Nanoparticle Properties Modeling Using Linear and Logistic Regression Model on Secondary Experiential Data

Abstract

The techniques used to characterize nanoparticles are important to the development of nanotechnology and materials science. Nevertheless, experimental optimization of a nanoparticle properties, including size, stability, and functionality, can be both inefficient and costly. In this study, a predictive framework, based on linear and logistic regressions is constructed, to derive predictive models to assess nanoparticle characteristics employing secondary experimental data sourced from literature and open-access repositories. In this case, linear regression is used to predict size-dependent continuous outcomes, while logistic regression is applied to predict categorical outcomes, like stability. The models are trained on a portion of the dataset which is then held out for validation, and the models' performance is evaluated on metrics including mean squared error, R^2 score, accuracy, and confusion matrices. The regression models built achieved a good degree of accuracy, which demonstrates modeling based on regression can provide reliable outcomes which will save a chemist significant time in experimental optimal synthesis parameter trials. This demonstrates the value of cross-disciplinary intelligent modeling in nanoscience to improve the decision-making process to speed up the development of multifunctional nanoparticles.