The most common causes of degradation of the track are uneven settling of the track due to inhomogeneity of the base, inadequate maintenance, poor state of the upper and lower structure, etc. In predicting the degradation of the track we need to know when the limits of the geometric parameters of the track are exceeded in order to decide in due time for effective planning of maintenance and renovation work and for taking other measures to ensure railway safety. The aim of this master’s thesis is to investigate the most important models of track degradation due to geometry defects, select a model suitable for use in Slovenia, and test this model in practice along a selected section of railway line. Based on the available data on traffic loads by individual rail sections and data on the measurement of the track with a measuring train, the Austrian and Swedish models of degradation of the track were found to be the most suitable for use in the Slovenian context. The Austrian degradation model is particularly well suited for predicting degradation of the track due to various geometric errors, depending on the initial state of the track quality and traffic load. The Swedish degradation model addresses the track geometry degradation as a function of the standard deviation of the longitudinal level defects and standard deviation of the right and left alignment defects. With the introduction of technical specifications on interoperability, determination of the standard deviation of defects has become a mandatory criterion for assessing the geometric quality of the track in Slovenia as well. The applicability of both models was tested in practice along a relatively new section of track between Gornje Ležeče and Divača and along a section of track between Ljubljana and Brezovica that was renovated 36 years ago. Degradation of the railway track was monitored after the last regulation by tamping machine in 2012 until 2015. The research showed a very good correlation between data actually measured and the results calculated using the proposed models.