relation: http://repository.mercubuana.ac.id/96609/
title: PENERAPAN MULTIMODAL LATE FUSION UNTUK IDENTIFIKASI SITUS WEB JUDI ONLINE DI INDONESIA
creator: RISKI, AZZAN DWI
subject: 004 Data Processing, Computer Science/Pemrosesan Data, Ilmu Komputer, Teknik Informatika
subject: 006.3 Artificial Intelligence/Kecerdasan Buatan
subject: 006.75 Social Multimedia/Multimedia Social
description: Online gambling has become a serious issue in Indonesia over the past few  years. Although the government has blocked millions of websites, new domains  continue to appear, making automated detection highly necessary. This study  proposes a Multimodal Late Fusion approach to detect online gambling websites  in Indonesia by combining visual and textual information from web pages. The  image classification model is built using EfficientNet-B3, while the text  classification model uses IndoBERT. Both models are trained independently, and  its prediction outputs are combined using a late fusion strategy. Experimental  results show that the text-based model (IndoBERT) outperforms the image-based  model. However, the fusion of both models achieves the best overall performance,  with accuracy, precision, recall, and F1-score all reaching 99.71%. Although the  fusion model delivers the highest performance, it also requires longer inference  time compared to single-modality models. Therefore, implementing this system in  real-world scenarios must consider the trade-off between accuracy and efficiency.  Keywords: online gambling, multimodal late fusion, EfficientNet-B3, IndoBERT, website detection    Perjudian online di Indonesia telah menjadi permasalahan serius selama  beberapa tahun terakhir. Meskipun pemerintah telah melakukan pemblokiran jutaan  situs, munculnya domain baru secara terus-menerus membuat deteksi otomatis  menjadi sangat dibutuhkan. Penelitian ini mengusulkan pendekatan Multimodal  Late Fusion untuk mendeteksi situs judi online di Indonesia berdasarkan kombinasi  informasi visual dan tekstual dari halaman web. Model klasifikasi gambar dibangun  menggunakan EfficientNet-B3, sementara model klasifikasi teks menggunakan  IndoBERT. Kedua model dilatih secara independen, kemudian hasil prediksi  digabungkan menggunakan strategi late fusion. Eksperimen menunjukkan bahwa  model teks berbasis IndoBERT memberikan performa lebih baik dibandingkan  model gambar. Namun, kombinasi keduanya melalui model fusion mampu  menghasilkan performa tertinggi, dengan accuracy, precision, recall dan F1-score  masing-masing sebesar 99.71%. Meskipun model fusion memberikan performa  terbaik, waktu inferensinya relatif lebih tinggi dibandingkan model tunggal. Oleh  karena itu, implementasi sistem ini dalam lingkungan nyata perlu  mempertimbangkan keseimbangan antara akurasi dan efisiensi.  Kata kunci: judi online, multimodal late fusion, EfficienNet-B3, IndoBERT, deteksi situs web
date: 2025-07-21
type: Thesis
type: NonPeerReviewed
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/1/01%20COVER.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/2/02%20BAB%201.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/3/03%20BAB%202.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/4/04%20BAB%203.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/5/05%20BAB%204.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/6/06%20BAB%205.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/7/07%20DAFTAR%20PUSTAKA.pdf
format: text
language: id
identifier: http://repository.mercubuana.ac.id/96609/8/08%20LAMPIRAN.pdf
identifier:   RISKI, AZZAN DWI  (2025) PENERAPAN MULTIMODAL LATE FUSION UNTUK IDENTIFIKASI SITUS WEB JUDI ONLINE DI INDONESIA.  S1 thesis, Universitas Mercu Buana Jakarta.