Predicting Students’ Academic Outcome in Econometrics: A Comparative Analysis between Traditional and Machine Learning Methods

Abstract

The Sustainable Development Goals acknowledge education as an important producer of human capital and sustainable economic development. As the educational engagements becoming complex, present a challenge to the applicability of traditional methods of analysis. This paper is a comparison of traditional and AI-enhanced machine learning (ML) methods to analyze student performance in Econometrics at Prince Abubakar Audu University, Nigeria. The study utilized 897 students sample and 13 features of student’s demography and academic performance attributes in Econometrics prerequisite undergraduate Economics courses as Statistics for Economists, Mathematics for Economists and Macroeconomics courses. The logistic regression model that explained traditional methods, found six predictors that significantly affected the academic performance of students in Econometrics but with a low accuracy. On the contrary, ML algorithms used in the study, such as K-Nearest Neighbor, Random Forest, and Support Vector Machine (SVM), worked much better. SVM had the highest accuracy of 85.48%. The study findings show the effectiveness of AI-enhanced methods in processing complex educational data compare to conventional approach. The research contributes to the desire need of more widely used of the innovative approach to produce credible information and inform decision-making based on the data to improve the learning outcomes and quality of education in Nigerian higher institutions.