MODELLING OF THE EFFECTS ON EDUCATIONAL SUCCESS BY MACHINE LEARNING ALGORITHMS

Zeynep Bakan; Filiz Kanbay

Makale detayları

MODELLING OF THE EFFECTS ON EDUCATIONAL SUCCESS BY MACHINE LEARNING ALGORITHMS

Yazarlar:

Zeynep Bakan | Filiz Kanbay

Yayıncı:

Istanbul Ticaret University

Ana Başlıklar:

Derin Öğrenme

Dergi:

İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi Vol.23 Issue 45, pp.27-41

Belge Tipi:

Makale

Yayın Tarihi:

6/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001318565

pISSN:

1305-7820

Anahtar kelimeler:

Özet:

 Machine learning, which is used for classification, clustering and prediction in the fields of health, media, banking and finance, is also used in the field of education today. In this study, by using the mathematical models established with machine learning classification methods such as K-nearest neighbour, naive bayes, random forest, support vector machines, decision trees and boosting; the factors affecting students academic success were investigated to guide educational institution the strategies , to determine the measures to be taken, and even to benefit the generalization of the results by applying them to a larger population, different types of schools or at different levels, in different sectors. Maximum model success was achieved by determining the hyperparameters that affected the success of the established mathematical model by the grid scanning method. In mathematical modelling, the academic success criterion is determined as the output; The changes in the model success of the output and input numbers in the established mathematical models were examined; The effects of reducing the number of outputs and inputs by various methods supervised and unsupervised methods on the success of the mathematical model have been observed. Finally the best accuracy scores were obtained from the data set with two class labels. The accuracy scores of the algorithms K-nearest neighbour, naive bayes, random forest, support vector machines, decision trees and boosting respectively were 0,62, 0,61, 0,96, 0,72, 0,86, 0,79.

 Machine learning, which is used for classification, clustering and prediction in the fields of health, media, banking and finance, is also used in the field of education today. In this study, by using the mathematical models established with machine learning classification methods such as K-nearest neighbour, naive bayes, random forest, support vector machines, decision trees and boosting; the factors affecting students academic success were investigated to guide educational institution the strategies , to determine the measures to be taken, and even to benefit the generalization of the results by applying them to a larger population, different types of schools or at different levels, in different sectors. Maximum model success was achieved by determining the hyperparameters that affected the success of the established mathematical model by the grid scanning method. In mathematical modelling, the academic success criterion is determined as the output; The changes in the model success of the output and input numbers in the established mathematical models were examined; The effects of reducing the number of outputs and inputs by various methods supervised and unsupervised methods on the success of the mathematical model have been observed. Finally the best accuracy scores were obtained from the data set with two class labels. The accuracy scores of the algorithms K-nearest neighbour, naive bayes, random forest, support vector machines, decision trees and boosting respectively were 0,62, 0,61, 0,96, 0,72, 0,86, 0,79.

 Machine learning, which is used for classification, clustering and prediction in the fields of health, media, banking and finance, is also used in the field of education today. In this study, by using the mathematical models established with machine learning classification methods such as K-nearest neighbour, naive bayes, random forest, support vector machines, decision trees and boosting; the factors affecting students academic success were investigated to guide educational institution the strategies , to determine the measures to be taken, and even to benefit the generalization of the results by applying them to a larger population, different types of schools or at different levels, in different sectors. Maximum model success was achieved by determining the hyperparameters that affected the success of the established mathematical model by the grid scanning method. In mathematical modelling, the academic success criterion is determined as the output; The changes in the model success of the output and input numbers in the established mathematical models were examined; The effects of reducing the number of outputs and inputs by various methods supervised and unsupervised methods on the success of the mathematical model have been observed. Finally the best accuracy scores were obtained from the data set with two class labels. The accuracy scores of the algorithms K-nearest neighbour, naive bayes, random forest, support vector machines, decision trees and boosting respectively were 0,62, 0,61, 0,96, 0,72, 0,86, 0,79.

 Sağlık, medya, bankacılık ve finans alanında sınıflandırma, kümeleme ve tahmin amacıyla kullanılan makine öğrenmesi günümüzde eğitim alanında da kullanılmaktadır. Bu çalışmada eğitim öğretim kurumlarının belirleyecekleri stratejilerde veya alacakları önlemlerde yol gösterici olması ve hatta daha büyük ana kütle, daha farklı okul türü ya da farklı kademelerde, farklı sektörlerde uygulanarak sonuçların genelleştirilmesine fayda sağlaması amacıyla makine öğrenmesi yöntemlerinden K-en yakın komşu, naive bayes, rastgele orman, destek vektör makineleri, karar ağaçları, boosting makine öğrenmesi sınıflandırma algoritmaları ile kurulan matematiksel modellemeler ile öğrencilerin akademik başarılarını etkileyen faktörler araştırılmıştır. Kurulan matematiksel modelin başarısına etki eden hiperparametreler ızgara taraması yöntemi ile belirlenerek maksimum model başarısı sağlanmıştır. Matematiksel modellemelerde akademik başarı ölçütü çıktı olarak belirlenerek; kurulan matematiksel modellerde çıktı ve girdi sayılarına ait model başarılarının değişimi incelenmiş; çıktıların ve girdilerin sayısının çeşitli yöntemlerle denetimli ve denetimsiz yöntemlerle azaltılması işlemlerinin matematiksel model başarısına etkileri gözlenmiştir. Sonuç olarak, en yüksek model başarılarının iki sınıf etiketli veri setine ait olduğu görülmüştür. K-en yakın komşu, naive bayes, rastgele orman, destek vektör makineleri, karar ağaçları, boosting model başarıları sırasıyla 0,62, 0,61, 0,96, 0,72, 0,86, 0,79 olarak elde edilmiştir.

Makale detayları

MODELLING OF THE EFFECTS ON EDUCATIONAL SUCCESS BY MACHINE LEARNING ALGORITHMS

Yazarlar:

Zeynep Bakan | Filiz Kanbay

Yayıncı:

Istanbul Ticaret University

Ana Başlıklar:

Derin Öğrenme

Dergi:

İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi Vol.23 Issue 45, pp.27-41

Belge Tipi:

Makale

Yayın Tarihi:

6/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001318565

pISSN:

1305-7820

Machine learning

k-nearest neighbors

random forest

naive bayes

support vector machines

education-trainingMakine öğrenmesi

k-en yakın komşu

rastgele orman

naive bayes

destek vektör makineleri

eğitim-öğretim

Sayı

Makale detayları

Yazarlar: Zeynep Bakan | Filiz Kanbay

Yayıncı: Istanbul Ticaret University

Ana Başlıklar: Derin Öğrenme

Dergi: İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi Vol.23 Issue 45, pp.27-41

Belge Tipi: Makale

Yayın Tarihi: 6/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001318565

pISSN: 1305-7820

Sayı

Yazarlar:

Zeynep Bakan | Filiz Kanbay

Yayıncı:

Istanbul Ticaret University

Ana Başlıklar:

Derin Öğrenme

Dergi:

İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi Vol.23 Issue 45, pp.27-41

Belge Tipi:

Makale

Yayın Tarihi:

6/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001318565

pISSN:

1305-7820