Bu çalışmada, yükseköğretim kurumlarındaki öğrenci başarısının tahmin edilmesi amacıyla 9 farklı makine öğrenmesi algoritmasının performansları karşılaştırılmıştır. Araştırma kapsamında, öğrencilerin demografik özellikleri, akademik geçmişleri ve sosyo-ekonomik durumlarını içeren kapsamlı bir veri seti üzerinde XGBoost, Random Forest, Gradient Boosting gibi ensemble yöntemleri ile geleneksel algoritmalar analiz edilmiştir. Veri ön işleme aşamasında eksik veriler median değerler ve kategorik kodlama teknikleriyle işlenmiş, özellik önem analiziyle kritik faktörler belirlenmiştir. Beş katlı çapraz doğrulama sonuçlarına göre XGBoost algoritması %90.1 test doğruluğuyla en yüksek performansı gösterirken, özellik önem analizi "önceki yeterlilik notu" ve "kabul puanı"nın en belirleyici faktörler olduğunu ortaya koymuştur. Bulgular, eğitim kurumlarının erken uyarı sistemleri geliştirirken ensemble yöntemlerinin kullanımının etkililiğini kanıtlamaktadır.
Öğrenci başarı tahmini makine öğrenmesi XGBoost eğitim veri madenciliği karşılaştırmalı algoritma analizi
This study presents a comparative performance evaluation of ten machine learning algorithms for predicting student academic success in higher education institutions. Using a comprehensive dataset encompassing demographic characteristics, academic history, and socio-economic status, we analyzed ensemble methods (XGBoost, Random Forest, Gradient Boosting) alongside traditional algorithms. During data preprocessing, missing values were handled through median imputation and categorical encoding techniques, while feature importance analysis identified critical predictive factors. Five-fold cross-validation results demonstrated that the XGBoost algorithm achieved superior performance with 90.1% test accuracy, with feature importance analysis revealing "previous qualification grade" and "admission score" as the most determinant factors. The findings substantiate the effectiveness of ensemble methods in developing early warning systems for educational institutions.
Student success prediction machine learning XGBoost educational data mining comparative algorithm analysis.
Birincil Dil | Türkçe |
---|---|
Konular | Öğretim Teknolojileri |
Bölüm | Eğitim Bilimleri |
Yazarlar | |
Yayımlanma Tarihi | 30 Nisan 2025 |
Gönderilme Tarihi | 29 Mart 2025 |
Kabul Tarihi | 15 Nisan 2025 |
Yayımlandığı Sayı | Yıl 2025 Cilt: 9 Sayı: 20 |
The Aim of The Journal
The Journal of Interdisciplinary Educational Researches (JIER) published by the Interdisciplinary Educational and Research Association (JIER)A) is an internationally eminent journal.
JIER, a nonprofit, nonprofit NGO, is concerned with improving the education system within the context of its corporate objectives and social responsibility policies. JIER, has the potential to solve educational problems and has a strong gratification for the contributions of qualified scientific researchers.
JIER has the purpose of serving the construction of an education system that can win the knowledge and skills that each individual should have firstly in our country and then in the world. In addition, JIER serves to disseminate the academic work that contributes to the professional development of teachers and academicians, offering concrete solutions to the problems of all levels of education, from preschool education to higher education.
JIER has the priority to contribute to more qualified school practices. Creating and managing content within this context will help to advance towards the goal of being a "focus magazine" and "magazine school", and will also form the basis for a holistic view of educational issues. It also acts as an intermediary in the production of common mind for sustainable development and education