Abstract: Track irregularity status was a key factor affecting driving safety. Track Quality Index(TQI) could reflect the change of track geometry state. It was a time series changed with time and was with random characteristic. In order to study the change trend of track status better, the gray interval prediction model was developed to model over TQI which was changed with the range of the unit section of track, and compared with conventional non- equidistant GM(1,1) model. Effectiveness of the prediction model was validated by the actual data of Jingjiu Line K467.8~K468 unit section, The results showed that the gray zone model was with higher prediction accuracy.