An earthquake casualty prediction model based on modified partial Gaussian curve

Earthquake casualty prediction is crucial for efficient and effective emergency management and response. In order to improve prediction reliability of earthquake casualties, correlation analysis and principal component analysis are used to select prediction covariates. Finally, five key indexes, including magnitude, epicenter intensity, population density, earthquake occurrence time and damaged building area, are chosen. According to the “two-stage” rule of earthquake casualties, a prediction model based on the modified partial Gaussian curve is proposed. In order to improve its prediction accuracy, the paper looked epicenter intensity and the casualty as the variables. And the partial Gaussian curve prediction model is modified by using the magnitude coefficient, population density coefficient, earthquake occurrence time coefficient and damaged building coefficient. The cross-validation experimental results show that the modified partial Gaussian curve has the advantages of good stability and high prediction accuracy comparing with the high-order nonlinearity, logarithmic curve, multivariate linearity, artificial neural network and so on. It can be used in practice from earthquake casualty prediction.