Comparıson Of Tradıtıonal Machıne Learnıng Methods And Metaheurıstıc Methods For Attrıbute Selectıon

İsmail Açar; İbrahim Berkan Aydilek

Article Details

Comparıson Of Tradıtıonal Machıne Learnıng Methods And Metaheurıstıc Methods For Attrıbute Selectıon

Authors:

İsmail Açar | İbrahim Berkan Aydilek

Publisher:

Dicle University

Main Topics:

Makine Öğrenme (Diğer)

Journal:

Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi Vol.15 Issue 2, pp.397-409

Document Type:

Article

Publication Date:

6/2024

Document Language:

Turkish

Document ID (Org No.):

2001320917

pISSN:

1309-8640

Keywords:

Abstract:

 Feature selection methods are very important for machine learning. As the size of the dataset increases, the performance of the machine learning model decreases. The feature selection method can also be considered as an optimization process. Success can be improved by eliminating noisy or irrelevant attributes for the model. In this study, 3 different datasets from the UCI repository were used and 5 different machine learning algorithms were used to measure the performance and runtime of the algorithms using 10-fold cross-validation. As feature selection methods, traditional machine learning feature selection methods and meta-heuristic algorithms were used for feature selection. The feature selection methods used are as follows; Correlation Analysis, Lasso, Hypothesis testing and metaheuristic algorithms Genetic Algorithm and Particle Swarm Optimization Algorithm. As a result of the experiments, it is seen that the performance increases with the application of feature selection methods. It was observed that feature selection with metaheuristic algorithms is more suitable in terms of performance and runtime.

 Feature selection methods are very important for machine learning. As the size of the dataset increases, the performance of the machine learning model decreases. The feature selection method can also be considered as an optimization process. Success can be improved by eliminating noisy or irrelevant attributes for the model. In this study, 3 different datasets from the UCI repository were used and 5 different machine learning algorithms were used to measure the performance and runtime of the algorithms using 10-fold cross-validation. As feature selection methods, traditional machine learning feature selection methods and meta-heuristic algorithms were used for feature selection. The feature selection methods used are as follows; Correlation Analysis, Lasso, Hypothesis testing and metaheuristic algorithms Genetic Algorithm and Particle Swarm Optimization Algorithm. As a result of the experiments, it is seen that the performance increases with the application of feature selection methods. It was observed that feature selection with metaheuristic algorithms is more suitable in terms of performance and runtime.

 Feature selection methods are very important for machine learning. As the size of the dataset increases, the performance of the machine learning model decreases. The feature selection method can also be considered as an optimization process. Success can be improved by eliminating noisy or irrelevant attributes for the model. In this study, 3 different datasets from the UCI repository were used and 5 different machine learning algorithms were used to measure the performance and runtime of the algorithms using 10-fold cross-validation. As feature selection methods, traditional machine learning feature selection methods and meta-heuristic algorithms were used for feature selection. The feature selection methods used are as follows; Correlation Analysis, Lasso, Hypothesis testing and metaheuristic algorithms Genetic Algorithm and Particle Swarm Optimization Algorithm. As a result of the experiments, it is seen that the performance increases with the application of feature selection methods. It was observed that feature selection with metaheuristic algorithms is more suitable in terms of performance and runtime.

 Öznitelik seçim yöntemleri, makine öğrenmesi için çok önemli bir yere sahiptir. Veri kümesinin boyutu arttıkça makine öğrenmesi modelinin performansı düşmektedir. Öznitelik seçim yöntemi aynı zamanda bir optimizasyon süreci olarak düşünülebilmektedir. Gürültülü olan ya da model için alakasız olan öznitelikler elenince başarı artabilmektedir. Bu çalışma kapsamında UCI veri deposundan 3 farklı veri kümesi kullanılmış ve 5 farklı makine öğrenmesi algoritması kullanılarak 10 katlı çapraz doğrulama yöntemiyle algoritmaların başarımları ve çalışma süreleri ölçülmüştür. Öznitelik seçim yöntemi olarak geleneksel makine öğrenmesi ile öznitelik seçim yöntemleri ve meta sezgisel algoritmalar kullanılarak öznitelik seçim işlemleri gerçekleştirilmiştir. Kullanılan öznitelik seçim yöntemleri şu şekildedir; Korelasyon Analizi, Lasso, Hipotez testi ve metasezgisel algoritmalardan Genetik Algoritma ve Parçacık Sürü Optimizasyon Algoritması yöntemleri kullanılmıştır. Yapılan deneyler sonucunda öznitelik seçim yöntemlerinin uygulanması ile birlikte başarımın arttığı görülmektedir. Metasezgisel algoritmalarla yapılan öznitelik seçiminin başarım ve çalışma süresi bakımından daha uygun olduğu gözlemlenmiştir.

Article Details

Comparıson Of Tradıtıonal Machıne Learnıng Methods And Metaheurıstıc Methods For Attrıbute Selectıon

Authors:

İsmail Açar | İbrahim Berkan Aydilek

Publisher:

Dicle University

Main Topics:

Makine Öğrenme (Diğer)

Journal:

Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi Vol.15 Issue 2, pp.397-409

Document Type:

Article

Publication Date:

6/2024

Document Language:

Turkish

Document ID (Org No.):

2001320917

pISSN:

1309-8640

Feature selection

Metaheuristic Algorithms

Machine Learning

Genetic Algorithms

Particle Swarm Optimization AlgorithmÖznitelik seçimi

metasezgisel algoritmalar

makine öğrenmesi

genetik algoritmalar

parçacık sürü optimizasyon algoritması

Issues

Article Details

Authors: İsmail Açar | İbrahim Berkan Aydilek

Publisher: Dicle University

Main Topics: Makine Öğrenme (Diğer)

Journal: Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi Vol.15 Issue 2, pp.397-409

Document Type: Article

Publication Date: 6/2024

Document Language: Turkish

Document ID (Org No.): 2001320917

pISSN: 1309-8640

Issues

Authors:

İsmail Açar | İbrahim Berkan Aydilek

Publisher:

Dicle University

Main Topics:

Makine Öğrenme (Diğer)

Journal:

Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi Vol.15 Issue 2, pp.397-409

Document Type:

Article

Publication Date:

6/2024

Document Language:

Turkish

Document ID (Org No.):

2001320917

pISSN:

1309-8640