A Machine Learning\nApproach for Efficiently Predicting\nPolymer Aging from UV–Vis Spectra

This research has introduced an innovative approach that\nproficiently\nforecasts the alterations in ultraviolet–visible spectroscopy\n(UV–Vis) of polymer solutions during the aging effect. This\nmethod combines readily accessible feature descriptors with classical\nmachine learning (ML) algorithms. Traditional spectral measurements,\nwhile precise in analyzing physical properties, are limited by their\ncost and efficiency. Therefore, this paper introduces a method that\nutilizes wavelength and the blue (<i>B</i>), green (<i>G</i>), and red (<i>R</i>) color values of the solutions\nas input features. We employed seven different ML models to train\non these features with 10-fold cross-validation to ensure the reliability\nand generalizability of our results. After comparative analysis, all\nof the models performed excellently. Among them, the ExtraTree model\ndemonstrated particularly high precision and excellent predictive\nability on the testing set, with a Pearson correlation coefficient\n(<i>r</i>) of 0.9859 and a mean absolute error (MAE) of\n0.0457. This study offers a practical solution for the rapid and cost-effective\nevaluation of polymer solutions’ aging effect.