Pemodelan Identifikasi Trafik Bittorrent Dengan Pendekatan Correlation Based Feature Selection (CFS) Menggunakan Algoritme Decision Tree (C4.5)

Abstract

Abstract— BitTorrent is a P2P file sharing software protocol that allows clients to apply data to other clients and can affect network performance. Bittorent client traffic data collection uses secondary data taken from official sources on the link https://unb.ca/cic/datasets/index.html in 2016. Traffic data is used as a model for BitTorrent traffic identification using feature-based correlation selection (CFS) and traffic analysis model analysis using Decision Tree Algorithm (C4.5). Feature selection is done to clean irrelevant features so that they can affect the results of the accuracy value. The results of feature selection obtained 7 features and 1 category with 244,689 records and the system connecting the rule tree data training model selected the four best accuracy values. Furthermore, the model training data is carried out by testing the BitTorrent traffic trial data. The results of data testing obtained the best BitTorrent traffic accuracy value of 98.82% with 73,406 records on the 30% data test. Keywords— BitTorrent, C4.5 algorithm, correlation based feature selection, traffic identification, modeling.