The adoption of Digital Twin (DT) technology has become a transformative force in the railway industry, offering innovative solutions to enhance operational efficiency, safety, and predictive maintenance. Predicting railway track stiffness is critical for maintaining structural integrity and safety of railway systems. Accurate predictions help identify potential issues early, reducing maintenance costs and minimizing disruptions. This paper demonstrates a DT framework for predicting track stiffness while the train is in motion. In this study, the subgrade and ballast stiffness varied with position, while rail and rail pad stiffness remained constant. The rail-track system was modeled using the multibody simulation software, Simpack®. An accelerometer was installed at the axle box to capture the train’s vertical vibration response. Various machine learning regression models were applied to train the collected dataset and predict track stiffness at specific locations and train speeds. Based on the evaluation metrics, the Random Forest regression model exhibited the highest predictive performance, with accuracies of 99.40% for variable ballast stiffness and 98.01% for variable subgrade stiffness. This high accuracy underscores the effectiveness of the random forest model in capturing complex relationships between position, frequency, and stiffness variations. By leveraging this model, the DT can provide precise real-time predictions of track stiffness, facilitating proactive maintenance and optimizing track performance.