Fitriani, Jasmir, Sharipuddin

Perbandingan Algoritma Naïve Bayes Dan Regresi Logistik Untuk Memprediksi Kesehatan Mental Mahasiswa Di Provinsi Jambi

Introduction

perbandingan algoritma naïve bayes dan regresi logistik untuk memprediksi kesehatan mental mahasiswa di provinsi jambi . Penelitian ini membandingkan akurasi Naïve Bayes & Regresi Logistik untuk prediksi kesehatan mental mahasiswa di Jambi. Naïve Bayes terbukti lebih unggul (akurasi 100%) untuk deteksi dini dan intervensi efektif.

Abstract

Kesehatan mental mahasiswa menjadi perhatian global karena dipengaruhi oleh tekanan akademik, masalah sosial, dan perubahan gaya hidup. Di Indonesia, sekitar 20% populasi diperkirakan mengalami gangguan mental, dengan lebih dari 12 juta orang mengalami depresi. Peningkatan kasus bunuh diri, termasuk di kalangan mahasiswa di Jambi, menunjukkan dampak tekanan yang besar terhadap kesehatan mental. Untuk mengatasi masalah ini, prediksi dini gangguan kesehatan mental menjadi langkah penting agar intervensi dapat dilakukan lebih awal. Penelitian ini membandingkan akurasi algoritma Naïve Bayes dan regresi logistik dalam memprediksi kesehatan mental mahasiswa di Provinsi Jambi. Data dikumpulkan dari 300 mahasiswa di tiga perguruan tinggi yang berbeda. Hasil penelitian menunjukkan bahwa Naïve Bayes memiliki akurasi sebesar 99,58% pada training set dan 100% pada testing set, sedangkan regresi logistik hanya mencapai 61,67% pada training set dan 63,33% pada testing set. Hasil ini menunjukkan bahwa Naïve Bayes lebih unggul dibandingkan regresi logistik dalam memprediksi kesehatan mental mahasiswa. Temuan ini dapat menjadi dasar pengembangan alat deteksi dini yang lebih efektif, sehingga institusi pendidikan dapat merancang strategi intervensi yang tepat guna mendukung kesejahteraan mahasiswa.

Review

The study titled "Perbandingan Algoritma Naïve Bayes Dan Regresi Logistik Untuk Memprediksi Kesehatan Mental Mahasiswa Di Provinsi Jambi" addresses a critically important and timely subject: the mental well-being of university students. The abstract effectively contextualizes the research by highlighting the global prevalence of student mental health issues, exacerbated by academic pressures and lifestyle changes, and provides concerning local statistics from Indonesia and Jambi. The stated aim to develop early prediction models using machine learning is highly relevant, aligning with the urgent need for proactive interventions to support student welfare. The authors' focus on a specific regional context, utilizing data from 300 students across three institutions, provides a valuable localized perspective. The research's methodology centers on a comparative analysis of two widely used classification algorithms, Naïve Bayes and Logistic Regression, for predicting student mental health. The results presented are quite pronounced: Naïve Bayes achieved remarkable accuracy scores of 99.58% on the training set and a perfect 100% on the testing set. In sharp contrast, Logistic Regression showed significantly lower performance, with accuracies of 61.67% on the training set and 63.33% on the testing set. These findings strongly support the authors' conclusion that Naïve Bayes is demonstrably superior to Logistic Regression for this specific predictive task, positioning it as a highly effective tool. The implications of these findings are significant, suggesting that a Naïve Bayes-based model could form the cornerstone of an effective early detection system for student mental health issues. Such a tool could empower educational institutions to design and implement timely and targeted intervention strategies, thereby fostering a more supportive environment for student welfare. However, the exceptionally high accuracy reported for Naïve Bayes, particularly the 100% on the testing set, warrants a more detailed discussion within the full paper regarding potential data characteristics, feature engineering, or the absence of noise, which could explain such perfect performance. Future research could further enhance this work by exploring the interpretability of the Naïve Bayes model, conducting external validation with diverse datasets, and considering the inclusion of other robust machine learning algorithms for broader comparative analysis and increased generalizability.

Full Text

Comments

You need to be logged in to post a comment.