Machine learning approaches for breast cancer diagnosis and prognosis

Abstract

For breast cancer diagnosis in patients, radiologists conduct Fine Needle Aspirate (FNA) procedure of breast tumor. This procedure reveal features such as tumor radius, concavity, texture and fractal dimensions. These features are further studied by medical experts to classify tumor as Benign or Malignant. The cardinal aim of this paper is to predict breast cancer as benign or malignant using data set from Wisconsin Breast Cancer Data using sophisticated classifiers such as Logistic Regression, Nearest Neighbor, Support Vector Machines. Furthermore, using Wisconsin Prognostic data set, probability of recurrence in affected patients in calculated. As a result, a concrete relationship between precision, recall and the number of features in the data set is achieved, which is shown graphically.