In the field of modern medicine, biomedical informatics is crucial for enhancing illness detection, management, and treatment. A strong basis for applications that can enhance people’s health and quality of life has been established thanks to the convergence of rich medical data with potent information technology. The detection and diagnosis of breast cancer is one of the key fields where biomedical data is used and labeled data plays an important role in supervised learning models. However, labeling data can costly and time consuming. In the field of breast cancer, there is a similar problem. So that, an approach will be proposed for automatic data labeling using active learning to improve breast cancer prediction by combining least confidence sampling with fine-tuned MobileNet model called BCPDAL. Experimental research on a number of fine-tuned deep learning models combined with uncertainty sampling methods and with the dataset used as BUSI. The results have shown that proposed approach BCPDAL can achieve 90% accuracy with only 50% of the labeled data samples in the entire original dataset.