New regression model for academic achievement and new classification method for school dropout based on Artificial Bee Colony Algorithm

Abstract

This article focuses on the analysis of two major phenomena in the field of education: academic achievement and student dropout. Academic achievement corresponds to the academic performance of students, generally assessed through their grades, averages or the achievement of educational objectives. It is influenced by various factors such as personal abilities, motivation, family support and the quality of education. On the other hand, school dropout refers to the premature abandonment of studies, often caused by academic, social or economic difficulties.  These two phenomena are among the greatest challenges facing educational institutions in most countries of the world, especially in developing countries. They have serious social and economic consequences for individuals and societies. To analyze the risks resulting from these two phenomena, it is necessary to use advanced forecasting techniques and methods, including statistical methods and artificial intelligence algorithms using available data. These methods allow us to understand the factors of each phenomenon individually and predict its negative risks. In order to improve the quality of predictions, we propose in this article a new regression model based on the multiple exponential regression model and the polynomial regression model. In order to identify the impact of social, economic and personal factors of the student and his environment on school dropout, we present an innovative classification method based on a generalization of the logistic regression model, replacing the linear term with a multiple polynomial term. To estimate the coefficients of the two proposed models, we use the ABC (Artificial Bee Colony) optimization algorithm. The two proposed approaches were applied to two different databases: the regression model was used to predict academic achievement and the classification method was used to predict the risks of school dropout. We carry out comparative studies with recent methods. The results obtained showed the reliability and superiority of the proposed approaches in terms of prediction and accuracy.