Public Sentiment Analysis Toward the Ministry of Finance 2025 Using Recurrent Neural Network Methods Based on Data from Sosial Media X

Abstract

The Ministry of Finance plays a strategic role in maintaining national economic stability through fiscal policy management, taxation, public debt administration, and state budget control. In today’s digital era, social media platforms such as X have become important channels for the public to express opinions about government policies. This study analyzes public perceptions of the Ministry of Finance’s performance using machine-learning-based sentiment analysis and identifies the most effective classification model. Data were collected from public posts on X and processed using text mining and Natural Language Processing (NLP). Three Recurrent Neural Network (RNN) models were tested: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and an improved variant, LSTM_G. The findings show that negative sentiment dominates at 43.0%, followed by neutral at 33.9% and positive at 23.1%. Among the models, LSTM_G achieved the highest accuracy of 78.98%, indicating strong capability in capturing sequential patterns in dynamic, unstructured social media text. These results reflect substantial public concerns regarding fiscal policies and demonstrate the usefulness of sentiment analysis as a data-driven tool for decision-making and for strengthening public communication strategies to enhance the Ministry’s digital reputation.