In this study, the genetic algorithm multiple linear regression (GA-MLR) method with 2D molecular descriptors is used to generate statistically robust and acceptable local QSAR models. Predictive R2, Q2 based on prediction of test set compounds (Q2F1, Q2F2 indicators) and the additional criterion for concordance correlation coefficient (CCC) to verify the consistency of experimental and predicted data are reported. Compounds with different numbers of hydrogen bonds showing markedly different binding affinities are elucidated in detail by molecular docking methods. The apolar surface area, electronegativity, chemical softness, hydrogen bonding, maximum condensed nucleophilic local softness, and maximum partial charge of hydrogen atoms can be used to investigate the physicochemical properties and estrogen potency of pesticides, phytochemicals and mycotoxins.