EVALUATING MACHINE LEARNING MODELS FOR BREAST CANCER SURVIVAL PREDICTION

ABSTRACT

Breast Cancer (BC) is a deadly disease and known globally as the second killer disease among women. It can be treated but with a low survival rate. Machine Learning (ML) techniques have been identified as one of the best computational models for Breast cancer survival (BCS) prediction. The aim of this research is to develop and compare ML techniques for predicting BCS. A total of 4,640 records were downloaded from U-BRITE’s repository which consists of 15 BC clinical features consisting of data for 3,408 living patients and 1,232 killed by BC was used to develop 10 ML models and sensitivity analysis was used for evaluation. RF model with Chi square features gave the best results of sensitivity, specificity, accuracy and AUC of 79.51%, 83.95%, 81.75% and 89.53% respectively. Random forest Model can be used to assist clinicians in predicting BCS.

Keywords: Breast Cancer Survival, Prediction, Feature Selection, Machine Learning