Electrode impedance modeling based on XGboost algorithm for analyzing the antioxidant properties of juice

Abstract

This study presents a novel approach integrating electrochemical impedance spectroscopy with an XGBoost algorithm for analyzing antioxidant properties in fruit juices. Three tyrosinase-immobilized electrode configurations were compared: screen-printed carbon electrodes with polyvinyl alcohol, glutaraldehyde, and human serum albumin crosslinkers. Chronoamperometry and cyclic voltammetry experiments were conducted to assess the redox kinetics and interfacial properties of the biosensors. The impedance spectra of the biosensors were recorded in juice samples and equivalent circuit modeling was performed to extract charge transfer resistance, double layer capacitance and other key parameters as input features for the XGBoost model. The model was trained on these EIS markers paired with reference antioxidant capacity assay values.]{.mark} The SPE/Tyr/HSA/GA sensor exhibited superior predictive accuracy, with a mean absolute error of 0.345, root mean squared error of 0.444, and an R-squared value of 0.980. Feature importance analysis revealed charge transfer resistance and double-layer capacitance as the most influential predictors. The XGBoost model outperformed multiple linear regression and random forest baselines, demonstrating high consistency with standard antioxidant assays. These findings highlight the potential of machine learning models combined with electrochemical impedance spectroscopy biosensors for rapid and accurate nutrient monitoring, providing substantial advancements in food analysis and health monitoring.