<p>This research proposes an automated system by combining a novel image preprocessing approach with a machine learning-based ensemble technique to identify and classify breast cancers through the analysis of ultrasound images. Breast cancer ranks as the second leading cause of mortality in women, surpassed only by lung cancer. Early detection can significantly reduce female death rates. Many studies have used machine learning techniques and artificial intelligence tools to automatically classify cancers from ultrasound images with varying success. This research focuses on image preprocessing and feature extraction to enhance machine learning classifier performance. The proposed method applies traditional image preprocessing techniques such as data augmentation, ROI identification, image scaling, and greyscale conversion, followed by fuzzy c-means clustering for unsupervised segmentation. The features are then extracted using a pre-trained VGG-19 model from the input images, which plays a significant role in classifying the images as Benign or Malignant. The extracted feathers are optimized via the binary grey wolf algorithm. Finally, the optimal features are then fed to a machine learning-based ensemble classifier (e.g., bootstrap aggregation based on bagging technique) for breast cancer detection. The methodology incorporated enhanced Breast Ultrasound Images, achieving an accuracy rate of 98.14%, precision 98.17%, recall 97.76%, F1-score 95.91%, and AUC-ROC of 98.99%. Compared to contemporary methods, the proposed framework demonstrates superior performance. The proposed approach ensures accurate and efficient breast cancer detection using optimized feature selection and classification.</p>