The accurate QSPR models for the prediction of nonionic surfactant cloud point

Quantitative structure–property relationship models were developed to predict cloud points and study the cloud phenomena of nonionic surfactants in aqueous solution. Four descriptors were selected by the heuristic method as the inputs of multiplier linear regression and support vector machine (SVM) models. Very satisfactory results were obtained. SVM models performed better both in fitness and in prediction capacity. For the test set, they gave a predictive correlation coefficient ( R) of 0.9882, root mean squared error of 4.2727, and absolute average relative deviation of 9.5490, respectively. The proposed models can identify and provide some insight into what structural features are related to the cloud points of compounds, i.e., the molecular size, structure, and isomerism of the hydrocarbon moiety and the degree of oxyethylation. They can also help to understand the cloud phenomena of nonionic surfactants in aqueous solution. Additionally, this paper provides two simple, practical, and effective methods for analytical chemists to predict the cloud points of nonionic surfactants in aqueous solution. Plot of predicted CP values vs experimental CP values for the training set and test set based on the 4-parameter model by SVM.