Interpretable seismic oscillation parameter probability density distribution prediction method based on NGBoost and SHAP values

Abstract

The invention discloses an interpretable seismic oscillation parameter probability density distribution prediction method based on NGBoost and SHAP values, and relates to the technical field of seismic engineering, and the method comprises the following steps: 1, determining a research region, collecting seismic event waveforms and metadata information in the research region, carrying out the analysis and processing, and building a strong vibration database; 2, performing data cleaning on the strong vibration data, and selecting data for machine learning model training; 3, training a machine learning model for predicting the probability density distribution of the seismic oscillation parameters by using the selected strong vibration records and a natural gradient lifting algorithm; 4, calculating SHAP values of all features of all samples, analyzing importance of each feature according to the SHAP values, analyzing how to influence seismic oscillation parameter prediction, and explaining the machine learning model; and 5, predicting the seismic oscillation parameter probability density distribution of a newly occurring or hypothetical earthquake by using the trained machine learning model.

Description

technical field [0001] The invention relates to the technical field of earthquake engineering, in particular to an interpretable earthquake motion parameter probability density distribution prediction method based on NGBoost and SHAP values. Background technique [0002] Casualties and property losses caused by earthquakes are mainly caused by the damage and collapse of building structures caused by strong ground movements. Strong ground motion is also the direct cause of secondary disasters such as landslides. After an earthquake, the rapid estimation of strong ground motion parameters (Peak Acceleration (PGA), Peak Velocity (PGV), Peak Displacement (PGD) and Acceleration Response Spectrum (SA)) can be used to determine the loss caused by the earthquake and guide emergency rescue work . Before the earthquake, the prediction of the earthquake caused by the hypothetical earthquake that may occur on the dangerous fault can be used for the probability analysis of the earthqua...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art, to provide an interpretable method for predicting the probability density distribution of ground motion parameters based on NGBoost and SHAP values, and to solve the problem of uncertainty in the prediction results of ground motion parameters, it is proposed to use natural gradient boosting (NGBoost )algorithm

Images