The global rollout of COVID-19 vaccines has marked a critical milestone in combating the pandemic. Despite widespread availability, vaccine hesitancy and misinformation continue to challenge public health efforts. Understanding public sentiment towards COVID-19 vaccination is vital to designing effective communication strategies that promote vaccine acceptance. Previous research on vaccine sentiment has largely focused on static snapshots or limited datasets, often neglecting the dynamic and multifaceted nature of public opinions across demographics and regions. Moreover, many studies employ basic classification techniques without exploring advanced feature selection or balancing methods to optimize model performance. This paper aims to bridge these gaps by conducting a comprehensive sentiment analysis on a large-scale Twitter dataset, “COVID-19 Vaccine Tweets,” containing over 375,000 unique entries. We apply robust data preprocessing and feature selection methods, including Variance Threshold and SelectKBest, to optimize input features for classification. Using logistic regression as a baseline classifier, we examine sentiment polarity (positive, neutral, negative) and explore the influence of follower count and geographic distribution on public perception. Our study provides deeper insights into evolving vaccine-related discourse and lays the groundwork for more targeted, data-driven public health interventions.