Sentiment Analysis of Presidential and Vice Presidential Candidates Using FastText CNN

Abstract

The Internet has become a platform for storing and accessing a wide range of public information. People tend to use social media to express opinions on products, political policies, and politicians, both as individuals and as parties. Sentiment analysis of presidential and vice-presidential candidates aims to understand public perspectives on each candidate. Based on sentiment analysis of the Indonesian public on social media, potential candidates for the presidential election, which is still over a year away, can be identified. This study categorizes sentiments into four classes: happy, love, sad, and angry. The model employs FastText embeddings with a Convolutional Neural Network (CNN). The best performance achieved was an F1-score of 0.9510. The findings indicate that Ganjar Pranowo leads as a presidential candidate, while Erick Thohir stands out as the leading vice-presidential candidate.