Detection of Volatile Organic Compounds by Plasma Jet Optical Emission Spectroscopy With Machine Learning Methods in Cloud Technology
IEEE Transactions on Plasma Science, cilt.54, sa.5, ss.2101-2114, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 54 Sayı: 5
- Basım Tarihi: 2026
- Doi Numarası: 10.1109/tps.2026.3674343
- Dergi Adı: IEEE Transactions on Plasma Science
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Compendex, INSPEC, Natural Science Collection (ProQuest), Earth, Atmospheric, & Aquatic Science Collection (ProQuest), Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Sayfa Sayıları: ss.2101-2114
- Anahtar Kelimeler: Cloud technology, machine learning, plasma jet, volatile organic compounds (VOCs)
- Isparta Uygulamalı Bilimler Üniversitesi Adresli: Evet
Özet
Optical emission spectroscopy (OES), a well-established technique for plasma characterization, can be effectively integrated with machine learning methods to distinguish disease-marker gases. In this study, several machine learning algorithms were employed to classify volatile organic compounds (VOCs) using spectral data obtained from a plasma jet OES system. Acetone and methanol—VOCs of particular relevance in breath-based diagnostics for diabetes and cancer—were selected as target analytes. A total of 3047 spectral data points were analyzed using four classification algorithms: random forest (RF), Bagging, AdaBoost, and artificial neural networks (ANNs). Model performance was evaluated using standard metrics, including accuracy, precision, recall, F1 score, area under the curve (AUC), and receiver operating characteristic (ROC) curves. Among the evaluated models, the Bagging algorithm demonstrated superior performance, achieving an accuracy of 99%. High classification accuracy was attained by utilizing the full spectral range (200–1000 nm) rather than restricting the analysis to selected emission lines. Overall, the RF, Bagging, AdaBoost, and ANNs models exhibited strong generalization capabilities and robust performance in high-dimensional datasets. Furthermore, a cloud-based web application was developed to enable real-time system access, allowing users to upload new signals, perform classification, and visualize results instantaneously. This development represents a significant step toward practical VOC classification, particularly in biomedical applications such as disease diagnosis through breath analysis.