Improving Machine Learning Algorithms for Breast Cancer Prediction

Abstract

Early prediction of breast cancer can prevent death or receiving late treatment. The purpose of this research is to improve machine learning algorithms in predicting breast cancer that will assist patients and healthcare systems. The machine learning algorithms for the prediction of breast cancer are the methods applied in this research by using these following algorithms which are decision tree, random forest, naive Bayes, and gradient boosting due to their high performance. This research uses data from the breast cancer of Wisconsin (diagnostic) dataset of the general surgery department. The results from this research are that by using the stratified k-fold cross validation as a part of the random forest classifier achieved 100% for all four performance scores which are accuracy, recall, precision and F1. The stratified k-fold also improved two machine learning algorithms. In addition, data visualization was applied to the random forest algorithm for result understanding. The implication from the best method is that it could increase the number of accurate breast cancer detections. The values by selecting the best method from this research could assist doctors in early breast cancer detection and increase the number of breast cancer survival rates by receiving early treatment from accurate prediction.