TY - GEN
T1 - Sentiment Analysis on YouTube Comment Data for the Candidate Debate in the 2024 Presidential Election of the Republic of Indonesia
AU - Purwandari, Kartika
AU - Jiwanggi, Meganingrum Arista
AU - Yulianti, Evi
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - The 2024 presidential election in Indonesia is just around the corner, and political figures are actively campaigning to increase their electability. Sentiment analysis has become a valuable tool in understanding public opinion, especially in the context of political campaigns. This study aims to analyze the sentiments expressed by YouTube comments towards the candidates during the presidential debates. We compared several machine learning (Support Vector Machine, Multinomial Linear Regression, Multinomial Naive Bayes, Random Forest) models along with TF-IDF, BoW, and Word2Vec feature techniques compare with deep learning models (IndoBERT and RoBERTa). From the experimental results, the best result is RoBERTa, with accuracy 83% and F1-score 81%. Evaluation of the best model results is applied to the remaining data and then used for the analysis process. According to our data, the result of the Indonesian presidential election in 2024 had a mean absolute error (MAE) of 6.92%. Based on the relatively low MAE, YouTube comments can potentially provide useful insights for predicting future elections considering a good agreement between the real count and our sentiment analysis.
AB - The 2024 presidential election in Indonesia is just around the corner, and political figures are actively campaigning to increase their electability. Sentiment analysis has become a valuable tool in understanding public opinion, especially in the context of political campaigns. This study aims to analyze the sentiments expressed by YouTube comments towards the candidates during the presidential debates. We compared several machine learning (Support Vector Machine, Multinomial Linear Regression, Multinomial Naive Bayes, Random Forest) models along with TF-IDF, BoW, and Word2Vec feature techniques compare with deep learning models (IndoBERT and RoBERTa). From the experimental results, the best result is RoBERTa, with accuracy 83% and F1-score 81%. Evaluation of the best model results is applied to the remaining data and then used for the analysis process. According to our data, the result of the Indonesian presidential election in 2024 had a mean absolute error (MAE) of 6.92%. Based on the relatively low MAE, YouTube comments can potentially provide useful insights for predicting future elections considering a good agreement between the real count and our sentiment analysis.
KW - deep learning
KW - feature engineering
KW - machine learning
KW - presidential election
KW - sentiment analysis
UR - http://www.scopus.com/inward/record.url?scp=85209656619&partnerID=8YFLogxK
U2 - 10.1109/AiDAS63860.2024.10730443
DO - 10.1109/AiDAS63860.2024.10730443
M3 - Conference contribution
AN - SCOPUS:85209656619
T3 - 2024 5th International Conference on Artificial Intelligence and Data Sciences, AiDAS 2024 - Proceedings
SP - 392
EP - 397
BT - 2024 5th International Conference on Artificial Intelligence and Data Sciences, AiDAS 2024 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 5th International Conference on Artificial Intelligence and Data Sciences, AiDAS 2024
Y2 - 3 September 2024 through 4 September 2024
ER -