Comparison of Regression Algorithms for Predicting Students' Academic Performance

Abstract

In recent years, higher education has reached the peak level in growth. Many Universities, educational institutes and colleges are being set up at private level and public level for the growth of the education sector and for the well being of students [1]. Each of the institutions aims at providing higher knowledge to the students by grooming their faculties and implementing best teaching and educational practices. But still the problem persists as most of the students are at risk of unemployment due to poor academic performance. DM techniques can be used to analyse vast quantity of data that is readily available in bulk, extract usable knowledge, and help decision-making in order to forecast the academic success of students. [2].
 The amount of data present in the education sector is examined, and by extracting the data from it, beneficial patterns are found. This is known as educational data mining. This paper discusses numerous regression techniques that can be used in educational institutions to forecast students' academic achievement. To forecast student performance, regression techniques such as Bayesian networks and decision trees can be used. Applications of the CART, J48, Random Forest, and ID3 decision tree algorithms are made to undergraduate student data obtained from private colleges in order to forecast how well they will perform on the final exam of the semester and determine whether or not they will be promoted for the following year. [3].