Pengembangan Metode Seleksi Fitur Dan Reduksi Data Berbasis Feature Importance Ranking Dan Local Outlier Factor Untuk Meningkatkan Performa Deteksi Pada Jaringan

Megantara, Achmad Akbar (2021) Pengembangan Metode Seleksi Fitur Dan Reduksi Data Berbasis Feature Importance Ranking Dan Local Outlier Factor Untuk Meningkatkan Performa Deteksi Pada Jaringan. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
05111950010039-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 April 2023.

Download (4MB) | Request a copy

Abstract

Tingkat keberhasilan proses machine learning untuk Intrusion Detection System (IDS) ditentukan oleh kualitas model yang dihasilkan dari proses training data. Training data yang bagus dapat diperoleh dengan melakukan pre-processing data pada dataset, seperti melakukan ekstraksi fitur, reduksi fitur maupun transformasi fitur. Namun, dalam penelitian tentang IDS proses pemilihan fitur mana yang digunakan dan yang tidak digunakan menjadi masalah utama yang terjadi karena akan mempengaruhi performa dari sistem dan waktu komputasi yang dibutuhkan untuk memproses data dari setiap intrusi. Penelitian ini bertujuan untuk membangun metode pendekatan seleksi fitur dan reduksi data baru dengan mengkombinasikan metode Feature Importance Ranking dan Local Outlier Factor (LOF). Feature Importance Ranking digunakan untuk menghitung tingkat kepentingan dari masing – masing fitur dan LOF digunakan untuk mendeteksi adanya outlier data pada dataset dan menghilangkannya. Setelah melewati tahap pre-processing, proses klasifikasi akan dilakukan untuk menguji hasil dari training data yang dihasilkan. Dengan menggunakan metode tersebut diharapkan dapat meningkatkan nilai accuracy, sensitivity, specificity, dapat menurunkan false alarm rate dari sistem IDS dan dapat menurunkan waktu komputasi dari program, sehingga dapat meningkatkan performa dari IDS. ============================================================ ======================================= The successfulness of the machine learning process for Intrusion Detection System (IDS) is determined by the quality of the model generated from training data process. Good training data can be obtained by performing pre-processing data in the dataset, such as feature extraction, feature reduction and feature transformation. However, the method used to specify which selected features will be used become main problem in IDS research topic. A good feature selection process will affect IDS performance and computational time needed for detecting some intrusion. In this research, we work on that problem by developing a new feature selection and data reduction approach by combining Feature Importance Ranking method with Local Outlier Factor (LOF). Feature Importan Ranking method is used to calculate the importance value of each feature or attribute in the dataset and LOF is used to detect the outlier data in the dataset and remove it. After pre-processing step is passed, classification process will be carried out to test the training data. By using this method, it is expected to increase the accuracy, sensitivity, and specificity values, reduce the false alarm rate of the IDS system and to reduce the computation time of the program, so the proposed method can be used to increase the performance of the IDS.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Intrusion Detection System (IDS), Feature Importance Ranking, Local Outlier Factor (LOF), Decision Tree, NSL-KDD, UNSW-NB15.
Subjects: T Technology > T Technology (General) > T55 Industrial Safety
T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T57.62 Simulation
T Technology > T Technology (General) > T58.62 Decision support systems
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User: Achmad Akbar Megantara
Date Deposited: 03 Mar 2021 01:28
Last Modified: 03 Mar 2021 01:28
URI: https://repository.its.ac.id/id/eprint/83204

Actions (login required)

View Item View Item