Sentiment Analysis of a 271 Trillion Rupiahs Corruption Case Using LSTM

Abstract

Corruption is one of the most pressing issues in Indonesia, significantly affecting public trust in governance and the nation’s development. Among the many corruption cases that have surfaced, the recent 271 trillion rupiah corruption case has drawn widespread attention and public discourse. Understanding the public's perception and sentiment regarding such cases can provide valuable insights into how these issues impact society. Researchers identified an opportunity to leverage sentiment analysis as a method to capture and analyze public sentiment in this context. The dataset for this study was collected from the social media platform Twitter (X) using a data crawling technique. Prior to analysis, preprocessing was performed to clean and prepare the data. After preprocessing, the data was categorized into three sentiment labels: negative, positive, and neutral. To perform sentiment classification, this study utilized the LSTM (Long Short-Term Memory) algorithm, a deep learning method particularly suited for sequential data analysis. The model was trained over a total of 10 epochs. The classification results demonstrated that the LSTM algorithm achieved an accuracy of 0.9365 at the 10th epoch, showcasing its effectiveness in analyzing public sentiment regarding 271 trillion rupiah corruption issues.