Breast cancer classification via statistical learning techniques, with applications to the Wisconsin diagnostic data set

Abstract

Breast cancer is one of the most reported types of cancer around the world and the third leading cause of death among women. The high mortality rate due to breast cancer can be best tackled via early detection so that prevention can be done in a timely and efficient manner. Several statistical-based approaches have been developed to support medical decision makers in early breast cancer detection. Various techniques may provide different desired accuracies and it is therefore vital to use the method which provides the best results. This thesis is concerned with a comparative analysis of a number of statistical learning methods, namely Random Forests, KNN algorithm, Logistic Regression and Lasso Regression. These techniques are applied to the Wisconsin breast cancer classification problem. All the above algorithms were implemented in the R programming language.