Abstract
A graphene–chitosan modified glassy carbon electrode (Gr-Chit/GCE) was evaluated for the differential pulse voltammetric (DPV) determination of ibuprofen (IBF) and paired with a transparent machine learning (ML) calibration audit. In a 0.1 M phosphate-based electrolyte (PBE, pH 7.5), the modified electrode gave a linear DPV response from 100 to 488 µmol/L after background subtraction, following Ip = 0.0828 + 0.0275C (R2 = 0.9970). The recalculated limit of detection (LOD) and the limit of quantification (LOQ) were 22.27 and 74.25 µM, respectively. Cyclic voltammetric characterisation with [Fe(CN)6]3–/4– showed a diffusion-controlled behaviour and a 4.2-fold increase in the electroactive surface area after Gr-Chit deposition. Interference tests indicated that ascorbic acid was resolved at lower potential, while acetaminophen contributed in the IBF potential region without masking the analyte response. The leave-one-out cross-validation of six calibration models showed that the ordinary least-squares inverse calibration gave the lowest prediction error (root-mean-square error, RMSE = 12.11 µmol/L), whereas more complex nonlinear models showed a poorer generalisation for the five-point calibration set. The study therefore positions ML not as a black-box replacement for electroanalytical calibration, but as an auditable model-selection layer for small voltammetric datasets.