This study presents a novel approach to optimizing breast cancer treatment recommendations through the integration of machine learning techniques and feature importance analysis. A real-time dataset of 154 patients was obtained from a private hospital, encompassing patient demographics, clinical features, tumor characteristics, key markers such as ER, PR, HER-2, Ki-67 and cancer stage. Our research aims to identify critical features influencing treatment decisions and guiding breast cancer treatment. Various machine learning models are evaluated, showcasing robust performance metrics such as accuracy, precision, recall, and F1 score. Through TOPSIS ranking, the ensemble model emerges as the most effective in guiding treatment recommendations. Internal validation of the ensemble model is performed using K-fold cross-validation, and the model has performed consistently. External validation is carried out by a doctor.