Cancer of the breast is one of the primary causes of mortality of women across the globe. Breast abnormalities may often be diagnosed and classified with the use of ultrasound imaging. To better pinpoint health issues, radiologist diagnosis have been supported by a variety of CAD technologies created in recent decades. Using Automatic Breast Ultrasound (ABUS) scans, this research employed CNN models to differentiate between benign tumor, malignant tumor, and Normal breast tissue. Pre-processing the data is crucial in data classification since it helps the learning model function more effectively. The implemented dataset contains unprocessed image data where classes have data imbalance issue. So, it is important to resolve this imbalance problem. To resolve this issue data augmentation approach is applied on the minority class. Before the learning models are employed to the preprocessed data, the data is first partitioned into two subsets: a training set and a test set. This study’s deep learning models for breast ultrasound classification are MobileNetV2, VGG16, AlexNet, ResNet, Inception V3, and Xception. Each model is trained and then evaluated at different epochs to keep track of its progress. The study found that as the epochs go higher, data loss lowers and accuracy rises. The accuracy of MobileNetV2, VGG16, AlexNet, ResNet, Inception V3, and Xception is 98.82%, 94.70%, 95.96%, 93.50%, 92.72%, and 93.65%, respectively. MobileNetV2 outperforms with 98.82% accuracy rate the comparative to other applied models. Finally, breast cancer is classified using the MobileNetV2 model with significant accuracy and precision.