This study is a single-center retrospective analysis of 930 consecutive congenital choledochal malformation surgeries screened from January 2019 to January 2024; 803 patients met criteria for analysis. Pancreaticobiliary maljunction (PBM) was present in 628 of 803 children, 78.2 percent. PBM was defined by intraoperative cholangiography as a common channel longer than 5 mm, assessed by two blinded surgeons with consensus arbitration. Patients were split 80 percent for training and 20 percent for an independent test set. Variables with more than 20 percent missingness were removed, remaining gaps were imputed with MICE. Categorical variables were one-hot encoded. Feature selection combined correlation filtering at |rho| greater than 0.7, recursive feature elimination using a Random Forest base estimator, and clinician review, yielding 43 features from seven clinical domains. Models trained with five-fold cross-validation included logistic regression, SVM, Random Forest, XGBoost, AdaBoost, LightGBM, and KNN. SMOTE was applied only within training folds. Performance was summarized on the held-out test set with nonparametric bootstrap confidence intervals. Random Forest was the best single model on the test set with AUC 0.86 (95% CI 0.79 to 0.92) and recall 0.92 (0.88 to 0.96). A soft voting ensemble that combined all seven models achieved AUC 0.87 (0.81 to 0.92), recall 0.91 (0.85 to 0.95), F1 0.91 (0.87 to 0.94), accuracy 0.85 (0.80 to 0.90), specificity 0.64 (0.46 to 0.80), PPV 0.91 (0.85 to 0.95), and NPV 0.63 (0.46 to 0.80). For comparison, logistic regression reached AUC 0.79 (0.70 to 0.86) and recall 0.78 (0.71 to 0.85). SHAP analysis highlighted 12 contributors: higher abdominal pain, surgical age, prealbumin, and C-reactive protein were associated with increased PBM probability, while higher gamma-glutamyl transferase, alkaline phosphatase, total cholesterol, globulin, D-dimer, urea, albumin, and prothrombin time were associated with decreased PBM probability. Median age was 989.7 days in PBM vs 202.7 days in non-PBM, and abdominal pain occurred in 63.4 percent vs 25.9 percent, p less than 0.01.