Implementation of XGBoost Algorithm for Sentiment Classification of Public Opinions on the Rupiah Redenomination Policy

Abstract

Although rupiah redenomination has long featured in Bank Indonesia’s monetary policy discourse as a means of simplifying currency denominations without altering the exchange rate or real purchasing power, public concerns about declining purchasing power and price rounding underscore the need for effective policy communication. This study analyzes Indonesian public sentiment toward rupiah redenomination using the Extreme Gradient Boosting (XGBoost) algorithm to classify YouTube comments. Data were collected through the YouTube Data API v3 from 1,763 comments posted on the Tribunnews video titled Purbaya Targets Rupiah Redenomination Bill to Be Completed in 2027. Following text cleaning and tokenization, 1,169 comments were retained and transformed using Term Frequency–Inverse Document Frequency (TF-IDF). Manual labeling classified 657 comments as positive and 512 as negative. The XGBoost model, trained using optimized hyperparameters, achieved an accuracy of 73.39%, with F1-scores of 0.772 for positive sentiment and 0.680 for negative sentiment. These results indicate that the model classified positive sentiment more effectively, although ambiguous comments remained challenging. The findings reveal the distribution of public responses to the proposed policy and emphasize the need for intensive public outreach to mitigate potential resistance and misconceptions. This study contributes a data-driven basis for developing more effective monetary policy communication and anticipating the social dynamics associated with rupiah redenomination.