Green processing based on supercritical carbon dioxide for preparation of nanomedicine: Model development using machine learning and experimental validation

Abstract

Solubility data for ANA (Anastrozole) drug in supercritical solvent was investigated in this study, and models were developed to estimate the solubility values. The main aim was to provide a predictive methodology for determination of drug solubility in wide range of operational parameters for advanced green pharmaceutical manufacture. The properties used are temperature and pressure which were considered as the models’ inputs. Modeling has been done using three models based on the support vector regression. These models include support vector regression (with polynomial kernel), boosted support vector machine with AdaBoost, and improved support vector machine with bagging. These models were evaluated after optimization, and all three models have a coefficient of determination (R2) higher than 0.98. Also considering RMSE, AdaBoosted SVR, Bagging SVR, and SVR have error rates of 2.31E-01, 4.31E-01, and 5.01E-01.