Penerapan Multimodel Deep Learning Dalam Pendeteksian Berita Hoaks Laman `Turnbackhoax.Id` Menggunakan Arsitektur CNN

Bayhaqi, Ahmad Rizal (2024) Penerapan Multimodel Deep Learning Dalam Pendeteksian Berita Hoaks Laman `Turnbackhoax.Id` Menggunakan Arsitektur CNN. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 2043201047-Undergraduate-Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (2MB) | Request a copy

Abstract

Meningkatnya penyebaran berita hoaks memicu kebutuhan publik untuk mengatasi permasalahan tersebut. Masyarakat Anti Fitnah Indonesia (Mafindo), sebagai komunitas independen, berupaya memberikan edukasi dan identifikasi terkait berita hoaks. Meski demikian, proses verifikasi berita yang masih manual dan tidak cukup efektif mengingat jumlah berita hoaks yang tersebar cepat dan dalam jumlah besar. Oleh karena itu, penelitian dilakukan untuk membangun model machine learning menggunakan deep learning, khususnya convolutional neural network (CNN), dalam mengklasifikasikan berita hoaks secara cepat dan otomatis. Penggunaan CNN berbasis data teks dan gambar telah menunjukkan performa klasifikasi yang baik, terutama ketika kedua data digabungkan dalam multimodel deep learning. Multimodel deep learning atau model CNN gabungan, menggabungkan model CNN berbasis teks (CNN 1D) dan gambar (CNN 2D) yang menunjukkan kinerja lebih baik dibandingkan dengan model tunggal (unimodel). Model tersebut kemudian dilatih dengan 3103 data training dan 775 data testing dan diperoleh nilai akurasi 99,35% dan AUC 99,81%. Hasil evaluasi menunjukkan kinerja yang baik dalam mengklasifikasikan berita hoaks di dalam dan di luar model.
==========================================================================================
The increasing spread of hoax news has prompted the public’s need to address this issue. The Indonesian Anti-Slander Society (Mafindo), as an independent community, strives to provide education and identification regarding hoax news. However, the manual verification process proves ineffective given the rapid and large-scale dissemination of hoax news. Therefore, research is conducted to build a machine learning model using deep learning, specifically convolutional neural network (CNN), for the fast and automatic classification of hoax news. The use of CNN based on text and image data has shown good classification performance, especially when both data types are combined in multimodal deep learning. Multimodal deep learning, or the combined CNN model, merges text-based CNN (CNN 1D) and image-based CNN (CNN 2D), demonstrating superior performance compared to a single model (unimodal). The model is then trained with 3103 training data and 775 testing data, resulting in an accuracy of 99.35% and an AUC of 99.81%. Evaluation results indicate excellent performance in classifying hoax news both within and outside the model.

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	CNN, Hoax, Multimodel Deep Learning
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Ahmad Rizal Bayhaqi
Date Deposited:	08 Aug 2024 06:41
Last Modified:	25 Sep 2024 03:20
URI:	http://repository.its.ac.id/id/eprint/113531

Actions (login required)

View Item