Objective To establish a prediction model based on machine learning (ML) algorithm for central lymph node metastasis (CLNM) in papillary thyroid carcinoma (PTC). Methods Clinical data of 262 PTC patients at Department of Thyroid and Breast Surgery, The Fifth Affiliated Hospital of Xinjiang Medical University from November 2022 to November 2025 were retrospectively collected and randomly divided into training (n=184) and testing (n=78) sets at a 7∶3 ratio. The LASSO (least absolute shrinkage and selection operator) regression combined with the Boruta algorithm was used to screen for predictors of CLNM. Eight machine learning algorithms, namely SVM (support vector machine), NN (neural network), RF (random forest), XGBoost (extreme gradient boosting), KNN (K-nearest neighbors), AdaBoost (adaptive boosting), LightGBM (light gradient boosting machine), and CatBoost (categorical boosting) were incorporated to construct predictive models. The performance of the established models was compared and evaluated using the area under the receiver operating characteristic curve, sensitivity, specificity, accuracy, precision, F1-score, calibration curve, and decision curve analysis. Results By combination of LASSO regression and the Boruta algorithm, five predictors (age, maximum lesion diameter, mean platelet volume, osmotic pressure, anti-thyroglobulin antibody) were selected. In the testing set, CatBoost achieved the highest area under curve (0.776), followed by RF (0.757). Considering discrimination, calibration, and clinical net benefit, RF was identified as the optimal model due to its best calibration performance (best agreement between predicted probability and actual risk). Furthermore, RF had the widest clinical threshold range (0.1–0.8) and the highest clinical net benefit in decision curve analysis. The sensitivity, specificity, accuracy, precision, and F1 score of the RF model were 0.619, 0.860, 0.795, 0.619, 0.619, respectively. Conclusion The RF based prediction model has high clinical practical value for early prediction of CLNM in PTC patients and guiding clinical decision-making.
