Ensemble Learning Applied to Classification of Malignant and Benign Breast Cancer

Abstract: In this study, we show how ensemble learning can be useful for the future of breast cancer diagnosis. The chosen ensemble learning method was bagging, which made use of the classifiers Support Vector Machine (SVM), Decision Tree (DT) and Naive Bayes (NB) in order to classify mammograms as benign or malignant. The results achieved with bagging were compared to the results of each individual classifier previously mentioned. Overall, the results showed that the benefits of ensemble learning were varying, dependent on certain factors. Affecting aspects were: which classifier that was used, chosen method for extracting input data, but also which tumor types that were used in training and evaluation of each classifier. While classification using DT improved significantly with bagging, SVM and NB gave negligible performance benefits. Finally, this study only scratched the surface of known ensemble learning methods, indicating that there may be a lot of room for future research in the area.