An integrated ML model for the prediction of the melting points, phase diagrams, and eutectic points of the Type III and V deep eutectic solvents

The environmental impact of conventional solvents is a widely acknowledged problem. Consequently, there has been an increasing interest in deep eutectic solvents (DESs) as green alternatives to conventional solvents. Types III and V DESs are particularly prevalent. However, the large number of potential combinations poses a significant challenge in identifying optimal DESs for industrial applications. A crucial parameter here is the melting point, which directly influences the liquid window range of DES. In this work, we propose an integrated model comprised of a multilayer perceptron (MLP), multiple linear regression (MLR), support vector machine regression (SVR), K-nearest neighbor regression (KNN), and random forest regression (RFR) models, all assembled in a stacked configuration. Our integrated model leverages a comprehensive database of DES melting points (2315 data points) and COSMO-RS (Conductor-like screening model for real solvents) descriptors, facilitating accurate predictions and boosting overall model performance (R2 = 0.99, AARD = 1.2402 %). This integrated model reliably predicts DES melting points, phase diagrams, and eutectic points, serving as a robust tool for screening DESs and enhancing their industrial applications.