The Implementation of TF-IDF and Word2Vec on Booster Vaccine Sentiment Analysis Using Support Vector Machine Algorithm
Candra Agustina, Rice Novita, Mustakim Mustakim, Nesdi Evrilyan Rozanda
Abstract
As a sort of technological advancement, social media is a medium used to transmit ideas on certain subjects. Sentiment analysis can be used to analyze public opinion. Feature extraction stage of sentiment analysis is crucial for transforming unstructured text into categorizable structured data. Using 13,297 records from Twitter and SVM algorithm, as well as the TF-IDF and Word2Vec feature extraction approaches, the combination of SVM + TF-IDF with 80:20 data split scenario and the RBF kernel produces the best results, with precision 85%, recall 86%, and f1-score 84%. In the 80:20 data split and RBF kernel, SVM+Word2Vec combination achieves the highest performance, with precision 83%, recall 82%, and f1-score 76%.