Classification of Cancer Drug Compounds for Radiation Protection Optimization Using CART
Heri Kuswanto, Rizky Mubarok · Procedia Computer Science · 2019
This paper applies Classification and Regression Tree (CART) to classify the selected compounds of cancer drugs related to radiation protection for cancer treatment. CART is one of machine learning methods that has been widely applied in drug design due to its simple algorithm and efficiency. The classification is applied to the 5%, 10%, 20%, 30%, 35% most important features selected by Mean Decreasing Gini Index. Moreover, the performance of CART on classification with full features is also investigated. The analysis shows that classification of cancer drug compounds using CART reached 79% accuracy when it uses 5% or 10% most important features. In this case, the performance of CART is slightly better than other complex machine learning methods applied in the previous researches.