FastText-Enhanced Hybrid Deep Learning with Genetic Algorithm for Sentiment Analysis of Indonesia’s 2024 Regional Election on X

The 2024 Regional Elections in Indonesia have ignited vigorous public discourse, rendering sentiment analysis an essential instrument for comprehending voter behavior, candidate popularity, and campaign plans. This research utilizes a novel methodology, incorporating a hybrid model of Convolutional Neural Networks (CNN) and Bidirectional Gated Recurrent Units (BiGRU), enhanced by Genetic Algorithms, to assess public mood. This research also investigates the use of FastText features to increase sentiment classification accuracy. The dataset includes 60,000 Indonesian tweets collected using keywords linked to the 2024 Regional Election. In this research, CNN is utilized to extract spatial features, BiGRU to capture temporal dynamics, and FastText and Term Frequency-Inverse Document Frequency (TF-IDF) to represent features, all of which are GA optimized. The experimental results demonstrate that GA optimization has a considerable effect on model performance. The CNN-BiGRU + GA models had the highest accuracy of 80.72 percent in the Top 10, representing a 1.22 percent improvement over the base model. Meanwhile, the BiGRU-CNN + GA models obtained 80.69 percent accuracy in the Top 5, representing a 1.49 percent improvement over the base model. These findings not only demonstrate the immense potential of hybrid methodologies in sentiment analysis, but also present a more in-depth and relevant approach to investigating public opinion surrounding the 2024 Regional Election. This research is expected to provide fresh and more precise insights into the political dynamics of Indonesian society during the 2024 Regional Elections.