A traffic congestion prediction method including the steps of: detecting an acceleration of a vehicle; calculating a power spectrum corresponding to a frequency from a frequency analysis of the detected acceleration; calculating a simple linear regression line of the power spectrum and calculating a maximum value of an amount of change in a gradient of the simple linear regression line in a predetermined frequency range as a maximum gradient value; detecting an inter-vehicle distance between the vehicle and a vehicle ahead; estimating an inter-vehicle distance distribution from the detected inter-vehicle distance by using a distribution estimation method; calculating a minimum value of covariance value from the estimated inter-vehicle distance distribution; estimating a distribution of a group of vehicles ahead from a correlation between the minimum value of covariance value and the maximum gradient value; and performing a real-time traffic congestion prediction based on the distribution of the group of vehicles.