Article Per Year (5 Year)
p-Index From 2021 - 2026
0.408
P-Index
This Author published in this journals
All Journal Journal of Vocational, Informatics and Computer Education
Irene R.H.T Tangkawarow
Universitas Negeri Manado
Computer Science & IT Education

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Comparison of the Performance of Support Vector Machines and Naïve Bayes in Analyzing Sentiment Regarding Money Politics in the 2024 General Election Olgraciella Manitik; Irene R.H.T Tangkawarow; Alfiansyah Hasibuan
Journal of Vocational, Informatics and Computer Education Vol 4, No 2 (2026): June 2026
Publisher : Academic Bright Collaboration

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.66053/voice.v4i2.494

Abstract

Purpose – This study aims to compare the performance of the Support Vector Machine (SVM) and Naïve Bayes algorithms in classifying public sentiment on money politics in the 2024 General Election in Indonesia, as well as to determine which algorithm has the highest accuracy, precision, recall, and F1-score. Methods – A total of 1,280 Indonesian-language tweets from the X platform containing the keyword “money politics” were collected and preprocessed. Of these, 384 tweets were manually labeled by two annotators (Cohen’s Kappa = 0.9493). TF-IDF and SMOTE feature extraction were applied before splitting the data into training (70%) and testing (30%) sets. Model evaluation used accuracy, precision, recall, F1-score, ROC-AUC, paired t-test, and bootstrap confidence intervals. Findings – Naïve Bayes (α=0.1) achieved accuracy 79.31%, precision 70.73%, recall 70.73%, F1-score 0.7073, ROC-AUC 0.8387. SVM (C=1, kernel RBF) achieved 77.59%, 85.71%, 43.90%, 0.5806, 0.8410. Paired t-test (p=0.8326) showed comparable performance across 10-fold CV. Naïve Bayes demonstrated better stability (bootstrap CI [0.5974–0.8101] vs. SVM [0.4193–0.7294]). Research implications – The methodological framework (preprocessing, SMOTE, TF-IDF, tuning, 10-fold CV, paired t-test, bootstrap CI) serves as a reference for election-related text classification in Indonesia. Results offer initial insights for election monitoring agencies (e.g., Bawaslu). Originality – This study provides a comparison of SVM and Naïve Bayes specifically for detecting money politics sentiment in the 2024 Indonesian election, a topic with limited prior research. It applies structured pipeline including manual labeling (high inter-annotator agreement), SMOTE, and statistical validation (paired t-test, bootstrap CI).
Implementation of the Multinomial Naïve Bayes Algorithm in a Web-Based System for Detecting Online News Hoaxes During the 2024 Elections Juliawati Haribae; Irene R.H.T Tangkawarow; Gladly C. Rorimpandey
Journal of Vocational, Informatics and Computer Education Vol 4, No 3 (2026): September 2026
Publisher : Academic Bright Collaboration

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.66053/voice.v4i3.608

Abstract

Purpose – This study addresses the escalation of disinformation during the 2024 Indonesian General Election by developing an automated hoax detection system. The primary focus is to evaluate the integration of data balancing methods to minimize detection failures in hoax narratives, which often appear less frequently than factual news in real-world scenarios. Methods – The dataset consists of 1,529 unique news documents, comprising 1,020 factual articles from Kompas.com and 509 hoax articles from TurnBackHoax.id. The modelling workflow involves text preprocessing, feature extraction via Term Frequency-Inverse Document Frequency (TF-IDF), and the application of the Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalance. The model was developed using the Multinomial Naive Bayes algorithm and integrated into a web-based platform using the Flask framework. Findings – Results from 5-fold cross-validation demonstrated a stable average accuracy of 92.64% (±1.27%). On an independent test set, the model achieved 89.54% accuracy with a hoax class recall of 0.93. This proves that SMOTE significantly enhances sensitivity in identifying disinformation, reducing the risk of false information bypassing the system. Research implications – The study is limited to textual data from curated sources and cannot yet capture multimedia disinformation or high levels of sarcasm. Future research should explore Transformer-based models for deeper semantic context understanding. Originality – This research provides a functional public verification platform with a response time of under one second. Its novelty lies in a domain-specific detection system that balances algorithmic efficiency with precision on imbalanced real-world election data.
Co-Authors Alfiansyah Hasibuan Gladly Caren Rorimpandey Juliawati Haribae Olgraciella Manitik