Pamukkale Univ Muh Bilim Derg. Baskýdaki Makaleler: PAJES-68623 | DOI: 10.5505/pajes.2024.68623

Bulanýk mantýk yöntemi ve yapay sinir aðlarý kullanýlarak farklý kokularýn sýnýflandýrýlmasý

Ümit Özsandýkcýoðlu, Bilal Talha Ayvaz, Ayten Atasoy
Karadeniz Teknik Üniversitesi Mühendislik Fakültesi, Elektrik Elektronik Mühendisliði Anabilim Dalý, Trabzon

Bu çalýþmada 8 adet metal oksit gaz sensörü kullanýlarak gerçekleþtirilen bir elektronik burun devresi kullanýlarak 9 farklý maddenin (yumurta, çürük yumurta, nane, naftalin, limon, melek otu kökü, aseton, oje ve gül suyu) ayýrt edilmesi saðlanmýþtýr. Çalýþmada veri sýnýflandýrma için bulanýk mantýk yöntemi ve yapay sinir aðlarý kullanýlmýþtýr. Yapay sinir aðlarý kullanýlarak gerçekleþtirilen veri sýnýflandýrma iþleminde farklý að mimarileri kullanýlarak sýnýflandýrma performanslarý incelenmiþtir. Bulanýk mantýk yöntemiyle gerçekleþtirilen sýnýflandýrma iþleminde ise kullanýlan farklý üyelik fonksiyonlarý ile sýnýflandýrma doðruluðu artýrýlmaya çalýþýlmýþtýr. Yapay sinir aðlarý ile yapýlan sýnýflandýrma iþleminde gizli katman ve çýktý katmanýnda sýrasýyla logaritmik sigmoid ve hiperbolik tanjant fonksiyonlarý kullanýlarak %96,41 sýnýflandýrma doðruluðu elde edilmiþtir. En yüksek sýnýflandýrma doðruluðunun elde edildiði bu yapay sinir aðýnýn gizli katmanýnda 8 adet yapay sinir hücresi kullanýlmýþtýr. Bulanýk mantýk yöntemi ile yapýlan sýnýflandýrma iþleminde ise çan üyelik fonksiyonunun kullanýlmasýyla sýnýflandýrma doðruluðu %95,55 olarak elde edilmiþtir.

Anahtar Kelimeler: Elektronik burun, veri boyutu indirgeme, yapay sinir aðlarý, bulanýk mantýk yöntemi, veri sýnýflandýrma

---

Classification of different odors using fuzzy logic method and artificial neural network

Ümit Özsandýkcýoðlu, Bilal Talha Ayvaz, Ayten Atasoy
Department Of Electrical And Electronics Engineering, Karadeniz Technical University, Trabzon, Turkey

In this study, 9 different substances (egg, rotten egg, mint, naphthalene, lemon, angelica root, acetone, nail polish and rose water) were distinguished by using an electronic nose using 8 metal oxide gas sensors. In the study, fuzzy logic method and artificial neural networks were used for data classification. In the data classification process performed using artificial neural networks, the classification performances were examined using different network architectures. In the classification process carried out with the fuzzy logic method, an attempt was made to increase the classification accuracy with the different membership functions used. In the classification process performed with artificial neural networks, 96.41% classification accuracy was achieved by using logarithmic sigmoid and hyperbolic tangent functions in the hidden layer and output layer, respectively. 8 artificial nerve cells were used in the hidden layer of this artificial neural network, where the highest accuracy value was achieved. In the classification process carried out with the fuzzy logic method, the classification accuracy was achieved as 95.55% by using the bell membership function.

Keywords: Electronic nose, data dimension reduction, artificial neural network, fuzzy logic method, data classification

---

ATIF KOPYALA