Evaluating chemometric strategies and machine learning approaches for a miniaturized near-infrared spectrometer in plastic waste classification

Optimizing the sorting of plastic waste plays a crucial role in improving recycling process. In this contribution, we report on comparative study multiple machine learning and chemometric approaches to categorize data set derived from analysis performed with handheld spectrometer working Near-Infrared (NIR) spectral range. Conducting cost-effective NIR requires identifying appropriate techniques improve commodity identification categorization. Chemometric techniques, such as Principal Component Analysis (PCA) Partial Least Squares - Discriminant (PLS DA), Support- Vector Machines (SVM), fine tree, bagged ensemble were compared. Various pre-treatments tested collected spectra. particular, Standard Normal Variate (SNV) Savitzky-Golay derivatives signal pre-processing tools compared feature selection Gaussian Curve Fit based Radial Basis Functions (RBF). Furthermore, results combined into single predictor by using likelihood-based aggregation formula. Predictive performances models terms classification parameters Non-Error Rate (NER) Sensitivity (Sn) confusion matrices, giving broad overview rational means for approach sorting.