Feasible features of Breast Cancer Recurrence (BCR) patients using machine learning algorithms

Cancer is a disease in which cells in the body grow out of control and is consistently named for the parts of the body where it starts from the breast. BCR is breast cancer that returns after initial treatment and may occur within months or years. This study aims to identify the feasible feature in predicting BCR using four machine learning algorithms. The study utilized 10377 secondary data from the official statistic of the Ministry of Health and Medical Education and the Iran Cancer Research Center. Naive Bayes (NB), Random Forest (RF), Gradient Boosted Tree (GBT), and Logistic Regression (LR) were utilized by using RapidMiner to obtain a good classifier's performance to be evaluated to determine the best model that can accurately predict the BCR, and the significant risk factors of BCR using the best model. The results show that the best model with the highest accuracy is GBT, with a ratio of51 %, and the most essential feature of this algorithm is radiotherapy.