Accurate Prediction of θ (Lower Critical Solution Temperature) in Polymer Solutions Based on 3D Descriptors and Artificial Neural Networks
Jie Xu, Biao Chen, Hao Liang · Macromolecular Theory and Simulations · 2008
Abstract Quantitative structure‐property relationships were studied between descriptors representing the three‐dimensional structures of molecules and θ (LCST, lower critical solution temperature) in polymer solutions with a database of 169 data containing 12 polymers and 67 solvents. Feed‐forward artificial neural networks (ANNs) combined with stepwise multilinear regression analysis (MLRA) were used to develop the models. With ANNs, the squared correlation coefficient (R2) for θ (LCST) of the training set of 112 systems is 0.9625, the standard error of estimation (SEE) is 13.43 K, and the mean relative error (MRE) is 1.99%; in prediction of θ (LCST) using the test set of 57 systems, the MRE is 2.26%. With MLRA, the MREs for the training and test sets are 4.02% (R2 = 0.8739, SEE = 25.88 K) and 5.05%, respectively. magnified image