Hansen solubility parameters (HSP) are widely recognized as a viable method to numerically estimate the interactions between solvents and other chemical compounds. A data-driven approach to predict the HSP is attractive as conventional molecular modeling of solvent systems is complex, time-consuming, and expensive. In the presented work, we employ a variety of supervised learning techniques such as simple linear regression methods (e.g., elastic nets), kernel-based methods (e.g., support vector machines), ensemble-based methods (e.g., random forest regressor), and artificial neural networks to predict the HSPs of solvents. In addition, we perform a hyper-parameter optimization of these models using genetic algorithms. In order to build physics into our data-derived prediction models, we explore the utility of <i>first-principles</i> molecular descriptors of our solvent systems from electronic structure calculations as features. Here, we present a comparison of results from different algorithms and different combinations of features, and perform a cost-benefit analysis of using expensive features in these models. We then apply this model to understand the behavior of different polymer-solvent systems. The results indicate that simple linear regression models can be used to give accurate predictions for HSPs. The accuracy of these predictions is comparable to, or an improvement on the best existing models reported in the literature. The inclusion of features from electronic structure calculations only marginally improves the accuracy of our models. All models were generated using our group’s machine learning and informatics program suit <i>ChemML</i>, along with a module to predict solvent-polymer interactions using the concept of the 3-dimensional Hansen space. The simple linear regression models readily yield insights into those molecular features that have a predominant impact on the HSP.

Machine learning methods for hansen solubility parameters and their application in predicting solvent-polymer interactions

Macromolecular Chemistry: The Second Century

[Author Benefits](https://scimeetings.acs.org/?utm_source=pubs_content_marketing&utm_medium=website&utm_campaign=0320_MCF_NPI_Launch_Spring_Homepage&ref=pubs_content_marketing)

[How to use SciMeetings](https://storage.googleapis.com/dl.morressier.com/acs/SciMeetings-ACSSpring2020-HowToGuide_M3020-Final.pdf)

[Frequently Asked Questions](https://storage.googleapis.com/dl.morressier.com/acs/SciMtng_ACSNMSp2020_FAQ_M3120_Final.pdf)

ACS Spring 2020 National Meeting & Expo

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.