David Leandro Wibisono, Zaenal Abidin

Prediction of Student Graduation Predicts using Hybrid 2D Convolutional Neural Network and Synthetic Minority Over-Sampling Technique

Introduction

Prediction of student graduation predicts using hybrid 2d convolutional neural network and synthetic minority over-sampling technique. Predict student graduation using a hybrid 2D CNN and SMOTE, achieving 96.31% accuracy with family background & academic data. Enhance early intervention in higher education.

Abstract

Abstract. With the rapid growth of technology, educational institutions are constantly looking for ways to improve their services and enhance student performance. One of the significant challenges in higher education is predicting the graduation outcome of students. Predicting student graduation can help educators and academic advisors to provide early intervention and support to students who may be at risk of not graduating on time. In this paper, we propose a hybrid 2D convolutional neural network (CNN) and synthetic minority over-sampling technique (SMOTE) to predict the graduation outcome of students. Purpose: Knowing the results and how the Hybrid 2D Convolutional Neural Network (CNN) and Synthetic Minority Over-sampling Technique (SMOTE) algorithms work in predicting student graduation predicates. This algorithm uses a dataset based on family background variables and academic data. Methods/Study design/approach: This study uses the Hybrid 2D CNN algorithm for the classification process and SMOTE for the minority class over-sampling. Result/Findings: The prediction accuracy of the model using SMOTE is 96.31%. Meanwhile, the model that does not use SMOTE obtains an accuracy of 95.32%. Novelty/Originality/Value: This research shows that the use of a Hybrid 2D CNN algorithm with SMOTE gives better accuracy than without using SMOTE. The dataset used also proves that family background and student academic data can be used as a reference for predicting student graduation predicates.

Review

This paper presents a timely contribution to the field of educational data mining by addressing the crucial challenge of predicting student graduation outcomes. The authors propose a hybrid model that intelligently combines a 2D Convolutional Neural Network (CNN) with the Synthetic Minority Over-sampling Technique (SMOTE) to enhance prediction accuracy. The stated purpose is to evaluate the effectiveness of this approach using a dataset comprising both family background variables and academic data. A key finding demonstrates that incorporating SMOTE significantly improves the model's predictive capability, achieving an accuracy of 96.31% compared to 95.32% without it. This research highlights the utility of a sophisticated deep learning architecture coupled with a technique to address class imbalance, leveraging a comprehensive dataset to inform early intervention strategies for at-risk students. The methodology adopted by the authors, integrating a 2D CNN with SMOTE, represents a novel and potentially powerful approach to student outcome prediction. The reported high accuracy figures are promising and suggest the model's potential for real-world application. However, the abstract's brevity leaves several critical details undefined that would be essential for a comprehensive evaluation. Specifically, the mechanism by which tabular family background and academic data are structured or transformed for input into a 2D CNN is not elaborated. Further insights into the dataset's size, specific features utilized, the architecture of the 2D CNN, and the validation strategy (e.g., cross-validation, train-test split ratio) would be beneficial. Additionally, while accuracy is provided, presenting a broader set of evaluation metrics such as precision, recall, F1-score, or AUC would offer a more robust assessment, particularly given the use of SMOTE to balance class distribution. If the full paper can substantiate these promising initial findings with detailed methodological descriptions and comprehensive evaluation, this model could serve as an invaluable tool for educational institutions seeking to proactively identify and support students facing graduation challenges. The inclusion of both academic and family background data underscores a holistic perspective on student success factors, which is a significant strength. For future work, it would be important to explore the interpretability of the 2D CNN's decision-making process when applied to non-image data, conduct comparative analyses with other state-of-the-art models, and discuss the generalizability of the findings across diverse educational settings and student demographics. Addressing these points would considerably enhance the practical utility and academic rigor of this innovative predictive model.

Full Text

Comments

You need to be logged in to post a comment.