them in time,it can help students to escape from academic failure.The existing prediction method is that the counselor can grasp the students' learning situation by asking the classes of the class cadres and the attendance of the class.However,because of the heavy workload of the student managers,they can not grasp the students' abnormal activities in time and timeliness is poor.With the growing popularity used of the campus card in universities,a large number of students' behavior trajectories data are collected.with the analysis of these data,the rules of students' activities can be found,and these rules can show the information of students' learning enthusiasm,learning time,life regularity,and then are used to predict students' achievement.Based on the above observations,this paper studies the prediction of academic abnormal students,find the factors that can reflect the students' academic achievements from the campus card data,and use these factors to predict academic abnormalities students.This paper establisted an overall forecasting model,which includes data preprocessing,attribute correlation analysis,prediction algorithm and so on.Firstly,the pre-processing part of the data are to achieve the data which are clean-up,integration,fine-grained,multi-type,and are processed into the form applicable to the application.This paper analyzes the main factors which affect the academic achievement,and calculates the correlation coefficient related to the scores,and chooses the attributes from the number of breakfasts and the number of the entrances of the library.Secondly,we use the Lasso model based on linear regression to filter the attributes and find that the attributes are related to academic achievement.Then,two kinds of methods are used to predict the abnormal students.One is based on the classification algorithm,using C4.5,CART,naive Bayes for the prediction of abnormal students.The other is the prediction algorithm,using Lasso algorithm for fractional prediction,and then the anomaly is determined based on the data distribution.The experimental results show that the Lasso prediction model is suitable for the prediction of abnormal scores,and the prediction stability is better.The classification algorithm is suitable for predicting the abnormal students clearly labeled before the situation,can directly predict the class of students,and the prediction accuracy is higher.In this paper,the experiment using the university campus card data proves that the relationship between students' academic achievements and their daily behavior are inseparable.Predictive model can be widely used in the prediction of academic abnormal students.
Keywords/Search Tags:Academic exception, Activities trajectory, Behavior patterns, Prediction model
PDF Full Text Request
Related items