Optimization of Support Vector Machine and XGBoost Methods Using Feature Selection to Improve Classification Performance

Abstract

Breast cancer still ranks first in the number of cancers in Indonesia and is one of the contributors to the death rate caused by cancer. Generally occurs in women, but does not rule out it can occur in men. Globocan data in 2020 shows the number of new cancer cases in Indonesia reached 65,858 or 16.6% of the total 396,914 new cancer cases, with the number of deaths reaching 22,430 cases. Early detection can allow patients to get the right therapy and increase their chances of survival. The purpose of this study is to implement a machine learning algorithm to detect breast cancer in women, the algorithms that will be used are Support Vector Machine (SVM) and XBGoost by implementing feature selection to obtain better accuracy. The classification results of the two algorithms will be compared to find out which algorithm has the best performance. The dataset used is from the SEER NCI program in November 2017 involving 4024 patients. The research shows that of the 16 attributes contained in the dataset, there are 3 attributes (features) that have a significant effect on the classification results, namely 6th stage, reginol node positive, and tumor size. XGBoost with feature selection has a better performance of 91.4% compared to SVM which is only 89.8%.