Abstract:  Applying the microtremer H/V method in urban areas faces many complex artificial noises. The microtremer data with a large number of noise needs to be denoised and extracted. In this paper, a   multiple weighted HVSR method based on XGBoost (extreme gradient boosting) is proposed to solve the problems that the existing methods are difficult to deal with the microtremer data with considerable interference and the signal extraction process is cumbersome. Firstly, the amplitude and frequency of the microtremer data are analyzed.  Amplitude-weighted spectral ratio, frequency-weighted spectral ratio, and multiple weighted spectral ratio are established. Then, the denoised HVSR curve is calculated by the XGBoost method. The proposed method is compared with the STA/LTA (short-time average/long-time average) method  to analyze the actual high-noise data, the results show that the proposed  method is better than the STA/LTA method for high-noise data extraction. 

Key words:  , microtremer H/V method, weighted spectral ratio, XGBoost, surface wave exploration