A Deep Learning Breast Cancer Prediction Framework

Abstract

Breast cancer (BrC) is now the world's leading cause of death for women. Early detection and effective treatment of this disease are the only rescues to reduce BrC mortality. The prediction of BrC diseases is very difficult because it is not an individual disease but a mixture of various diseases. Many researchers have used different techniques such as classification, Machine Learning (ML), and Deep Learning (DL) of the prediction of the breast tumor into Benign and Malignant. However, still there is a scope to introduce appropriate techniques for developing and implementing a more effective diagnosis system. This paper proposes a DL prediction BrC framework that uses a selected Bidirectional Recurrent Neural Network (BRNN). An efficient fast and accurate optimizer is needed to train the neural network used. The more recent Dynamic Group-based Cooperative Optimization Group (DGCO) algorithm is modified MDGCO for this purpose. The Deep Learning Breast Cancer Prediction Framework (DLBCPF) includes four layers: preprocessing, feature selection, optimized Recurrent Neural Networks, and prediction. Four different Wisconsin BrC datasets are used to test the validity of the proposed framework and optimizer against others. The results obtained have shown the superiority of both the framework DLBCPF and the optimizer MDGCO when they are compared to others.