A lot of research has been done on deep learning techniques for finding breast cancer. Nevertheless, because of the complex and multidimensional nature of healthcare data, correctly diagnosing this disease continues to be a tough challenge. Data preparation is crucial to improving prediction accuracy in the context of deep learning. Preprocessing of mammography images is particularly important for the identification and categorization of breast cancer. Inherent difficulties like low image contrast and differences in breast tissue density are mostly to blame for this. This study presents a series of preprocessing procedures to solve these issues, ultimately aiding accurate breast cancer classification by improving the quality of breast images.The method described in this work makes use of thresholding and contrast enhancement techniques to enhance the visibility of breast tumor boundaries in digital mammograms. Using a Convolutional Neural Network (CNN) model, classification with labels is performed. The CNN model locates regions of masses and categorizes them as ductal carcinoma, inflammatory, triple-negative, or invasive cancer. A breast cancer dataset used for experimental evaluation produced results with a remarkable accuracy of 88.6% for both classification and prediction tasks.