Patient appointment no-shows constitute a significant challenge in healthcare institutions, leading to a substantial waste of medical service resources and a deterioration of operational efficiency. Consequently, accurately predicting patient visit propensity following initial appointment absenteeism holds considerable practical significance for effective healthcare management. This study leverages longitudinal data from the Korea Association of Health Promotion, spanning 2020–2022, to predict the likelihood of patient visitation versus non-visitation subsequent to appointment defaults. Crucially, the research employs eXplainable Artificial Intelligence (XAI) methodologies to elucidate key contributory factors. A comparative analysis was conducted across multiple machine learning and ensemble modeling approaches, specifically Random Forest, Decision Trees, and LightGBM algorithms. The empirical findings demonstrate that LightGBM exhibited superior predictive performance for visit prediction, whereas Random Forest achieved optimal efficacy for non-visitation forecasting. Application of the SHapley Additive exPlanations (SHAP) methodology revealed that historical rescheduling behavior and new patient status emerged as pivotal determinants across both predictive models. Furthermore, Local Interpretable Model-agnostic Explanations (LIME) techniques were employed to provide a visual interpretation of variable-specific influence directionality and feature importance magnitudes. This investigation contributes significantly to the advancement of healthcare appointment management systems and provides actionable insights for efficient medical resource allocation and personalized patient engagement strategies through predictive modeling of post- default patient behavioral patterns.