Atıf Sayısı
1
Görüntüleme
20
İndirme
2
Veri madenciliği algoritmalarının kullanımına hazır hale getirilmesi için, “Kaggle’da Veri Bilimi Anketi” isimli veri seti, veri madenciliği problem tanıma aşaması ile analiz edilmiştir. Analiz edilen veri seti ve karar verilen hedef doğrultusunda veri ön işleme aşaması da gerçekleştirilmiştir. Ön işleme aşaması çıktısı olarak elde edilen işlenmiş veri seti, veri madenciliği teknolojisi sınıflandırma yöntemine ait C4.5, Rastgele Orman ve K-En Yakın Komşu Algoritmaları ile modellenmiştir. Bu üç algoritmanın model başarı oranları hesaplanmıştır. Aralarındaki başarı oranı sapma değerleri analiz edilerek sapma değerlerine sebep olan durumlara değinilmiştir. Model başarı oranını etkileyen durumlara farklı bir açıdan daha bakılabilmesi için, bu üç algoritma ile yeni bir modelleme daha gerçekleştirilmiştir. İşlenmiş veri seti için karar verilen üç algoritma ve bu algoritmaların modelleme kriterleri ile gerçekleştirilen modelleme işlemi, orijinal veri seti kullanılarak da gerçekleştirilmiştir. İşlenmiş veri seti kullanılarak elde edilen modellerin başarı oranı hesaplamaları, orijinal veri seti için de hesaplanmış olup kıyaslamaları gerçekleştirilmiştir. Kullanılan veri seti, karar verilen yöntem, algoritma, algoritma kriter değerleri gibi model başarısını etkileyen etmenlerin kıyaslama işlemleri, gerçekleştirilen modelleme uygulamaları sayesinde somutlaştırılarak aktarılmıştır. Elde edilen bu kıyaslamalı örnekler referans alınarak, model başarı oranını etkileyen etmenler değerlendirilmiş olup, veri seti nitelik analizi ve Veri Madenciliği süreçleri hakkında sırasıyla çıkarımlar gerçekleştirilmiştir.
To prepare for the use of data mining algorithms, the data set called "Data Science Survey in Kaggle" was analyzed by the data mining problem recognition stage. The analyzed set of data and the data pre-processing stage according to the determined goal have also been carried out. The processed data set, obtained as a pre-processing phase output, is modeled with the data mining technology classification method C4.5, Rastgele Forest and K-The Nearest Neighbor's Algorithms. The success rates of these three algorithms are calculated. The success rate between them is analyzed by the deviation values and the circumstances that cause deviation values. In order to be able to look more from a different perspective to the situations affecting the success rate of the model, a new modeling has been implemented with these three algorithms. The three algorithms determined for the processed data set and the modeling process performed with the modeling criteria of these algorithms was also performed using the original data set. Calculations of the success rate of the models obtained using the processed data set were also calculated for the original data set and comparisons were made. The comparison processes of the factors affecting the success of the model, such as the data set used, the method determined, the algorithm, the algorithm criteria values, are transmitted through the practices of modeling performed. Referring to these comparative examples, the factors affecting the model success rate have been evaluated, and results on the quality analysis of the data set and the processes of data mining have been made accordingly.
In order to make data mining algorithms ready for use, the data set named "Data Science Questionnaire in Kaggle", was analyzed in the data mining problem recognition stage. In line with the analyzed data set and the decided target, the data preprocessing stage was also carried out. The processed data set obtained as the output of the pre-processing stage is modeled by C4.5, Random Forest and K-Nearest Neighbor classification algorithms of data mining technology. The model success rates of these three algorithms were calculated. The success rate deviation values between them were analyzed and the situations that caused the deviation values were mentioned. In order to look at the issues affecting the model success rate from a different perspective, new modeling was performed with these three algorithms. The modeling process, which was carried out with the three algorithms decided for the processed data set and the modeling criteria of these algorithms, was also carried out using the original data set. The success rate calculations of the models obtained using the processed data set were also calculated for the original data set and their comparisons were made. The comparison of the factors affecting the success of the model, such as the data set used, the method decided, algorithm, algorithm criterion values, were embodied and expressed thanks to the modeling applications carried out. By taking these comparative examples as a reference, the factors affecting the model success rate were evaluated, and inferences were made about the data set quality analysis and data mining processes, respectively.
Alan : Fen Bilimleri ve Matematik; Mühendislik
Dergi Türü : Uluslararası
Makale Detay 
Atıf Yapanlar
Benzer Makaleler
PDF Görüntüle
Dergi Bilgisi
Eseri Dinleyin
Bu Sayfayı Yazdırın
Paylaş
Mendeley
