Imbalanced Breast Cancer Prediction Using CNN Through Multi-Net Framework

Cancer remains to be one of the most prevailing diseases in the world in which breast cancer remains unaddressed. There are several methods for identifying breast cancer, but still, breast cancer detection using automated algorithms remains a problem within the literature. Though several automated algorithms are used to identify breast cancer, the one with the highest accuracy needs to be identified. CNN abbreviated as Convolutional neural network, a fragment of a deep learning model, was created to accurately identify breast cancer. There are two objectives in this research where the introductory step is to look into assorted DL methods for cataloguing breast cancer images. Secondly, to identify the best feature extraction model. The proposed system, analysis of different performance metrics results that were obtained for the deep neural networks showed that other factors, such as transfer learning approaches and pre-processing techniques, may have an impact on the model’s ability to achieve improved prediction of breast cancer. Approaches like EfficientNet, Inception V3, VGG-16, and Modified VGG-19 have been used to identify the most effective results for the ImageNet database. In this experiment the experimental design for each model tested on breast cancer histopathology images is carefully analyzed and discussed.