BU EC503--Final Project F'16
Member: Andrew Levy, Yixuan Xiao and Zhe Cai
In this paper, we evaluate neural networks and support vector machines as possible automated solutions to breast cancer diagnosis and prognosis. The Wisconsin Diagnostic and Prognostic Breast Cancer datasets were used to train and test the classifiers. Neural networks were optimized using grid search to determine structure and regularization parameters. SVMs were optimized using grid search to find optimal regularization and kernel parameters. Top neural network and SVM classification accuracies on the diagnosis data set were 98.1% and 97.7%, respectively. Top neural network and SVM classification accuracies on the prognosis data set were 84.5% and 81.9%.