E-NoseFlavNet: Sensör Dizisi Teknolojisi ile Çeşitli Makine Öğrenmesi Modelleriyle Güçlendirilmiş E Burun Tabanlı Aroma Lezzet Analizine Doğru

Yıl 2026, Cilt: 8 Sayı: 1 , 67 - 76 , 30.04.2026

İlyas Özer , Onursal Çetin , Kutlucan Görür , Feyzullah Temurtaş

https://doi.org/10.46387/bjesr.1854260

https://izlik.org/JA78DN79JP

Öz

Gıda güvenliği için güvenilir aroma tespiti gereklidir, ancak geleneksel yöntemler (test kağıdı, kromatografi) doğruluk, hız, taşınabilirlik ve maliyet açısından yetersizdir. Bu çalışmada, aroma VOC analizi için istatistiksel yöntemler (t-SNE, MANOVA) ve makine öğrenmesi modellerini birleştiren sensör dizili bir E-burun sistemi önerilmiştir. Çikolata, karanfil, tarçın, zencefil ve aromasız olmak üzere beş sınıf incelenmiştir. Literatürdeki %95.73 doğruluk referansı ile Random Forest, ExtraTrees, XGBoost, CatBoost ve Stacking ensemble modelleri değerlendirilmiştir. Tüm modeller ikili testlerde çikolatayı mükemmel şekilde tespit etmiştir. Stacking aromasız sınıfında %50, diğer aromalarda %94'e kadar başarı göstermiştir. İstatistiksel yöntemler özellikle ikili testlerde belirgin VOC ayrımı sağlamıştır.

Anahtar Kelimeler

Aroma Lezzet Analiz , E Burun , Makine Öğrenimi Modelleri , İstatistiksel Ayrıştırma

E-NoseFlavNet: Towards E-Nose based Aroma Flavour Analysis Empowered by Diverse ML Models via Sensor Array Technology

https://izlik.org/JA78DN79JP

Öz

Ensuring food-supply safety requires reliable aroma/flavour detection, yet conventional tools (test paper, cyclotron, chromatography) often lack accuracy, broad detection range, speed, portability, and low cost. We propose a sensor array electronic nose (E nose) for aroma VOC analysis and prediction, combining statistical visualization/discrimination (t SNE, MANOVA) with machine learning models. Five classes were studied: chocolate, clove, cinnamon, ginger, and an unflavoured VOC set. Using a literature benchmark of 95.73% accuracy, we evaluated Random Forest, ExtraTrees, XGBoost, CatBoost, and a Stacking ensemble for 5 class prediction and one vs rest tests. All models perfectly identified chocolate in one vs rest prediction. Stacking performed poorly for unflavoured vs others (50%), whereas other aromas reached up to 94%. Statistical methods showed clear VOC separation, especially in one vs rest analyses.

Anahtar Kelimeler

Aroma Flavour Analysis , E-Nose , Machine Learning Models , Statistical Discrimination

Kaynakça

• Y. Durmuş and A.F. Atasoy, "Application of multivariate machine learning methods to investigate organic compound content of different pepper spices," Food Biosci., vol. 51, art. no. 102216, Jan. 2023.
• X. Yang, M. Li, X. Ji, et al., "Recognition algorithms in E-Nose: A review," IEEE Sens. J., vol. 23, no. 18, pp. 20460–20472, Sep. 2023.
• A. Ren, A. Zahid, A. Zoha, et al., "Machine learning driven approach towards the quality assessment of fresh fruits using non-invasive sensing," IEEE Sens. J., vol. 20, no. 4, pp. 2075–2083, Feb. 2020.
• M. Pardo and G. Sberveglieri, "Coffee analysis with an electronic nose," IEEE Trans. Instrum. Meas., vol. 51, no. 6, pp. 1334–1339, Dec. 2002.
• H. Wang, Y. Sui, J. Liu, et al., "Analysis and comparison of the quality and flavour of traditional and conventional dry sausages collected from northeast China," Food Chem. X, vol. 20, art. no. 100979, Dec. 2023.
• S. Wang, Q. Zhang, C. Liu, et al., "Synergetic application of an E-tongue, E-nose and E-eye combined with CNN models and an attention mechanism to detect the origin of black pepper," Sens. Actuators A, Phys., vol. 357, art. no. 114417, Aug. 2023.
• A. Flammini, D. Marioli, and A. Taroni, "A low-cost interface to high-value resistive sensors varying over a wide range," IEEE Trans. Instrum. Meas., vol. 53, no. 4, pp. 1052–1056, Aug. 2004.
• K. Brudzewski, S. Osowski, and A. Dwulit, "Recognition of coffee using differential electronic nose," IEEE Trans. Instrum. Meas., vol. 61, no. 6, pp. 1803–1810, Jun. 2012.
• F. Herrero-Carrón, D. J. Yáñez, F. de B. Rodríguez, and P. Varona, "An active, inverse temperature modulation strategy for single sensor odorant classification," Sens. Actuators B, Chem., vol. 206, pp. 555–563, Jan. 2015.
• R. Gosangi and R. Gutierrez-Osuna, "Active temperature programming for metal-oxide chemoresistors," IEEE Sens. J., vol. 10, no. 6, pp. 1075–1082, Jun. 2010.
• X. Yin, L. Zhang, F. Tian, and D. Zhang, "Temperature modulated gas sensing E-nose system for low-cost and fast detection," IEEE Sens. J., vol. 16, no. 2, pp. 464–474, Jan. 2016.
• M. Anly Antony and R. S. Kumar, "A comparative study on predicting food quality using machine learning techniques," in Proc. 7th Int. Conf. Adv. Comput. Commun. Syst. (ICACCS), Coimbatore, India, Mar. 2021, pp. 1771–1776.
• "UDOO Bolt Gear Mini PC." [Online]. Available: https://www.wdlsystems.com/UDOO-Bolt-Gear-Mini-PC
• "Aroma samples." [Online]. Available: https://www.hammaddeler.com/urun/kurabiye-aromalari-seti-12-parca
• MQ-series gas sensor." [Online]. Available: https://robu.in/mq-series-gas-sensor/
• S. Chehreh Chelgani, H. Nasiri, A. Tohry, and H. R. Heidari, "Modeling industrial hydrocyclone operational variables by SHAP-CatBoost - A 'conscious lab' approach," Powder Technol., vol. 420, art. no. 118416, Apr. 2023.
• U. Saeed, S. U. Jan, Y.-D. Lee, and I. Koo, "Fault diagnosis based on extremely randomized trees in wireless sensor networks," Rel. Eng. Syst. Safety, vol. 205, art. no. 107284, Jan. 2021.
• T. Chen and C. Guestrin, "XGBoost: A scalable tree boosting system," in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining (KDD), San Francisco, CA, USA, Aug. 2016, pp. 785–794.
• D. H. Wolpert, "Stacked generalization," Neural Netw., vol. 5, no. 2, pp. 241–259, 1992.
• I. Ozer, C. K. Ozer, A. C. Karaca, et al., "Species-level microfossil identification for globotruncana genus using hybrid deep learning algorithms from the scratch via a low-cost light microscope imaging," Multimed. Tools Appl., vol. 82, no. 10, pp. 13689–13718, Apr. 2023.