Objective: To develop and validate a machine learning-based model for predicting pin tract infection in children after external fixation surgery . Methods: A retrospective study was conducted on 338 children (213 males, 125 females) who underwent external fixation between 2021 and 2024. Patients were randomly split into development (n=237) and test (n=101) sets (7:3). Demographic, surgical, and laboratory data were collected. Lasso regression identified significant predictors, which were used to build four models: Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), and Naïve Bayes (NB). Model performance was evaluated using ROC curve and decision curve analysis (DCA); variable importance was ranked for the optimal model. Results: In the test set, the RF model achieved the highest AUC (0.99), sensitivity (1.00), accuracy (0.98), and F1-score (0.93), outperforming LR (0.69, 0.63, 0.71, 0.38), SVM (0.72, 0.52, 0.85, 0.49), and NB (0.68, 0.70, 0.66, 0.36). The RF model demonstrated superior discriminative ability . Conclusion: The Random Forest model exhibited excellent predictive performance for pin tract infection in children postexternal fixation, providing an effective tool for early prediction and prevention in clinical practice.