A reproducible ATR-FT-IR and chemometric workflow for forensic fibre classification: Extending to culturally significant textiles in the Gulf cooperation council (GCC)

This study presents a transparent and reproducible workflow for forensic fibre classification using ATR-FT-IR spectroscopy integrated with chemometric modelling. Guided by the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework, the approach systematically combines spectral pre-processing, unsupervised exploration, and supervised learning to evaluate both conventional natural fibres and culturally significant Emirati Abaya textiles. A total of ninety-two fibre samples were analysed using principal component analysis (PCA), linear discriminant analysis (LDA), support vector machines (SVM), partial least squares regression (PLS-R), and support vector regression (SVR). For natural fibres, a minimal pre-processing sequence, multiplicative scatter correction and baseline correction, enabled clear differentiation among cotton, wool, paper, and silk, achieving 97.9% classification accuracy with LDA. In contrast, the chemically heterogeneous Abaya dataset required an extended pipeline incorporating multiplicative scatter correction, baseline correction, Savitzky–Golay first derivative, and normalization to mitigate scatter effects and resolve overlapping polyester and polyamide bands. This strategy yielded near-perfect classification performance (LDA 100%, SVM 97.7%) and strong regression accuracy (PLS-R R 2 = 0.982). Across all models, pre-processing quality exerted a greater influence on predictive performance than algorithm selection, underscoring its role as a chemically informed analytical decision rather than a purely computational adjustment. The results demonstrate that ATR-FT-IR chemometrics provides a rapid, non-destructive, and interpretable forensic tool capable of converting complex spectral data into reproducible, evidence-ready outputs. The workflow supports both exclusionary comparisons and associative linking of fibre evidence while offering a scalable foundation for regionally relevant spectral reference databases. Importantly, this study represents the first structured chemometric investigation of Emirati Abaya textiles, extending vibrational spectroscopy to culturally specific materials prevalent in Gulf Cooperation Council casework yet underrepresented in existing fibre databases.