A Study on Dropout Prediction for University Students Using Machine Learning

Abstract

Student dropout is a serious issue in that it not only affects the individual students who drop out but also has negative impacts on the former university, family, and society together. To resolve this, various attempts have been made to predict student dropout using machine learning. This paper presents a model to predict student dropout at Sahmyook University using machine learning. Academic records collected from 20,050 students of the university were analyzed and used for learning. Various machine learning algorithms were used to implement the model, including Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Deep Neural Network, and LightGBM (Light Gradient Boosting Machine), and their performances were compared through experiments. We also discuss the influence of oversampling used to resolve data imbalance issues in the dropout data. For this purpose, various oversampling algorithms such as SMOTE, ADASYN, and Borderline-SMOTE were tested. Our experimental results showed that the proposed model implemented using LightGBM provided the best performance with an F1-score of 0.840, which is higher than the results of previous studies discussing the dropout prediction with the issue of class imbalance.

Ensemble-SMOTE: Mitigating Class Imbalance in Graduate on Time Detection

“…In binary scenarios, researchers meticulously define the thresholds for high class imbalance, typically when the minority class constitutes less than 8% of the dataset [17,20,21,23,25] while acknowledging imbalance when the minority class falls below 35% [18,31]. For example, in a recent study, a random split allocated 85% of the dataset to training data, revealing a distribution of 53.38% for on-time graduations and 46.62% for late graduations [11].…”

“…Nevertheless, SMOTE has shown promising results in terms of Area Under the Curve (AUC), with predictive models achieving the highest AUC of 69% compared to other class imbalance techniques when dealing with minority class instances as low as 4.7% [23]. Additionally, predictive models enhanced with SMOTE have demonstrated the highest average F1-score of 76.20% after implementing feature selection with Pearson correlation [25]. Moreover, in their study, researchers observed an improvement in average recall from 75.20% to 88.20% by increasing the number of minority class instances through SMOTE.…”

“…Moreover, in their study, researchers observed an improvement in average recall from 75.20% to 88.20% by increasing the number of minority class instances through SMOTE. However, the performance may degrade when the classification becomes more challenging due to increased noise in the minority class and overlapping regions between majority and minority classes [25].…”

“…To develop and evaluate predictive models in identifying the likelihood of students graduating on time during their studies in university.   [11]  [12]    [13]  [14]  [15]  [16]  [17]  [18]  [19]  [20]  [21]  [22]  [23]  [24]  [25]  [26]    [27]  [28]  [29]  [30]  [31]  Table 1 illustrates the prevalence of minority classes within datasets utilized by researchers to tackle the challenge of class imbalance in identifying GOT. This imbalance arises when the dominance of the majority class eclipses the presence of minority classes, resulting in biased predictive models that yield unpredictable outcomes [12,15,16,27,28,30].…”