Performance Analysis of IndoBERT-Lite for Indonesian Hoax News Detection

DOI

10.2991/978-94-6239-828-3_21How to use a DOI?

Keywords

hoax news detection; deep learning; IndoBERT-lite; ALBERT; model evaluation

Abstract

The growth of internet access in Indonesia, reaching 72.78% in 2024 according to the Central Bureau of Statistics (BPS), has become one of the key factors behind the increasing spread of hoax news. The Ministry of Communication and Digital (Komdigi) recorded 12,547 hoax content cases from August 2018 to December 2023. However, manual verification requires substantial time and resources. This study aims to train and evaluate the IndoBERT-lite model based on the ALBERT architecture as an automated hoax news detection solution. The research encompasses the collection of 2,000 news articles, data labeling, pre-processing, tokenization using BertTokenizer, model training, and evaluation using a confusion matrix alongside analysis of memory usage, training time, and prediction speed. The results show that IndoBERT-lite achieved an accuracy and F1-score of 98%, equivalent to IndoBERT-base. IndoBERT-lite attained a prediction time of 104.73 ms per article, outperforming IndoBERT-base which required 128.29 ms. Training for IndoBERT-lite was completed in 813 seconds using 3.83 GB of RAM and 8.19 GB of VRAM, whereas IndoBERT-base required 1,199 seconds with 6.61 GB of RAM and 8.34 GB of VRAM. During prediction, IndoBERT-lite required 1.91 GB of RAM and 0.36 GB of VRAM, compared to IndoBERT-base which needed 2.31 GB of RAM and 2.11 GB of VRAM. Based on these findings, IndoBERT-lite represents an effective and efficient solution for automated Indonesian hoax news detection.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Dzikri Muhammad Raihan
AU  - Hariz Farisi
AU  - Buce Trias Hanggara
PY  - 2026
DA  - 2026/11/04
TI  - Performance Analysis of IndoBERT-Lite for Indonesian Hoax News Detection
BT  - Proceedings of the 2026 11th International Conference on Sustainable Information Engineering and Technology (SIET 2026)
PB  - Atlantis Press
SP  - 235
EP  - 245
SN  - 2589-4900
UR  - https://doi.org/10.2991/978-94-6239-828-3_21
DO  - 10.2991/978-94-6239-828-3_21
ID  - Raihan2026
ER  -