Makine öğrenmesi yöntemleri ile ikinci el araçlarda fiyat tahmini
Küçük Resim Yok
Tarih
2023
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
İstanbul Ticaret Üniversitesi
Erişim Hakkı
info:eu-repo/semantics/openAccess
Özet
Son yıllarda ikinci el araçların pazar hacmi artmıştır. Bu pazarda satıcı ve alıcı için dog?ru fiyatlandırma oldukça o?nemlidir. Son kullanıcı veya kurumlar için ikinci el araç deg?erlemesine ya da kiralanmasında yardımcı olacak sistemsel bir yapıya ihtiyaç bulunmaktadır. Kurulacak bu yapıyla hem araç alacak olan hem de aracını satacak olan kullanıcı, kilometresi, kaza kaydı, kasa tipi, benzin tipi, ağır hasar kaydı gibi fiyata etki edecek özelliklerine göre fiyat çıkarabilir. Bu çalışmada ilgili ikinci el araç ilanların yer aldıg?ı sitelerden Selenium test aracı ile 13.000 u?zerinde veri toplanmış ve bu veriler u?zerinde veri o?nişleme (temizleme, do?nu?ştu?rme vs.) adımları uygulanmıştır. Makine o?g?renme teknikleri KNIME Analytics Platform veri madencilig?i programının 4.7.3 su?ru?mu?nde uygulanarak ikinci el araç fiyatı tahmin edilmeye çalışılmış ve sonuçlar birbiriyle karşılaştırılmıştır. Performans o?lçu?lu?rken R² kullanılmıştır. Dog?rusal (Linear) regresyon 0,71 R², Rastgele Orman (Random Forest) 0,85 R², Gradyan artırıcı (GBoosted) 0,83 R², ag?aç toplulug?u (Tree Ensemble) 0,85 R² ve yapay sinir ağları (Artificial Neural Network) 0,72 R² tahmin oranları elde edilmiştir. Elde edilen sonuçlarda en düşük tahmin başarısı doğrusal (linear) regresyon, en yüksek başarılı tahmin algoritması 0,85 R² ile rastgele orman algoritması olmuştur.
The market volume of second-hand vehicles has increased in recent years. In this market, accurate pricing is very important for the seller and the buyer. There is a need for a systematic structure that will help end-users or institutions in the valuation or leasing of second-hand vehicles. With this structure to be established, both the user who will buy a vehicle and the user who will sell his vehicle can make a price according to the features that will affect the price such as mileage, accident record, body type, gasoline type, heavy damage record. In this study, over 13,000 data were collected from the relevant used car classifieds websites with the Selenium test tool and data preprocessing (cleaning, transformation, etc.) steps were applied on this data. Machine learning techniques were applied in version 4.7.3 of the KNIME Analytics Platform data mining program to predict the used car price and the results were compared with each other. R² was used to measure performance. Linear regression 0.71 R², Random Forest 0.85 R², Gradient Boosted 0.83 R², Tree Ensemble 0.85 R² and Artificial Neural Network 0.72 R² prediction rates were obtained. In the results obtained, the lowest prediction success was linear regression and the highest successful prediction algorithm was the random forest algorithm with 0.85 R².
The market volume of second-hand vehicles has increased in recent years. In this market, accurate pricing is very important for the seller and the buyer. There is a need for a systematic structure that will help end-users or institutions in the valuation or leasing of second-hand vehicles. With this structure to be established, both the user who will buy a vehicle and the user who will sell his vehicle can make a price according to the features that will affect the price such as mileage, accident record, body type, gasoline type, heavy damage record. In this study, over 13,000 data were collected from the relevant used car classifieds websites with the Selenium test tool and data preprocessing (cleaning, transformation, etc.) steps were applied on this data. Machine learning techniques were applied in version 4.7.3 of the KNIME Analytics Platform data mining program to predict the used car price and the results were compared with each other. R² was used to measure performance. Linear regression 0.71 R², Random Forest 0.85 R², Gradient Boosted 0.83 R², Tree Ensemble 0.85 R² and Artificial Neural Network 0.72 R² prediction rates were obtained. In the results obtained, the lowest prediction success was linear regression and the highest successful prediction algorithm was the random forest algorithm with 0.85 R².
Açıklama
Fen Bilimleri Enstitüsü, Bilgisayar Mühendisliği Ana Bilim Dalı, Bilgisayar Mühendisliği Bilim Dalı
Anahtar Kelimeler
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control
