Analisis Sentimen Media Sosial X Terhadap Isu Tapera Menggunakan Metode XGBoost Berbasis PSO.

This study aims to analyze public sentiment towards the issue of People's
Savings (Tapera) using data from X social media. Data was collected by crawling
on social media X using the keyword “Tapera Issue” and obtaining 2,268 data. After
the data is collected, a manual data labeling process is carried out to determine the
sentiment categories, namely negative, positive, neutral and others. Followed by the
data pre-processing stage. Next, feature extraction is carried out using the Term
Frequency-Inverse Document Frequency (TF-IDF) method, then the data is divided
into training data and test data with a ratio of 80:20. The algorithm used in this
research is Extreme Gradient Boosting (XGBoost) which is then optimized using
Particle Swarm Optimization (PSO) to get the best parameters that can improve
model accuracy. The results showed that the application of PSO-based XGBoost
resulted in an accuracy of 85%, higher than the accuracy of the XGBoost model
before optimization. This result shows that the combination of XGBoost and PSO
methods is effective in classifying public sentiment on the Tapera issue, and shows
its potential application in social media-based public opinion analysis.