Prediction of Empty Railcar Quantity Based on Two-Stage Machine Learning

In response to the continuous expansion of railway transport capacity and the enhancement of its intelligentization level, enabling the decision-making for optimal allocation of empty railcar resources to transition from passive reaction to proactive deployment, this paper proposes a two-stage machine learning-based framework for predicting railway empty car quantities. Currently, research on the accurate prediction of empty railcar quantities remains relatively underdeveloped, with no systematic methodological framework yet established. This study constructs a dual-stage predictive framework through the synergistic integration of a Random Forest classifier and a Gradient Boosting Regression Tree, encompassing both existence judgment and quantitative estimation. Experimental results demonstrate that the proposed model achieved an AUC of 0.97 in the classification task and an \({R}^{2}\) of 0.87 in the regression prediction, both indicating robust performance.