Student performance prediction based on data mining classification techniques

Abstract

The process of predicting student performance has become a crucial factor in academic environment and plays significant role in producing quality graduates. Several statistical and machine learning algorithms have been proposed for analyzing, predicting and classifying student performance. However, these classification algorithms still posed issue in terms of the performance classification. This paper presents a method to predict student performance using Iterative dichotomiser 3 (ID3), C4.5 and Classification and Regression tree (CART). The experiment was performed on Waikato Environment for Knowledge Analysis (Weka). The experimental results showed that an ID3 accuracy of 95.9% , specificity of 95.9%, precision of 95.9%, recall of 95.9%, f-measure of 95.9% and incorrectly classified instance of 3.83. The C4.5 gave an accuracy of 98.3%, specificity of 98.3%, precision of 98.4%, recall of 98.3%, f-measure of 98.3% and incorrectly classified instance of 1.70. The CART results showed an accuracy of 98.3%, specificity of 98.3%, precision of 98.4%, recall of 98.3%, f-measure of 98.3% and incorrectly classified instance of 1.70. The time taken to build the model of ID3 is 0.05 seconds, C4.5 is 0.03 seconds and CART of 0.58 seconds. Experimental results revealed that C4.5 outperforms other classifiers and requires reasonable amount of time to build the model.

Keywords: Student performance, ID3, C4.5, CART, classification, Education data mining