To develop models predicting Clobetasol Propionate (CP) solubility in supercritical CO₂ using machine learning techniques.
Approach:
Machine Learning Models: Adaptive Boosting (AdaBoost) was employed with Decision Tree (DT), Lasso, and Gaussian Process Regression (GPR) as core models.
Model Optimization: The Firefly Algorithm (FA) was used to tune the models and determine unknown coefficients.
Key Findings:
FB-GPR achieved the highest accuracy with an R2 score of 0.977 and RMSE of 1.11 × 10⁻².
FB-DT and FB-LASSO achieved R2 scores of 0.934 and 0.813, respectively.
The models effectively estimate CP solubility across varying temperature and pressure conditions.
Limitations:
The dataset consists of only 45 rows, which may limit the generalizability of the models.
The study focuses solely on Clobetasol Propionate and may not be applicable to other APIs.
