Search for collections on Undip Repository

PERBANDINGAN KINERJA KLASIFIKASI SINYAL DRONE MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK (CNN) DAN LONG SHORT-TERM MEMORY (LSTM)

TAQIYUDDIN, Muhammad and Adi, Kusworo and Nurhayati, Oky Dwi (2023) PERBANDINGAN KINERJA KLASIFIKASI SINYAL DRONE MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK (CNN) DAN LONG SHORT-TERM MEMORY (LSTM). Masters thesis, UNIVERSITAS DIPONEGORO.

[img] Text
1. COVER.pdf

Download (1MB)
[img] Text
2. BAB I.pdf

Download (228kB)
[img] Text
3. BAB II.pdf

Download (960kB)
[img] Text
4. BAB III.pdf
Restricted to Repository staff only

Download (557kB)
[img] Text
5. BAB IV.pdf
Restricted to Repository staff only

Download (2MB)
[img] Text
6. BAB V.pdf

Download (230kB)
[img] Text
DAFTAR PUSTAKA.pdf

Download (231kB)

Abstract

Dalam usaha menangani penyusupan informasi yang dilakukan secara rahasia melalui penggunaan drone, penerapan kecerdasan buatan melalui pemodelan klasifikasi sinyal drone menggunakan paradigma pemrograman Deep Learning membantu mengidentifikasi drone sebelum terjadi aksi penyusupan. Sistem yang dikembangkan mengintegrasikan Deep Learning untuk mengenali pola dari data latih, memungkinkan klasifikasi data dengan tujuan yang ditentukan. Meskipun studi sebelumnya telah fokus pada penerapan Deep Learning dalam klasifikasi sinyal dan deteksi drone, beberapa aspek penelitian masih terbuka. Penelitian sebelumnya cenderung terbatas pada klasifikasi biner, yakni adanya drone atau tidak, yang berakibat pada pengelompokan sinyal serupa dengan sinyal drone. Penelitian ini mengatasi kekurangan tersebut dengan mengembangkan pengklasifikasian dalam empat kategori, meliputi drone, 5G, Wi-Fi 11ax, dan tanpa sinyal. Penelitian ini juga memperkaya variasi dataset dengan mengakomodasi jarak variasi sinyal drone, antara 10 hingga 60 meter dengan interval 5 meter. Metodologi Deep Learning, terutama Convolutional Neural Network (CNN) dan Long-Short Term Memory (LSTM), diadopsi dalam pemodelan ini sesuai dengan karakteristik data sinyal yang dihasilkan melalui Waveform Generator pada Matlab. Analisis menyeluruh dilakukan terhadap akurasi model yang dihasilkan oleh Deep Learning dengan mempertimbangkan kondisi variasi yang terimplementasi pada dataset. Hasil penelitian menunjukkan bahwa model yang dilatih menggunakan Convolutional Neural Network (CNN) menghasilkan kinerja terbaik, dengan rata-rata akurasi pengujian melebihi 99%, dibandingkan dengan model yang menggunakan Long Short-Term Memory (LSTM) yang mencapai akurasi pengujian di atas 92%.
Kata Kunci: Deep Learning, Klasifikasi Sinyal, Convolutional Neural Network (CNN), Long-Short Term Memory (LSTM), Drone

In an effort to address covert information infiltration through the use of drones, the application of artificial intelligence through modeling the classification of drone signals using the Deep Learning programming paradigm helps identify drones before the act of infiltration occurs. The developed system integrates Deep Learning to recognize patterns from the training data, enabling the classification of data with a defined purpose. Although previous studies have focused on the application of Deep Learning in signal classification and drone detection, several aspects of research remain open. Previous research tends to be limited to binary classification, i.e. the presence of drones or not, which results in the clustering of signals similar to drone signals. This research addresses this shortcoming by developing classifications in four categories, including drones, 5G, Wi-Fi 11ax, and no signal. This research also enriches the dataset variation by accommodating the distance variation of drone signals, between 10 to 60 meters with an interval of 5 meters. Deep Learning methodologies, especially Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM), were adopted in this modeling according to the characteristics of the signal data generated through the Waveform Generator in Matlab. A thorough analysis was conducted on the accuracy of the model generated by Deep Learning considering the variation conditions implemented in the dataset. The results showed that the model trained using Convolutional Neural Network (CNN) produced the best performance, with an average test accuracy exceeding 99%, compared to the model using Long Short-Term Memory (LSTM) which achieved a test accuracy above 92%.
Keywords: Deep Learning, Klasifikasi Sinyal, Convolutional Neural Network (CNN), Long-Short Term Memory (LSTM), Drone

Item Type: Thesis (Masters)
Uncontrolled Keywords: Deep Learning, Klasifikasi Sinyal, Convolutional Neural Network (CNN), Long-Short Term Memory (LSTM), Drone
Subjects: Sciences and Mathemathic
Divisions: Postgraduate Program > Master Program in Information System
Depositing User: ekana listianawati
Date Deposited: 23 Nov 2023 08:27
Last Modified: 23 Nov 2023 08:27
URI: https://eprints2.undip.ac.id/id/eprint/18279

Actions (login required)

View Item View Item