Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms

This study presents a comparative analysis of Support Vector Machine (SVM) and Bidirectional Encoder Representations from Transformers (BERT) for sentiment analysis on electric motorcycles in Indonesia using data from the social media platform X, formerly known as Twitter. The dataset of 128,711 twe...

Full description

Saved in:
Bibliographic Details
Main Authors: Calvin Adiwinata, Afiyati Afiyati
Format: Article
Language:Indonesian
Published: Universitas Muhammadiyah Purwokerto 2025-08-01
Series:Jurnal Informatika
Subjects:
Online Access:http://jurnalnasional.ump.ac.id/index.php/JUITA/article/view/26152
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849392952886951936
author Calvin Adiwinata
Afiyati Afiyati
author_facet Calvin Adiwinata
Afiyati Afiyati
author_sort Calvin Adiwinata
collection DOAJ
description This study presents a comparative analysis of Support Vector Machine (SVM) and Bidirectional Encoder Representations from Transformers (BERT) for sentiment analysis on electric motorcycles in Indonesia using data from the social media platform X, formerly known as Twitter. The dataset of 128,711 tweets collected between 2015 and 2024 was refined through systematic preprocessing, reducing the corpus to 38,954 entries after data cleaning, tokenization, and feature selection. The objective was to evaluate algorithm performance in classifying public sentiment, with metrics including accuracy, precision, recall, and computational efficiency. Results showed that SVM achieved higher overall accuracy 89.74% with strong precision for positive sentiment 91%, while BERT, specifically the IndoBERT variant, demonstrated superior recall for negative sentiment 91% despite slightly lower accuracy 87.90%, effectively capturing nuanced contextual language, such as sarcasm, informal expressions, and emotionally ambiguous statements that require deeper semantic understanding beyond literal word meanings. Computational analysis revealed that SVM required approximately 53 minutes of CPU training, compared to BERT’s 3.3 hours on GPU. The study suggests that SVM is optimal for rapid, resource-constrained applications, whereas BERT excels in detailed contextual analysis. These findings guide stakeholders in selecting algorithms based on analytical priorities, such as monitoring public reception or addressing consumer concerns
format Article
id doaj-art-eefca152fd45432d9d89c8e50cc81123
institution Kabale University
issn 2086-9398
2579-8901
language Indonesian
publishDate 2025-08-01
publisher Universitas Muhammadiyah Purwokerto
record_format Article
series Jurnal Informatika
spelling doaj-art-eefca152fd45432d9d89c8e50cc811232025-08-20T03:40:37ZindUniversitas Muhammadiyah PurwokertoJurnal Informatika2086-93982579-89012025-08-0111912610.30595/juita.v13i2.2615221154Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT AlgorithmsCalvin Adiwinata0Afiyati Afiyati1Universitas Mercu BuanaUniversitas Mercu BuanaThis study presents a comparative analysis of Support Vector Machine (SVM) and Bidirectional Encoder Representations from Transformers (BERT) for sentiment analysis on electric motorcycles in Indonesia using data from the social media platform X, formerly known as Twitter. The dataset of 128,711 tweets collected between 2015 and 2024 was refined through systematic preprocessing, reducing the corpus to 38,954 entries after data cleaning, tokenization, and feature selection. The objective was to evaluate algorithm performance in classifying public sentiment, with metrics including accuracy, precision, recall, and computational efficiency. Results showed that SVM achieved higher overall accuracy 89.74% with strong precision for positive sentiment 91%, while BERT, specifically the IndoBERT variant, demonstrated superior recall for negative sentiment 91% despite slightly lower accuracy 87.90%, effectively capturing nuanced contextual language, such as sarcasm, informal expressions, and emotionally ambiguous statements that require deeper semantic understanding beyond literal word meanings. Computational analysis revealed that SVM required approximately 53 minutes of CPU training, compared to BERT’s 3.3 hours on GPU. The study suggests that SVM is optimal for rapid, resource-constrained applications, whereas BERT excels in detailed contextual analysis. These findings guide stakeholders in selecting algorithms based on analytical priorities, such as monitoring public reception or addressing consumer concernshttp://jurnalnasional.ump.ac.id/index.php/JUITA/article/view/26152electric motorcyclessentiment analysissvmbertsocial media
spellingShingle Calvin Adiwinata
Afiyati Afiyati
Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms
Jurnal Informatika
electric motorcycles
sentiment analysis
svm
bert
social media
title Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms
title_full Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms
title_fullStr Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms
title_full_unstemmed Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms
title_short Sentiment Analysis of X Users Toward Electric Motorcycles Using SVM and BERT Algorithms
title_sort sentiment analysis of x users toward electric motorcycles using svm and bert algorithms
topic electric motorcycles
sentiment analysis
svm
bert
social media
url http://jurnalnasional.ump.ac.id/index.php/JUITA/article/view/26152
work_keys_str_mv AT calvinadiwinata sentimentanalysisofxuserstowardelectricmotorcyclesusingsvmandbertalgorithms
AT afiyatiafiyati sentimentanalysisofxuserstowardelectricmotorcyclesusingsvmandbertalgorithms