Abstract: Reinforced concrete (RC) frame structures are important targets in urban seismic risk assessment with their natural vibration periods exhibiting significant inter-building variability, i.e. different periods for the same height. This is neglected in seismic design stage by using a deterministic building height-period formula, as a result, the seismic performance of existing buildings may be incorrectly estimated, and thus the natural vibration period needs to be reasonably quantified in regional seismic risk assessment. In this study, a database containing measured building heights and natural periods in the longitudinal and transverse directions of 641 RC frame buildings is constructed through literature searching, and a height-dependent random period function model is proposed based on stringent data filtering and a physics-guided data-driven methodology to consider the interbuilding variability. The comparison with the classical statistical regression random period model shows that the proposed model can better represent the difference in physical mechanisms of interbuilding natural periods and meanwhile well capture the joint probability characteristics of observed data, therefore could be applicable in regional seismic risk assessment.