Journals / Erciyes Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Dergisi / 2026 / Cilt: 42 Sayı: 3

Classification of Synthetic Biosensor Data with Quantum Simulations Using Feature Optimization and Explainable Machine Learning Supported by RFECV

Öznitelik Optimizasyonu ve RFECV Destekli Açıklanabilir Makine Öğrenmesi ile Kuantum Simülasyonlu Sentetik Biyosensör Verilerinin Sınıflandırılması

Halil İbrahim Sarı(KUTAHYA DUMLUPINAR UNIVERSITY)0009-0006-3964-3101
Kübra Keser(KUTAHYA DUMLUPINAR UNIVERSITY)0000-0001-7215-7232
Pages
—
Publication Date
2026-09-16
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

The results indicate that the combined use of GroupKFold-based validation, RFECV, and explainable AI methods facilitates the reliable and interpretable classification of biosensor data.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Background/Aim: Accurate detection of trace amounts of heavy metal toxins in environmental samples requires sensitive and reliable diagnostic systems.
📊
Key Findings & Data:This study presents a machine learning framework to classify synthetic biosensor signals—generated via quantum simulation—representing five toxin classes: Control, Lead, Mercury, Cadmium, and Arsenic.
💡
Conclusion & Impact:The results indicate that the combined use of GroupKFold-based validation, RFECV, and explainable AI methods facilitates the reliable and interpretable classification of biosensor data.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Toksin · sınıflandırma · biyosensör

Abstract

Background/Aim: Accurate detection of trace amounts of heavy metal toxins in environmental samples requires sensitive and reliable diagnostic systems. This study presents a machine learning framework to classify synthetic biosensor signals—generated via quantum simulation—representing five toxin classes: Control, Lead, Mercury, Cadmium, and Arsenic. Methods: Methods: A total of 75 pre-calculated temporal, spectral, entropy, nonlinear, and wavelet features were utilized. To mitigate the risk of data leakage and obtain reliable performance estimates, the Recursive Feature Elimination with Cross-Validation (RFECV) method was integrated into a GroupKFold-based validation framework. The number of features selected across the five outer folds was 47, 73, 55, 47, and 41, respectively, resulting in an average of 52.6 features. Six ensemble learning algorithms were evaluated using a fixed 47-feature structure. Model interpretability was examined via SHAP analysis, while the contributions of feature groups were assessed through ablation and Wilcoxon signed-rank analyses. Results: Results: Configurations with 75 and 41 features were evaluated in a separate direct comparison experiment, yielding average accuracies of 84.06% and 83.73%, respectively. The main evaluation was conducted on the 47-feature configuration. Among the six ensemble learning algorithms, HistGradientBoosting demonstrated the highest performance, achieving an average accuracy of 85.27% and a weighted F1-score of 85.25%. The Random Forest model exhibited the highest discriminative performance, with a Macro-AUC of 0.9799. The Friedman test indicated that performance differences among the algorithms were statistically significant (p = 0.0186), while Nemenyi post-hoc analysis assessed the relative performance differences between the models. The 47-feature configuration was used for the ablation analysis. The most significant drop in performance occurred upon the removal of frequency-domain features, followed by the removal of non-linear features. The Wilcoxon signed-rank test, applied to evaluate the contribution of non-linear features, demonstrated that this group provided a positive trend and contribution to model performance (p = 0.0625). Conclusion: SHAP-based analyses have demonstrated that features such as Hjorth Mobility, D1 Energy, D1 Standard Deviation, Petrosian Fractal Dimension, and Spectral Entropy contribute significantly to the model's decisions. The results indicate that the combined use of GroupKFold-based validation, RFECV, and explainable AI methods facilitates the reliable and interpretable classification of biosensor data.

Özet

Çevresel örneklerdeki eser miktardaki ağır metal toksinlerinin doğru bir şekilde tespiti, yüksek hassasiyete sahip teşhis sistemleri gerektirir. Kuantum hesaplama yeni öznitelik temsil yetenekleri sunsa da, fiziksel kuantum biyosensörleri henüz büyük ölçüde deneysel aşamadadır. Bu çalışmada, üretilen kuantum simülasyonlu sentetik biyosensör sinyallerini sınıflandırmak amacıyla tasarlanmış bir makine öğrenmesi mimarisi sunulmaktadır. Beş hedef toksin sınıfı (Kontrol, Kurşun, Cıva, Kadmiyum, Arsenik) genelindeki sinyal dinamiklerini temsil etmek üzere toplam 75 önceden hesaplanmış zamansal, spektral, entropi, doğrusal olmayan ve dalgacık özniteliği kullanılmıştır. Veri sızıntısı riskini azaltmak ve daha güvenilir performans tahminleri elde etmek amacıyla, RFECV tabanlı öznitelik seçimi GroupKFold tabanlı çapraz doğrulama çerçevesinde gerçekleştirilmiştir. Beş dış katmanda RFECV tarafından seçilen öznitelik sayıları 47, 73, 55, 47 ve 41 olup ortalama 52.6 öznitelik olarak bulunmuştur. Ayrıca, 75 ve 41 öznitelikli yapılar ayrı bir doğrudan karşılaştırma deneyinde değerlendirilmiş ve ortalama doğruluk sırasıyla %84.06 ve %83.73 olarak elde edilmiştir. Ana değerlendirme 47 öznitelikli yapı üzerinde gerçekleştirilmiştir. Altı ensemble öğrenme algoritması arasında HistGradientBoosting %85.27 ortalama doğruluk ve %85.25 Weighted F1 skoru ile en yüksek performansı sergilemiştir. Rastgele Orman modeli 0.9799 Macro-AUC ile en yüksek ayırt etme performansını göstermiştir. Friedman testi algoritmalar arasındaki performans farklarının istatistiksel olarak anlamlı olduğunu göstermiş (p = 0.0186), Nemenyi post-hoc analizi ise modeller arasındaki göreli performans farklılıklarını değerlendirmiştir. Ablasyon analizinde 47 öznitelikli yapı kullanılmıştır. En belirgin performans düşüşü frekans alanı öznitelliklerinin çıkarılmasıyla meydana gelmiş, bunu doğrusal olmayan özniteliklerin çıkarılması takip etmiştir. Doğrusal olmayan özniteliklerin katkısını değerlendirmek için uygulanan Wilcoxon signed-rank testi, bu grubun model performansına olumlu bir eğilim ve katkı sağladığını göstermiştir (p=0.0625). SHAP tabanlı analizler, Hjorth Mobility, D1 Energy, D1 Standard Deviation, Petrosian Fractal Dimension ve Spectral Entropy gibi özniteliklerin model kararlarına önemli katkılar sağladığını göstermiştir. Sonuçlar, GroupKFold tabanlı doğrulama, RFECV ve açıklanabilir yapay zekâ yöntemlerinin birlikte kullanımının biyosensör verilerinin güvenilir ve yorumlanabilir sınıflandırılmasına katkı sağladığını göstermektedir.

🔖

Cite This Article

Sarı, H. İ., Keser, K. (2026). Classification of Synthetic Biosensor Data with Quantum Simulations Using Feature Optimization and Explainable Machine Learning Supported by RFECV. Erciyes Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Dergisi, Cilt: 42 Sayı: 3. https://doi.org/10.65520/erciyesfen.1994336

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (27)

  1. D. Chakif, J. Furrer, Heavy Metal Toxicity in Clinical and Environmental Health: Sources, Mechanisms, Diagnostics, and Evidence-Based Management of Mercury, Lead, Cadmium, and Arsenic, Int. J. Mol. Sci. 27 (2026). https://doi.org/10.3390/ijms27083513.
  2. K. Keser, M.Ç. Soylu, Detection of Phenylarsine Oxide in Drinking Water Using an Impedimetric Electrochemical Sensor with Gelatin-Based Solid Electrolyte Enriched with Mercaptoethanol: A Novel Prospective Green Biosensor Methodology, ACS Omega 7 (2022) 43111–43121. https://doi.org/10.1021/acsomega.2c05516.
  3. Z. Ling, L. Yang, W. Zhang, T. Yao, H. Xu, Detection of food contaminants: A review of established rapid analytical techniques and their applications and limitations, Food Safety and Health 2 (2024) 72–95. https://doi.org/10.1002/fsh3.12032.
  4. J. Yan, R. Liang, W. Yan, X. Wang, Machine Learning-Enhanced MEC Sensors with Feature Engineering for Quantitative Analysis of Multi-Component Toxicants, Biosensors (Basel). 16 (2026). https://doi.org/10.3390/bios16030144.
  5. S.A. Lahari, N. Kumawat, K. Amreen, R.N. Ponnalagu, S. Goel, IoT integrated and deep learning assisted electrochemical sensor for multiplexed heavy metal sensing in water samples, NPJ Clean Water 8 (2025). https://doi.org/10.1038/s41545-025-00441-x.
  6. M. Kim, S. Adjei-Nimoh, J. Park, S.S. Egon, E. Choi, G. Pak, K. You, A.H.M.A. Sadmani, K.T. Kim, W.H. Lee, Calibration-free on-site detection of microcystin-LR using integrated biosensing, multi-parameter water quality monitoring, and machine learning, Water Res. 298 (2026). https://doi.org/10.1016/j.watres.2026.125832.
  7. K. Tafadzwa Mpofu, P. Mthunzi-Kufa, Recent Advances in Quantum Biosensing Technologies, in: 2024. https://doi.org/10.5772/intechopen.1007858.
  8. U. Zvi, D.R. Candido, A.M. Weiss, A.R. Jones, L. Chen, I. Golovina, X. Yu, S. Wang, D. V. Talapin, M.E. Flatté, A.P. Esser-Kahn, P.C. Maurer, Engineering spin coherence in core-shell diamond nanocrystals, Proc. Natl. Acad. Sci. U. S. A. 122 (2025). https://doi.org/10.1073/pnas.2422542122.
  9. S. Srivastava, W. Wang, W. Zhou, M. Jin, P.J. Vikesland, Machine Learning-Assisted Surface-Enhanced Raman Spectroscopy Detection for Environmental Applications: A Review, Environ. Sci. Technol. 58 (2024) 20830–20848. https://doi.org/10.1021/acs.est.4c06737.
  10. H. Tan, J. Jin, C. Fang, Y. Zhang, B. Chang, X. Zhang, H. Yu, W. Shi, Deep Learning in Environmental Toxicology: Current Progress and Open Challenges, ACS ES&T Water 4 (2024) 805–819. https://doi.org/10.1021/acsestwater.3c00152.
  11. K. Jaganathan, H. Tayara, K.T. Chong, An Explainable Supervised Machine Learning Model for Predicting Respiratory Toxicity of Chemicals Using Optimal Molecular Descriptors, Pharmaceutics 14 (2022). https://doi.org/10.3390/pharmaceutics14040832.
  12. C.-C. Chang, Y. Chen, C.-Y. Chen, Towards decision-ready explainable machine learning for water quality management using a consistency index and cross-validated feature-selection framework, Environ. Res. 297 (2026) 124207. https://doi.org/10.1016/j.envres.2026.124207.
  13. A. Barredo Arrieta, N. Díaz-Rodríguez, J. Del Ser, A. Bennetot, S. Tabik, A. Barbado, S. Garcia, S. Gil-Lopez, D. Molina, R. Benjamins, R. Chatila, F. Herrera, Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI, Information Fusion 58 (2020) 82–115. https://doi.org/10.1016/j.inffus.2019.12.012.
  14. M. Khajavian, J.-H. Jang, J.-Y. Kwon, J.-M. Lee, S. Lee, M.-H. Hwang, E. Yang, J.K. Jang, K.-J. Chae, Harnessing interpretable machine learning: SHapley additive exPlanations (SHAP)-driven insights, transformative impact, and controversies in adsorption-based environmental remediation, Inorg. Chem. Commun. 186 (2026) 116269. https://doi.org/10.1016/j.inoche.2026.116269.
  15. L.T. Yogarathinam, S.I. Abba, J. Usman, D.U. Lawal, I.H. Aljundi, Predicting micropollutant removal through nanopore-sized membranes using several machine-learning approaches based on feature engineering, RSC Adv. 14 (2024) 19331–19348. https://doi.org/10.1039/d4ra02475c.
  16. https://zenodo.org/records/15345426, (n.d.).
  17. Pagidipalla, M. S. (2025). Dataset and Code for Quantum-AI Enhanced Biosensor for Environmental Toxin Detection: A Synthetic Data Approach [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.15345426, (n.d.).
  18. zenodo.15345425, (n.d.).
  19. M. Awad, S. Fraihat, Recursive Feature Elimination with Cross-Validation with Decision Tree: Feature Selection Method for Machine Learning-Based Intrusion Detection Systems, Journal of Sensor and Actuator Networks 12 (2023). https://doi.org/10.3390/jsan12050067.
  20. C. Chen, J. Liang, W. Sun, G. Yang, X. Meng, An automatically recursive feature elimination method based on threshold decision in random forest classification, Geo-Spatial Information Science 28 (2025) 1494–1519. https://doi.org/10.1080/10095020.2024.2387457.
  21. Z. Zheng, Y. Jiang, Q. Zhang, Y. Zhong, L. Wang, A Feature Selection Method Based on Relief Feature Ranking with Recursive Feature Elimination for the Inversion of Urban River Water Quality Parameters Using Multispectral Imagery from an Unmanned Aerial Vehicle, Water (Switzerland) 16 (2024). https://doi.org/10.3390/w16071029.
  22. W.H. Alawee, A. Basem, L.A. Al-Haddad, Advancing biomedical engineering: Leveraging Hjorth features for electroencephalography signal analysis, J. Electr. Bioimpedance 14 (2023) 66–72. https://doi.org/10.2478/joeb-2023-0009.
  23. G.L. Turdean, Design and Development of Biosensors for the Detection of Heavy Metal Toxicity, International Journal of Electrochemistry 2011 (2011) 1–15. https://doi.org/10.4061/2011/343125.
  24. A.F. Mohamed, V. Jusas, Advancing Fractal Dimension Techniques to Enhance Motor Imagery Tasks Using EEG for Brain–Computer Interface Applications, Applied Sciences (Switzerland) 15 (2025). https://doi.org/10.3390/app15116021.
  25. Y. Ma, W. Shi, C.K. Peng, A.C. Yang, Nonlinear dynamical analysis of sleep electroencephalography using fractal and entropy approaches, Sleep Med. Rev. 37 (2018) 85–93. https://doi.org/10.1016/j.smrv.2017.01.003.
  26. C. Fan, W. Li, J. Ning, D. Yang, Z. Zhu, W. Cheng, Y. Liang, H. Hu, Interfacing chemistry and materials for heavy metal ion sensing: mechanistic foundations and adaptive design strategies, Mater. Chem. Front. 10 (2026) 1368–1408. https://doi.org/10.1039/D6QM00046K.
  27. J. Demšar, Statistical Comparisons of Classifiers over Multiple Data Sets, 2006.