Identification of Breast Tumor Using Hybrid Approach of Independent Component Analysis and Deep Neural Network

Among the most prevalent and serious diseases that affect women is breast cancer. A large number of women succumb to breast cancer each year. Breast cancer must be detected in its early stage. To deal with this challenge, Deep Neural Network (DNN) is used to achieve the success. In medical science, DNN has played a vital role in the diagnosis of a wide range of illnesses. In this study, we investigate the use of Regularized Deep Neural Network (R-DNN) for the prediction of breast cancer. A variety of optimization techniques, such as Limited-memory Broyden Fletcher Goldfarb Shanno (L-BFGS), Stochastic Gradient Descant (SGD), Adaptive Moment Estimation (Adam), and activation functions like as Tanh, Sigmoid, and Rectified Linear Unit (ReLu) are used in the simulation of R-DNN. The Independent Component Analysis (ICA) approach is used to identify the most effective features to be used in the study. To measure the efficacy of the model, training and testing of the proposed network is carried out using the Wisconsin Breast Cancer (WBC) (Original) dataset from the University of California at Irvine (UCI) Machine Learning repository. The detailed analysis of the accuracy is carried out and compared to the accuracy of other author’s model. We find that the proposed network attains the highest accuracy.