Analysing the Effect of Metaheuristic Based Feature Selection Algorithms on the Classification of the Power Quality Events

Birsen Gümüş; Melih Çoban; Suleyman Sungur Tezcan

Makale detayları

Analysing the Effect of Metaheuristic Based Feature Selection Algorithms on the Classification of the Power Quality Events

Yazarlar:

Birsen Gümüş | Melih Çoban | Suleyman Sungur Tezcan

Yayıncı:

Kirikkale University

Ana Başlıklar:

Power Plants

Dergi:

International Journal of Engineering Research and Development Vol.16 Issue 2, pp.646-658

Belge Tipi:

Makale

Yayın Tarihi:

6/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001314842

pISSN:

1308-5506

Anahtar kelimeler:

Özet:

 In this study, two different optimization algorithms have been used for the feature selection stage, which plays a crucial role in the classification of power quality disturbances. In the first part of the study, signals containing power quality events were generated to initiate the classification process. Discrete wavelet transform DWT has been used for feature extraction. Two different datasets were obtained by normalizing and taking logarithm of the dataset obtained after the feature extraction process. The Equilibrium Optimizer EO and the Salp Swarm Optimization Algorithm SSA, which are named metaheuristic based feature selection algorithms, were used for the feature selection process. The K Nearest Neighbour Algorithm KNN is preferred for classification. The highest accuracy rate in classification was achieved at 96.05 when utilizing EO as the feature selection algorithm and using the logarithmic dataset. The lowest classification accuracy rate was obtained as 90.62 when the feature selection algorithm was SSA and the normalized data set was used. In the second part of the study, a histogram graph was created to identify the most frequently selected features from the first part. The highest accuracy rate obtained for the classification using the histogram graph was observed to be 95.8 and the lowest accuracy rate was 93.83.

 In this study, two different optimization algorithms have been used for the feature selection stage, which plays a crucial role in the classification of power quality disturbances. In the first part of the study, signals containing power quality events were generated to initiate the classification process. Discrete wavelet transform DWT has been used for feature extraction. Two different datasets were obtained by normalizing and taking logarithm of the dataset obtained after the feature extraction process. The Equilibrium Optimizer EO and the Salp Swarm Optimization Algorithm SSA, which are named metaheuristic based feature selection algorithms, were used for the feature selection process. The K Nearest Neighbour Algorithm KNN is preferred for classification. The highest accuracy rate in classification was achieved at 96.05 when utilizing EO as the feature selection algorithm and using the logarithmic dataset. The lowest classification accuracy rate was obtained as 90.62 when the feature selection algorithm was SSA and the normalized data set was used. In the second part of the study, a histogram graph was created to identify the most frequently selected features from the first part. The highest accuracy rate obtained for the classification using the histogram graph was observed to be 95.8 and the lowest accuracy rate was 93.83.

 In this study, two different optimization algorithms have been used for the feature selection stage, which plays a crucial role in the classification of power quality disturbances. In the first part of the study, signals containing power quality events were generated to initiate the classification process. Discrete wavelet transform DWT has been used for feature extraction. Two different datasets were obtained by normalizing and taking logarithm of the dataset obtained after the feature extraction process. The Equilibrium Optimizer EO and the Salp Swarm Optimization Algorithm SSA, which are named metaheuristic based feature selection algorithms, were used for the feature selection process. The K Nearest Neighbour Algorithm KNN is preferred for classification. The highest accuracy rate in classification was achieved at 96.05 when utilizing EO as the feature selection algorithm and using the logarithmic dataset. The lowest classification accuracy rate was obtained as 90.62 when the feature selection algorithm was SSA and the normalized data set was used. In the second part of the study, a histogram graph was created to identify the most frequently selected features from the first part. The highest accuracy rate obtained for the classification using the histogram graph was observed to be 95.8 and the lowest accuracy rate was 93.83.

 Bu çalışmada, güç kalitesi bozulmalarının sınıflandırmasında önemli bir role sahip olan özellik seçme aşaması için iki farklı optimizasyon algoritması kullanılmıştır. Çalışmanın birinci kısmında, sınıflandırma sürecinin başlaması için güç kalitesi olaylarını içeren sinyaller üretilmiştir. Özellik çıkarma için Ayrık Dalgacık Dönüşümü DWT kullanılmıştır. Özellik çıkarma işleminden sonra elde edilen veri seti, normalize edilerek ve logaritması alınarak iki farklı veri seti elde edilmiştir Özellik seçme işlemi için Denge Optimize Edici Algoritması EO ve Salp Sürü Optimizasyon Algoritması SSA olarak isimlendirilen metasezgisel tabanlı optimizasyon algoritmaları özellik seçme algoritmaları olarak kullanılmıştır. Sınıflandırma için K En Yakın Komşu Algoritması KNN tercih edilmiştir. En yüksek sınıflandırma doğruluk oranı, özellik seçme algoritması olarak EO ve veri seti olarak logaritmik veri setinin kullanıldığı durumda, 96.05 olarak elde edilmiştir. En düşük sınıflandırma doğruluk oranı ise özellik seçme algoritmasının SSA olduğu ve normalize veri setinin kullanıldığı durumda,  90.62 olarak elde edilmiştir. Çalışmanın ikinci kısmında ise, birinci kısımda en çok seçilen özellikleri tespit etmek için histogram grafiği oluşturulmuştur. En çok seçilen sekiz özellik ile sınıflandırma işlemi tekrarlanmıştır. Histogram grafiği kullanılarak yapılan sınıflandırmaya ilişkin elde edilen en yüksek doğruluk oranı  95.8 ve en düşük doğruluk oranı  93.83 olarak gözlenmiştir.

Makale detayları

Analysing the Effect of Metaheuristic Based Feature Selection Algorithms on the Classification of the Power Quality Events

Yazarlar:

Birsen Gümüş | Melih Çoban | Suleyman Sungur Tezcan

Yayıncı:

Kirikkale University

Ana Başlıklar:

Power Plants

Dergi:

International Journal of Engineering Research and Development Vol.16 Issue 2, pp.646-658

Belge Tipi:

Makale

Yayın Tarihi:

6/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001314842

pISSN:

1308-5506

Power quality events

Classification

Machine LearningGüç Kalitesi olayları

Sınıflandırma

Makine Öğrenmesi

Sayı

Makale detayları

Analysing the Effect of Metaheuristic Based Feature Selection Algorithms on the Classification of the Power Quality Events

Yazarlar: Birsen Gümüş | Melih Çoban | Suleyman Sungur Tezcan

Yayıncı: Kirikkale University

Ana Başlıklar: Power Plants

Dergi: International Journal of Engineering Research and Development Vol.16 Issue 2, pp.646-658

Belge Tipi: Makale

Yayın Tarihi: 6/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001314842

pISSN: 1308-5506

Power quality events

Classification

Machine LearningGüç Kalitesi olayları

Sınıflandırma

Makine Öğrenmesi

Sayı

Yazarlar:

Birsen Gümüş | Melih Çoban | Suleyman Sungur Tezcan

Yayıncı:

Kirikkale University

Ana Başlıklar:

Power Plants

Dergi:

International Journal of Engineering Research and Development Vol.16 Issue 2, pp.646-658

Belge Tipi:

Makale

Yayın Tarihi:

6/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001314842

pISSN:

1308-5506