Font Size: a A A

Early Warning Model For Learning Based On Educational Data Mining

Posted on:2021-12-09Degree:MasterType:Thesis
Country:ChinaCandidate:Y F LiFull Text:PDF
GTID:2517306467969829Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the continuous development of data mining and deep learning technology,it is possible for us to get valuable information from massive data and make use of it.At the same time,the development of education informatization has greatly increased the investment in the construction of Smart Campus by universities,and thus many information management systems have been deployed,such as student status management system,E-learning system,campus card system,etc.These systems store lots of data about students.How to mine valuable information from educational big data makes data-driven decision-making become the core content of Education Data Mining.Early warning for learning refers to identity at-risk students by mining and analyzing of education data such as students' background information,achievement,learning behaviors and then provide intervention.Based on multi-source heterogeneous data,combined with the education data mining and deep learning technology,this paper tries to identity at-risk students as early as possible,so as to correct the students' learning attitude and habits and improve students' performance.First,the paper studied the related research fields and processes of education data mining.Secondly,Some attributes of students' personal information,consumption,living habits and other aspects are extracted by data pre-processing.Descriptive statistics and clustering methods were used to analyze the attributes,and thus students' consumption level are divided into three categories:“high”,“medium” and “low”,which is convenient for discretization of continuous variables in the data set.Then the Apriori algorithm was used to find the correlation between the those attributes and whether the students are failed.The results showed that whether the students are failed or not is most related to the college entrance examination results and their native place,but less related to the students' bathing level,number of times of their taking school bus and sleeping late.Finally,in view of the problem that time dimension is not taken into account in the existing learning early warning,a prediction method of dividing data sets by time dimension is proposed and a deep bidirectional long short-term memory model based on Bi LSTM is proposed to predict students' achievement.The proposed model consists of input layer,Bilstm part,full connection layer part,dropout layer and output layer.In the part of Bi LSTM and FC,n-layer bilstm and full connection layer are superposed as hidden layer for deeply feature mined.At the same time,dropout layer was introduced to prevent over-fitting and improve generalization ability.Results showed that DBi LSTM-SMOTE model was capable of predicting students' achievement in the middle of the semester and trustworthy to be an early warning model.
Keywords/Search Tags:Educatioanl data mining, Deep learning, Early warning for learning, BiLSTM
PDF Full Text Request
Related items