Both the propagation velocity and the direction of atmospheric waves are important factors for analyzing and forecasting meteo-tsunami. In this study, a total of 14 events of meteo-tsunami over 11 years (2006-2016) are selected through analyzing sea-level data observed from tidal stations along the west coast of the Korean peninsula. The propagation velocity and direction are calculated by tracing the atmospheric disturbance of each meteo-tsunami event predicted by the WRF model. Then, the Froude number is calculated using the propagation velocity of atmospheric waves and oceanic long waves from bathymetry data. To derive the critical condition for the occurrence of meteo-tsunami, supervised learning using a logistic regression algorithm is conducted. It is concluded that the threshold distance of meteo-tsunami occurrence, from a propagation direction, can be calculated by the amplitude of air-pressure tendency and the resonance factor, which are found using the Froude number. According to the critical condition, the distance increases logarithmically with the ratio of the amplitude of air-pressure tendency and the square of the resonance factor, and meteo-tsunami do not occur when the ratio is less than 5.11 hPa/10 min.