Comparative Analysis of Classification Algorithms in Educational Data Mining

Abstract

Data mining is a field of computer science within Machine Learning (ML) and Artificial Intelligence. With the
advent of the internet and the increase in computing speed and storage capacity, the amount of data collected is also
on the rise in different data warehouses. Hence, the need to explore the humongous volume of data being generated
in order to extract patterns representing knowledge for decision-making arose. Data mining is a fast-gaining
application in several facets of life, and the educational sector is not left out. This work starts with a brief
overview of data mining techniques, then delved into a comparative analysis of 3 different Classification techniques such as decision tree (j48), Bayes (NaiveBayes) and Rules (oneR) on student academic performance datasets, using statistics such as F-Measure and Percentage – Correct. The methodology adopted is the CRISP-DM due to its advantages over other data mining methodologies as it is popularly used and provides a uniform framework for planning and managing a project. The tool employed in our data mining analysis is the Wekaito Environment for Knowledge Analysis (WEKA 3.0). Our results show that J48 performed better with 95.5556% prediction accuracy than the other algorithms. Consequently, it was chosen to build the prediction model of student academic performance.