Article Per Year (5 Year)
p-Index From 2020 - 2025
1.503
P-Index
This Author published in this journals
All Journal JOIV : International Journal on Informatics Visualization International Journal of Artificial Intelligence Research JURNAL MEDIA INFORMATIKA BUDIDARMA Majalah Farmaseutik Journal of Information Systems and Informatics Jurnal Penelitian Farmasi dan Herbal Jurnal INFOTEL
Azka Khoirunnisa
Telkom University, Indonesia
Computer Science & IT Control & Systems Engineering Electrical & Electronics Engineering Medicine & Pharmacology

Published : 8 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : International Journal of Artificial Intelligence Research

 1 
An Evaluation of SMOTE Effectiveness in Handling Class Imbalance in Public Opinion Data on the MBG Program Ramadhan, Nur Ghaniaviyanto; Khoirunnisa, Azka
International Journal of Artificial Intelligence Research Vol 9, No 1.1 (2025)
Publisher : Universitas Dharma Wacana

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29099/ijair.v9i1.1.1495

Abstract

The “Makan Bergizi Gratis” (MBG) Program is one of the strategic policies of the Government of Indonesia that reaps various opinions from the public, especially through social media. This study aims to classify public sentiment towards the MBG program with an ensemble learning-based machine learning approach, as well as evaluate the effectiveness of the SMOTE algorithm in dealing with class imbalance in opinion data. The dataset was collected from platform X (formerly Twitter) for the January–April 2025 period, totaling 4,374 tweets with label distributions: 1,783 positive, 1,634 negative, and 957 neutral. The preprocessing process includes data cleansing, normalization, stemming, and vectorization with TF-IDF. Five ensemble algorithms were used, namely Random Forest, AdaBoost, Bagging, Stacking, and Voting, tested in two scenarios: with and without the implementation of SMOTE. The results of the experiments showed that Random Forest provided the best and most consistent performance, with the F1-score increasing from 72.03% to 72.66% after the implementation of SMOTE. However, not all models benefit from SMOTE, such as Voting which experienced a drop in F1-score. These findings suggest that SMOTE is effective in increasing the sensitivity of the model to minority classes, but its success depends on the characteristics of the algorithm used. This study suggests the selective selection of balancing methods as well as the development of a more adaptive approach to handle unstructured opinion data.
Co-Authors Adiwijaya, - Ahmad Shobrun Jamil Didit Adytia Hashimoto, Takako Kurnianingsih Kurnianingsih, Kurnianingsih M. Artabah Muchlisin Muchlisin, M. Artabah Nur Ghaniaviyanto Ramadhan