Predicting Dropouts among a Homogeneous Population using a Data Mining Approach
Loading...
Date
2019-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The British University in Dubai (BUiD)
Abstract
Student retention is one the biggest challenges facing academic institutions worldwide. Failure to retain students not only affects the student in a negative way but also hinders institutional quality and reputation. While there are several theoretical perspectives of retention, which study the factors that cause students to drop out, more recent studies rely on a data mining and machine learning approach to explore the problem of retention.
In this research, we present a novel data mining approach to predict retention among a homogeneous group of students, with similar social and cultural background, at an academic institution based in the UAE. Our model successfully identifes dropouts at an early stage. It provides an early warning system that enables the institution to promptly intervene with assertive measures. Moreover, our model also effectively determines the top predictive variables of retention. Several researchers study retention by focusing on student persistence from one term to another while our study builds a predictive model to study retention until graduation. Moreover, other works use additional student data for predictions, thereby reducing the dataset size, which is counter productive to data mining. Our research relies solely on pre-college and college performance
data available in the institutional database.
Our research reveals that the Gradient Boosted Trees is a robust algorithm that predicts dropouts with an accuracy of 79.31% and AUC of 88.4% using only pre-enrollment data. High School Average and High School stream of study are observed to be the top predictive variables of on-time graduation when a student joins college. Our study also reveals that ensemble machine learning algorithms are more reliable and outperform standard algorithms.
Description
Keywords
student retention, homogeneous population, data mining, United Arab Emirates (UAE), academic institutions, machine learning, machine learning algorithms