Breast cancer is the most frequently diagnosed malignancy among women globally, making it the leading cause of cancer incidence worldwide. This type of cancer is particularly challenging due to its high degree of molecular heterogeneity, with many different subtypes exhibiting distinct genetic and biological characteristics. Estrogen receptors (ERs) play a crucial role in breast cancer by acting as receptors for the hormone estrogen, often referred to as the hormone receptor (HR). The presence or absence of ERs, known as ER status, significantly impacts treatment decisions, making accurate prediction of ER status essential for personalized treatment plans. Various methods have been proposed in the literature to predict ER status based on gene expression profiles using microarray genome-wide gene expression profiling. The issue of class imbalance, indicated by a notable variation in the number of samples per class and the curse of dimensionality, is a significant concern with microarray datasets. This paper introduces a novel technique, Fuzzy XGBoost, to predict ER status in breast cancer cases while addressing the class imbalance problem. The proposed algorithm enhances the traditional XGBoost method by incorporating fuzzy logic principles, improving its robustness, computational power, and accuracy in handling imbalanced datasets. Our research demonstrates the effectiveness of the proposed Fuzzy XGBoost algorithm in accurately predicting ER status across multiple gene expression profile datasets. This approach not only contributes to advancements in breast cancer treatment but also showcases the power of computational methods in bioinformatics. We evaluate our method using four datasets: GSE2990, GSE3494, GSE6532, and GSE7390, achieving predictive accuracies of 100%, 92%, 100%, and 96%, respectively. The successful application of Fuzzy XGBoost in this context highlights its potential for broader use in bioinformatics, where leveraging computational power to address challenges such as imbalanced data is critical for predictive accuracy and personalized treatment. DOI: https://doi.org/10.52783/pst.619