The development of Indonesia's new capital city, Ibu Kota Nusantara (IKN), has sparked extensive public discourse on social media, positioning sentiment analysis as a strategic approach to understanding public opinion. This study assesses the performance of the Naïve Bayes algorithm enhanced through Particle Swarm Optimization (PSO) in classifying public sentiment related to the IKN project, using Indonesian-language comments extracted from the social media platform X. The initial Naïve Bayes model achieved an accuracy of 78.3%, while the PSO-optimized model demonstrated an improved accuracy of 79.7% under optimal parameter settings. These findings indicate the potential of PSO to enhance feature selection effectiveness and reduce classification errors, particularly for positive sentiments. Despite the observed improvements, limitations such as reliance on automated sentiment labeling and challenges posed by linguistic context remain. This study contributes an early exploration of optimization-based methods for public opinion classification and highlights the need for further research involving advanced approaches such as deep learning tailored to the Indonesian language.