Abstract
This research shows the utilization of various tree-based machine learning algorithms with a specific focus on predicting Salicylic acid solubility values in 13 solvents. We employed four distinct models: cubist regression, gradient boosting (GB), extreme gradient boosting (XGB), and extra trees (ET) for correlation of drug solubility to pressure, temperature, and solvent composition. The dataset was preprocessed using the Standard Scaler to standardize it, ensuring each feature has a mean of zero and a standard deviation of one, followed by outlier detection with Cook's distance. Hyperparameter optimization made using the Differential Evolution (DE) method improved the performance of models. Monte Carlo Cross-Valuation was used in evaluation of the models. Measures including the R(2) score, Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) helped to measure their performance. With an R(2) value of 0.996, the Extra Trees model displayed remarkable accuracy and consistency, so showing better performance than other models. This study emphasizes the resilience of ensemble methods in capturing intricate data patterns and their effectiveness in regression tasks for application of pharmaceutical manufacturing.