دانلود Classification via clustering for predicting final marks based on student participation in forums

Abstract

This  paper  proposes  a  classification  via  clustering  approach  to predict the final marks in a university course on the basis of forum data.  The  objective  is  twofold:  to  determine  if  student participation in the course forum can be a good predictor of the final marks for the course and to examine whether the proposed classification via clustering approach can obtain similar accuracy to traditional classification algorithms. Experiments were carried out  using  real  data  from  first-year  university  students.  Several clustering algorithms using the proposed approach were compared with  traditional  classification  algorithms  in  predicting  whether students pass or fail the course on the basis of their Moodle forum usage data. The results show that the Expectation-Maximisation (EM)  clustering  algorithm  yields  results  similar  to  those  of  the best classification algorithms, especially when using only a group of selected attributes. Finally, the centroids of the EM clusters are described to show the relationship between the two clusters and the two classes of students.