Klasifikasi Fragmen Metagenome Menggunakan Metode Support Vector Machine (SVM)

Abstract

Metagenome analysis is one of the most important bioinformatics field. This field is related to genome which is taken directly from the environment. The purpose of this research is to classify metagenome fragment into some taxonomic levels using support vector machine (SVM) method. Feature extraction is performed using spaced k-mers. Classification process is conducted by creating model using the training data from 381 organisms. The evaluation results show that the accuracies for short fragments (400 bp) are 65.3% and 82.1% at genus level and phylum level, respectively. Meanwhile, the accuracies increase significantly for long fragments (10 kbp), with a value of 95.4% at genus level and 97.6% at phylum level. It can be stated that the accuracy will be increased with the increasing of fragments length and higher taxonomic levels. In addition, the results of the study also conclude that the feature extraction methods used was very good and produce data with linearly separable conditions.