Afiani Agus Abdillah, Yono Cahyono, Teti Desyani, Perani Rosyani

Optimalisasi Prediksi Kelulusan Mahasiswa Menggunakan Algoritma Decision Tree CART

Introduction

Optimalisasi prediksi kelulusan mahasiswa menggunakan algoritma decision tree cart. Optimalkan prediksi kelulusan mahasiswa menggunakan Decision Tree CART. Integrasikan data akademik & non-akademik, raih akurasi 92.1%, ideal untuk sistem peringatan dini pendidikan tinggi.

Abstract

Timely student graduation is a key indicator of higher education quality and institutional effectiveness. This study aims to optimize student graduation prediction using a Decision Tree algorithm based on Classification and Regression Tree (CART) by integrating academic and non-academic variables. The dataset used in this study is the open-source Student Graduation Dataset obtained from Kaggle, consisting of 379 student records with graduation status as the target variable. The research stages include data preprocessing through mean imputation for missing values, categorical variable transformation, data splitting with an 80:20 ratio, and model optimization using CART hyperparameter tuning as a form of post-pruning. Model performance was evaluated using accuracy, precision, recall, F1-score, and a confusion matrix. The experimental results show that the optimized CART model achieved an accuracy of 92.1%, with F1-scores above 0.90 for both graduation classes and a balanced trade-off between precision and recall. Furthermore, the resulting decision tree structure is relatively simple and highly interpretable. These findings indicate that the optimized CART algorithm is effective and suitable for implementation as an early warning system to support academic decision-making in higher education institutions.

Review

This study presents a timely and relevant investigation into optimizing student graduation prediction, a critical aspect of higher education quality assurance. The authors propose an approach leveraging the CART Decision Tree algorithm, integrating both academic and non-academic variables to enhance predictive accuracy. The use of an open-source Kaggle dataset with 379 student records provides a reproducible basis for the research. The paper's core contribution lies in demonstrating a robust methodology for developing an interpretable early warning system, positioning it as a valuable tool for academic decision-makers. Methodologically, the study exhibits commendable rigor. The research stages clearly outline standard machine learning practices, including thorough data preprocessing steps such as mean imputation for missing values and categorical variable transformation. The 80:20 data splitting ratio is appropriate, and the emphasis on model optimization through CART hyperparameter tuning (post-pruning) indicates a commitment to achieving optimal performance. The comprehensive evaluation metrics employed – accuracy, precision, recall, F1-score, and confusion matrix – provide a holistic view of the model's capabilities. The reported results, with an impressive 92.1% accuracy and F1-scores above 0.90 for both graduation classes, along with a balanced trade-off between precision and recall, strongly support the effectiveness of the optimized CART model. Furthermore, the inherent simplicity and interpretability of the resulting decision tree structure are significant practical advantages for real-world implementation. While the abstract demonstrates a strong foundation, the full paper would benefit from elaborating on the specific types of "non-academic variables" included, as their nature can significantly influence model generalizability. Given the use of an open-source dataset, future work could explore validation of this optimized model using institution-specific datasets to confirm its applicability across diverse educational contexts and student demographics. Despite these minor considerations, the presented research offers a promising and well-executed solution for improving student success prediction, making a solid contribution to the field of educational data mining and supporting proactive academic intervention strategies.

Full Text

Comments

You need to be logged in to post a comment.