Genetics Algorithm Feature Selection for Improving Aqueous Solubility Prediction

Aqueous solubility is an important property for conducting chemical reactions of the compound. In this research, we develop several machine learning models predicting aqueous reaction molecules. The open public dataset, AqSolDB, was used model development which contains 9982 data on molecule solubility. Several regression were trained dataset and their performance evaluated using mean absolute error. use model-based tree development. result showed that best prediction Categoric Boosting Regressor achieving 0.854 importance feature affected can also be calculated from calculation. It shown variable MolLogP strongly correlated with reaction. To further improve our model, selected features a genetics algorithm learning-based models. lowest error obtained 0.771 provides improvement previous calculation without selection.