Accident Prediction Accuracy Assessment for Highway-Rail Grade Crossings Using Random Forest Algorithm Compared with Decision Tree

•Random Forest Application in Highway Rail Grade Crossing Safety Analysis is introduced•Complete variable importance analysis is demonstrated•AADT, rail day through traffic, night through traffic, train speed, number of road traffic lanes are the top five important attributes in predicting highway rail grade crossing safety analysis•Comprehensive model prediction assessment is performed with 8 measurements: Sensitivity, specificity, accuracy, precision, false positive rate (FPR), true negative rating (TNR), F1 score, and Matthews correlation coefficient (MCC).•The RF model is able to improve both crash and non-crash forecasting performance and effectively reduce false alarm rates for unbalanced data.