FTIR-Based Microplastic Classification: A Comprehensive Study on Normalization and ML Techniques DOI Creative Commons
Octavio Villegas-Camacho, Iván Francisco-Valencia, R. Alejo

и другие.

Recycling, Год журнала: 2025, Номер 10(2), С. 46 - 46

Опубликована: Март 18, 2025

This study examines the potential of machine learning (ML) and deep (DL) techniques for classifying microplastics using Fourier-transform infrared (FTIR) spectroscopy. Six commonly used industrial plastics (PET, HDPE, PVC, LDPE, PP, PS) were analyzed. A significant contribution this research is use broader more varied spectral ranges than those typically reported in state art. Furthermore, impact different normalization (Min-Max, Max-Abs, Sum Squares, Z-Score) on classification accuracy was evaluated. The assessed performance ML algorithms, such as k-nearest neighbors (k-NN), support vector machines (SVM), naive Bayes (NB), random forest (RF), artificial neural networks architectures (including convolutional (CNNs) multilayer perceptrons (MLPs)). Models trained validated FTIR-PLASTIC-c4 dataset with a 10-fold cross-validation approach to ensure robustness. results showed that Z-score significantly improved stability generalization across most models, CNN, MLP, RF achieving near-perfect values accuracy, precision, recall, F1-score. In contrast, sum squares less effective, particularly CNNs, due its sensitivity scale data distribution. Notably, consistently underperformed because limitations analyzing complex data. findings highlight effectiveness FTIR spectra broad variable automated techniques, along appropriate methods.

Язык: Английский

FTIR-Based Microplastic Classification: A Comprehensive Study on Normalization and ML Techniques DOI Creative Commons
Octavio Villegas-Camacho, Iván Francisco-Valencia, R. Alejo

и другие.

Recycling, Год журнала: 2025, Номер 10(2), С. 46 - 46

Опубликована: Март 18, 2025

This study examines the potential of machine learning (ML) and deep (DL) techniques for classifying microplastics using Fourier-transform infrared (FTIR) spectroscopy. Six commonly used industrial plastics (PET, HDPE, PVC, LDPE, PP, PS) were analyzed. A significant contribution this research is use broader more varied spectral ranges than those typically reported in state art. Furthermore, impact different normalization (Min-Max, Max-Abs, Sum Squares, Z-Score) on classification accuracy was evaluated. The assessed performance ML algorithms, such as k-nearest neighbors (k-NN), support vector machines (SVM), naive Bayes (NB), random forest (RF), artificial neural networks architectures (including convolutional (CNNs) multilayer perceptrons (MLPs)). Models trained validated FTIR-PLASTIC-c4 dataset with a 10-fold cross-validation approach to ensure robustness. results showed that Z-score significantly improved stability generalization across most models, CNN, MLP, RF achieving near-perfect values accuracy, precision, recall, F1-score. In contrast, sum squares less effective, particularly CNNs, due its sensitivity scale data distribution. Notably, consistently underperformed because limitations analyzing complex data. findings highlight effectiveness FTIR spectra broad variable automated techniques, along appropriate methods.

Язык: Английский

Процитировано

0