IMPLEMENTATION OF FEATURE IMPORTANCE XGBOOST ALGORITHM TO DETERMINE THE ACTIVE COMPOUNDS OF SEMBUNG LEAVES (BLUMEA BALSAMIFERA)

Abstract

Sembung is a medicinal plant native to Indonesia that grows optimally in tropical climates. The secondary metabolite compounds found in the leaves of sembung are biopharmaceutical active ingredients. Fourier Transform Infrared (FTIR) spectroscopy can identify the functional compounds in sembung leaves by analyzing unique peaks in the spectrum, which correspond to specific functional groups of the compounds. In this research, 35 observations were made with 1,866 explanatory variables (wavelengths). Data in which the number of explanatory variables surpasses the number of observations is known as high-dimensional data. One method that can handle high-dimensional problems is to select important variables that affect the objective variable. The XGBoost algorithm can calculate the feature importance score that affects the goal variable so that it does not have to include all variables in the modeling, this can overcome problems in high-dimensional data. The results of the calculation of feature importance found Lignin Skeletal Band, CH, and CH2 aliphatic Stretching Group, C=C, C=N, C–H in ring structure, DNA and RNA backbones, NH2 Aminoacidic Group, C=O Ester Fatty Acid that the active compounds contained in the leaves of sembung.