In railways, it is often not possible to avoid real-time disturbances altogether. A disturbed train may easily spread its delay to other trains, resulting in reduced train punctuality and substantial socio-economic costs. During disturbances, it is important to effectively prioritize trains, resolve the potential conflicts in the train timetable and quickly arrive at a rescheduled timetable as per the infrastructure manager’s (IM’s) goals. Currently, this task is handled manually by traffic controllers. The train timetable rescheduling problem is typically hard to solve, even for modern computers. Solving it requires exploring many alternative rescheduling decisions and quickly finding a good- enough solution based on the IM’s goals in real-time. Two main solution approaches in computer science to solve real-world problems are traditional and machine learning (ML) approaches. In a traditional approach, a computer is programmed to do various computations on input data to solve the problem and generate the solution as output. In an ML approach, a computer uses input data and expected outputs to learn how to solve the problem at hand without being explicitly programmed. The learning process results in an ML model/program, which helps to make data-driven decisions in the future using its learned experience. ML approaches have received remarkable attention in recent years. They have been shown to work well in transportation and many other problem domains, independently or in combination with traditional approaches, to tackle complex scheduling problems and get better solutions. This project aims to improve decision support for effective conflict resolution during disturbances by devising an ML model. The developed ML model will be able to assist traffic controllers in resolving a conflict during disturbances by suggesting the train to be prioritized (along with a confidence score). The model will be used in combination with a rescheduling algorithm developed at BTH (2016–2021) to speed it up and make data-driven decisions. We will assess the devised model, using various disturbances on a Swedish railway network, by evaluating the quality of the rescheduled timetables obtained by following the model’s suggestions.