Hybrid IndoBERT–BiLSTM with Attention Mechanism for Sentiment Analysis in Imbalanced Digital Society Data

Sentiment analysis plays an important role in understanding public opinion within digital society, particularly from social media data that is often noisy and imbalanced. This study proposes a hybrid deep learning model that integrates IndoBERT, Bidirectional Long Short-Term Memory (BiLSTM), and an attention mechanism to improve sentiment classification performance. IndoBERT is utilized to capture contextual semantic representations, while BiLSTM models sequential dependencies in text data. The attention mechanism further enhances the model by focusing on important features within sequences. To address class imbalance, a weighted cross-entropy loss is applied during training. The model is evaluated using accuracy, precision, recall, and F1-score, along with confusion matrix and visualization analysis. Experimental results show that the proposed model achieves high overall performance, with an accuracy of 91.16% and a weighted F1-score of 0.9005. However, class-wise evaluation reveals limitations in handling minority classes, particularly the negative class, due to severe data imbalance. These findings highlight that while hybrid architectures can significantly improve performance, additional strategies are required to ensure balanced classification. This study contributes to the development of robust sentiment analysis models for digital society applications.

Authors

Institutions

Publication Details

Journal
Journal of Digital Society
Published
2026-11-17
DOI
https://doi.org/10.63913/jds.v2i4.45
Primary Topic
Sentiment Analysis and Opinion Mining
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Hybrid IndoBERT–BiLSTM with Attention Mechanism for Sentiment Analysis in Imbalanced Digital Society Data

Aris Budiasono
Journal of Digital Society
Sentiment Analysis and Opinion Mining
article

Hybrid IndoBERT–BiLSTM with Attention Mechanism for Sentiment Analysis in Imbalanced Digital Society Data

Aris Budiasono
article en

Abstract

Sentiment analysis plays an important role in understanding public opinion within digital society, particularly from social media data that is often noisy and imbalanced. This study proposes a hybrid deep learning model that integrates IndoBERT, Bidirectional Long Short-Term Memory (BiLSTM), and an attention mechanism to improve sentiment classification performance. IndoBERT is utilized to capture contextual semantic representations, while BiLSTM models sequential dependencies in text data. The attention mechanism further enhances the model by focusing on important features within sequences. To address class imbalance, a weighted cross-entropy loss is applied during training. The model is evaluated using accuracy, precision, recall, and F1-score, along with confusion matrix and visualization analysis. Experimental results show that the proposed model achieves high overall performance, with an accuracy of 91.16% and a weighted F1-score of 0.9005. However, class-wise evaluation reveals limitations in handling minority classes, particularly the negative class, due to severe data imbalance. These findings highlight that while hybrid architectures can significantly improve performance, additional strategies are required to ensure balanced classification. This study contributes to the development of robust sentiment analysis models for digital society applications.

Journal of Digital SocietyVol. 2(4)
Universitas Amikom Purwokerto (ID)
Sentiment Analysis and Opinion Mining
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon