Conference article

In order to enable widespread use of data driven analysis and machine learning methods for rail operations problems, large volumes of operational data are needed. This data has the potential to contain erroneous or missing values, especially given its size and dimensionality. In this work a data reconciliation problem for rail dispatch data is proposed to identify and correct errors, as well as to impute missing data. The data reconciliation problem finds the least-perturbed modification of the historical data that satisfies operational constraints, such as feasibility of meet and overtake events, safety headway, siding allocation, and running time. It also imputes missing values with estimates that satisfy all operational constraints. The data reconciliation method is applied to a large historical dataset from freight rail territory in Tennessee, United States, containing over 3,000 train records over six months. The method identifies and corrects errors in the historical data, and is able to impute data on a synthetically decimated version of the historical data. The quality of the imputed data from data reconciliation is compared to imputed data using naive interpolation. The results show that data reconciliation reduces timing error of imputed points by up to 15% and increases the number of meet and overtake events estimated at the correct historical location from less than 40% to approximately 95%. These findings indicate that the data reconciliation method is a useful preprocessing step for analysis and modeling of railroad operations that are based on real-world physical dispatching data.

data reconciliation, dispatching, modeling, optimization