Acute kidney injury (AKI) affects outcomes and mortality in trauma patients, and early recovery of AKI predicts a good prognosis. We aimed to develop an explainable machine learning model that predicts recovery from AKI within 24 hours among hospitalized trauma patients. We conducted a retrospective cohort study of 1,156 trauma patients who developed AKI between 2015 and 2024 at the Dankook University Trauma Center. AKI and severity were defined by KDIGO criteria using serum creatinine bases (Stage 1, n=663; Stage 2, n=218; Stage 3, n=275). We utilized clinical and laboratory time-series data from 909 patients (training set) to train an Extreme Gradient Boosting (XGBoost) machine learning model. The model was validated using a temporally distinct cohort of 247 patients. The primary outcome was AKI recovery at 24 hours, defined by improvement in eGFR. Predictive performance was evaluated using the area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity. Feature importance was assessed via time-series SHapley Additive exPlanations (SHAP) analysis. The XGBoost model predicted AKI recovery at 24 hours, achieving an AUROC of 0.8429. At the cut-off probability of 0.5, sensitivity was 74.6%, and specificity was 71.2% (figure 1). The time-series SHAP approach provided dynamic, interpretable visualizations, highlighting patient-specific clinical trajectories and influential recovery determinants (figure 2). Our study demonstrates that an explainable XGBoost machine learning model can reliably predict 24-hour AKI recovery in trauma patients using clinical data. Integrating explainable AI through time-series SHAP plots with the actual patient’s status enables clinicians to understand individualized predictions and identify modifiable clinical factors, enhancing targeted interventions and potentially improving patient outcomes.