An Analysis of Track Geometry Data in Combination with Supporting Exogenous Sources Using Linear Regression Techniques

Abstract

We have investigated relationships between track geometry and weather, to determine if weather has any effect on degradation of track. The data provides an appropriate testbed, covering two Engineers’ Line Reference (ELR) IDs of known geological differences, allowing us to explore how weather affects different areas of the railway. We have justified the decision to exploit linear regression modelling, in order to provide a preliminary analysis as the basis for future work. As such, we process and develop a feature table from raw track geometry data that details the track geometry in 200m sections, named Location IDs. From this data, we have applied single and multivariate linear regression models to the dataset and provided an array of visualisations and supporting data. We confirm that linear regression was a suitable investigatory technique, supplying R2 values of up to 69.7%.