Penerapan Metode Pemangkasan Dalam Cart (Classification And Regression Tree)

Abstract

CART (Classification and Regression Tree) is an exploratonj data rnet!lod which analyze the relationship between dependent and independent variables zclhich have large and complex data size i.e.: categorical data (nominal and ordinal) and numerical data (interval and ratio). This method is usefill to determine the independent variables. 112 this study,CART is applied predict the clzild n11trieizt. The application of prulzing methods in CART is beneficial to avoid over fitting and zinderfitting in order to make the tree regression. This application uses after treatment clzild nutrient's data in Kecanzatan Bogor Tinzur. This study shows that the best method is pruning with high decreasing deviance values because it's simple nzethod and can be used for small sample. However, pruning with high decreasing resubstitution relative error and high decreasing cross-validated relative error are more accurate fo determine the independent variables also to predict dependent variables with small sample. The application of CART'S nzodel to predict child nutrient reslllts consistence and logic independent variables. Therefore CART'S nzodel is a precise model to predict child nutrient. Key words : CART Methods