Background: Acute appendicitis poses diagnostic challenges due to symptom overlap with other abdominal conditions, often leading to misdiagnosis or missed diagnosis. This study aimed to develop and validate an interpretable machine learning model based on routine hematological indicators to facilitate rapid diagnosis. Methods: A retrospective analysis was conducted on 408 patients with acute abdominal pain, including both adult and pediatric patients. The median age of patients in the appendicitis group was 37.5 years (IQR: 26.5 years). Univariate logistic regression revealed significant group differences in hematological indicators (all P   <  0.001). Three feature selection methods—LASSO, ElasticNet, and Random Forest—were applied, with neutrophil percentage (NE%) and eosinophil percentage (EO%) consistently identified across all methods, and red blood cell (RBC) and white blood cell (WBC) repeatedly selected by at least two methods. Eleven commonly used machine learning classifiers were developed and evaluated on an independent test set. Results: The support vector machine with a radial basis function kernel (SVM-RBF) using LASSO-selected features achieved the best performance, with an AUC (area under the curve) of 0.903 (95% CI: 0.84–0.96), accuracy of 90.2%, sensitivity of 80.3%, and specificity of 100%. The average precision exceeded 0.92, and the calibration curve demonstrated good agreement (Brier score: 0.092). Interpretability analyses with SHAP (Shapley additive explanations) and LIME (local interpretable model-agnostic explanations) applied to the LightGBM (Light Gradient Boosting Machine) model confirmed EO%, RBC, and WBC as the most influential predictors. Conclusion:This parsimonious and interpretable model, relying solely on routine blood indicators, may enable timely and accurate diagnosis of acute appendicitis while providing additional insights, particularly in resource-limited settings.