Implementasi Metode SVM Pada Sentimen Analisis Terhadap Pemilihan Presiden (Pilpres) 2024 Di Twitter

Abstrak

The focus of the research is the use of Twitter as a platform to express the political opinions of the Indonesian people regarding the 2024 Presidential Election. By utilizing sentiment analysis using the Support Vector Machine (SVM) method, this research aims to evaluate the accuracy of SVM in classifying tweets and compare the performance of four types of SVM kernels. Visualizations of positive and negative sentiments are also generated to provide a clearer picture. The stages of the research involve Twitter data collection, and pre-processing with steps such as data cleansing, case folding, tokenizing, stemming, and filtering. Labeling is done to identify sentiment, then feature extraction using TF-IDF. SVM implementation with linear, polynomial, RBF, and sigmoid kernels is performed, followed by model evaluation using precision, recall, F-measure, and accuracy metrics. The study used SVM to analyze the sentiment of the 2024 presidential election on Twitter data. As a result, out of 3938 tweets, 1575 were positive and 2363 were negative. The SVM model achieved 95.05% accuracy, superior in predicting negative sentiment. Comparison of SVM kernels shows the highest accuracy in the linear kernel 95.43%. Sentiment analysis on tweets shows a majority of positive support for Ganjar 54.9%, while Anies and Prabowo have support levels of 15.8% and 29.3% respectively.