Comparative Analysis of Decision Tree Algorithms for Predicting Undergraduate Students’ Performance in Computer Programming
Loading...
Date
2015
Journal Title
Journal ISSN
Volume Title
Publisher
Faculty of Science, Adeleke University, Ede, Nigeria - Journal of Advances in Scientific Research & Applications, A Multidisciplinary Journal Publication of the Faculty of Science, Adeleke University, Ede, Nigeria
Abstract
Educational data are increasing tremendously with little exploration by the education managers. Hidden
knowledge can be discovered from the huge data sets available in educational databases through the use of data
mining techniques. Educational data mining (EDM) is an emerging discipline, concerned with developing
methods for exploring the unique types of data that reside in educational databases. One of the significant areas
of the application of EDM is the development of student models for predicting students’ performances in their
educational institutions. The focus of this work is to identify the optimal decision tree algorithms for predicting
students’ performance in a computer programming course taken in 200 level based on their ordinary level
results in Mathematics and Physics and their 100 level results in Mathematics and Physics courses. One hundred
and thirty one (131) students’ records from computer science programme at Kwara State University (KWASU)
between 2009 and 2013 were used. The attributes used are students’ ordinary level scores in Mathematics and
Physics, 100 level results in Mathematics and Physics courses and the score in a 200 level computer
programming course (CSC 203). C4.5 (known as J48 in WEKA) Classification and Regression Tree (CART),
and Best-First Tree (BF Tree) decision tree algorithms were used in Waikato Environment for Knowledge
Analysis data mining software to generate three classification models employed in predicting students’
performance in CSC 203. The results of these algorithms were compared using 10-fold cross validation method
in terms of prediction accuracy and computational time. Our results showed that J48 tree has the highest
prediction accuracy of 70.37% and least execution time of 0.02 seconds while CART and BFTree has prediction
accuracy of 60.44% and 60.30% respectively and both having execution time of 0.22 seconds based on the data
set used in this study. This study also revealed that previous knowledge of Mathematics and Physics both at
Ordinary level and 100 level are essential determinants of students’ performance in a computer programming
course.
Keywords: Data Mining, Educational Data Mining, Decision Tree algorithm, Students’ Performance