Machine Learning and Deep Learning Approaches in Breast Cancer Survival Prediction Using Clinical Data

摘要

Breast cancer survival prediction can havean extreme effect on selection of best treatment pro-tocols. Many approaches such as statistical or ma-chine learning models have been employed to predictthe survival prospects of patients, but newer algo-rithms such as deep learning can be tested with theaim of improving the models and prediction accura-cy. In this study, we used machine learning and deeplearning approaches to predict breast cancer surviv-al in 4,902 patient records from the University ofMalaya Medical Centre Breast Cancer Registry. Theresults indicated that the multilayer perceptron (MLP),random forest (RF) and decision tree (DT) classifierscould predict survivorship, respectively, with 88.2 %,83.3 % and 82.5 % accuracy in the tested samples.Support vector machine (SVM) came out to be lowerwith 80.5 %. In this study, tumour size turned out tobe the most important feature for breast cancer sur-vivability prediction. Both deep learning and ma-chine learning methods produce desirable predictionaccuracy, but other factors such as parameter con-figurations and data transformations affect the ac-curacy of the predictive model.