Early diagnosis and treatment of cancer play a crucial role in improving patient outcomes and quality of life. In recent years, deep learning has significantly advanced medical imaging analysis, supporting clinicians in achieving more accurate and efficient diagnoses. However, individual deep learning models often exhibit sensitivity to variations in input data, potentially leading to inconsistent predictions and degraded classification performance. In this paper, we propose a robust ensemble framework that integrates multiple deep learning models and optimizes their contribution weights to minimize classification errors. Prior to training, the input data undergo comprehensive preprocessing, including noise reduction, normalization, and image enhancement, to ensure high-quality and consistent representations. By systematically optimizing model weights within the ensemble, the proposed approach achieves superior stability and improved generalization capability. Experimental results demonstrate that our method attains a classification accuracy of 97.85% on the test dataset, along with notable improvements in precision, recall, and F1-score. The findings highlight the potential of the proposed ensemble-based strategy to enhance the reliability of early breast cancer detection systems and promote the effective integration of advanced information technologies into medical diagnostics. Ultimately, this research contributes to better clinical decision support and fosters more timely and accurate patient care.