Pemodelan Biplot pada Klasifikasi Fragmen Metagenom dengan K-Mers Sebagai Ekstraksi Ciri Dan Probabilistic Neural Network Sebagai Classifier

Abstract

Metagenomics is the study of the genetic material taken directly from the organism's source environment by performing deoxyribonucleic acid sequencing procedure. Sequencing was performed on a set of genomes that have been mixed with a variety of other organisms in the environment. This led to the identification of the organism is required to prevent fragment assembly error between organisms. This study objectives to classify metagenome fragments at the genus level with Probabilistic Neural Network algorithms and k-mers as the features extraction. Biplot modelling is used to reduce features of the k-mers results that have dimensions large enough. This study uses 3 genuses of microorganisms which are divided into 2 dataset groups, namely known organisms and unknown organisms. The length of fragments used for each dataset is 500 bp, 1 kbp, 5 kbp, and 10 kbp. From the results of this study, the best accuracy for the known organisms was 98.10% and 92.80% for unknown organisms. Biplot modelling managed to reduce 68.75% fragment features with classification efficiency of 12%.