Trained SVM model parameters for stress detection using salivary Raman spectra

This repository contains trained parameter sets of Support Vector Machine (SVM) models developed to identify the presence of stress from Raman spectra of saliva samples. The deposit includes two related SVM models: (1) A general SVM parameter set, provided for methodological reference and reuse in similarclassification problems. (2) A specific SVM model trained using 12 predictors, optimized for stress detection fromsalivary Raman spectra. For each model, the following parameters are provided: • μ (mu): mean values used for predictor standardization. • σ (sigma): standard deviation values used for predictor standardization. • SV: support vectors selected by the trained SVM model. • α (alpha): SVM coefficients associated with the support vectors. • Bias: independent bias term of the classifier. • Kernel parameters: information related to the Gaussian (RBF) kernel, including the kernel scale parameter. • Classes: classification labels (stress / no stress). • isSV: logical index indicating which original observations correspond to support vectors (when available). These parameter sets allow the trained classifiers to be reproduced, validated, and reusedwith new Raman spectroscopy datasets, as well as compared with other machine learning approachesapplied to optical biosensing and biomedical spectroscopy. All models were trained in MATLAB using a Gaussian (RBF) kernel and 5-fold cross-validation.