Randomized trial demonstrates high accuracy in lignin solubility prediction using a hybrid COSMO-RS and machine learning framework, highlighting its potential for solvent design.
Key Points
The aim is to develop a framework for predicting lignin solubility using deep eutectic solvents (DES) combined with machine learning techniques.
Developed a hybrid COSMO-RS and machine learning framework for solvent screening.
Utilized residual learning to enhance prediction accuracy for lignin solubility.
Applied multilayer perceptron (MLP) for improved model performance.
Achieved an increase in test R2 from 0.9182 to 0.9941 with MLP.
Reduced average absolute relative deviation (AARD %) to below 8.94%.
Hydrogen bond interactions and COSMO-RS solubility calculations are crucial for solubility predictions.