Word2vec Architecture in Sentiment Classification of Fuel Price Increase Using CNN-BiLSTM Method

Abstract

Fuel price increases have been a frequent problem in recent years, mainly due to unstable international price fluctuations. This research uses sentiment analysis to examine the increase in fuel prices and its impact on public sentiment. Sentiment analysis is a data processing method to obtain information about an issue by recognizing and extracting emotions or opinions from existing text. In this research, two architectures, Word2vec Continous Bag of Words (CBOW) and Skip-gram, were tested with different vector dimension sizes in each architecture using the CNN-BiLSTM hybrid deep learning method. The results showed that the CBOW model with 300 vector dimensions produced the best performance with 87% accuracy, 87% Recall, 89% Precision, and 88% F1-score on the tested Indonesian language dataset.