Breast cancer is one common cancer with high fatality rate worldwide. Based on consideration of pharmacokinetics and toxicity, this study aims to build a model to investigate compounds which can be used for candidate drugs to resist breast cancer. Using SVR model, GBRT model and random forest algorithm, this study established a quantitative prediction model for the bioactivity of compounds to Erα, and SVM model, SGD model and random forest algorithm were constructed for predicting the 5 ADMET properties of the compounds. Finally, based on the above models and analysis, compounds with better biological activity against ERα and better ADMET properties were selected in this paper. This study definitively established the model for selecting compounds which can be used as anti-breast drug candidates, and further research is needed to add more pharmacokinetic properties in the model to improve its applicability.