Analisis Sentimen Berbasis Aspek terhadap Pilihan Presiden Menggunakan Algoritma LSTM pada Sosial Media X

Abstract

Democracy serves as the fundamental foundation of Indonesia’s governmental system, where citizens play a significant role in determining their leaders through general elections. The election results decide who will lead the nation and represent the public interest. Many people express their opinions and perspectives regarding presidential candidates through the social media platform X (formerly known as Twitter). The abundance of public opinions surrounding the 2024 presidential election on X generates various sentiments—positive, negative, and neutral—that can be utilized as data sources for analysis. This study employs the Long Short-Term Memory (LSTM) and Latent Dirichlet Allocation (LDA) algorithms for topic extraction and sentiment classification of tweets into positive, negative, or neutral responses. The analyzed data consist of 30,000 tweets, equally divided among Anies Baswedan, Prabowo Subianto, and Ganjar Pranowo, collected between January and April 2023. The text preprocessing stages include normalization, tokenization, stopword removal, and stemming. After cleaning, the texts are labeled with sentiments using TextBlob, followed by data balancing with the SMOTE technique. The dataset is then split into training and testing data, with the training data used to build the sentiment analysis model. The implementation of the LSTM algorithm in this research successfully achieved an accuracy of 72.5%.