Strong-motion Earthquake Prediction Model using Convolutional Extreme Learning Machine
Phiphat Chomchit, Paskorn Champrasert · 2022
The earthquake occurrence is difficult to predict. However, the ground motion can be measured by using the network of seismic telemetry stations. In this case, the sensing seismic wave data can be used as the input to predict the strong-motion earthquake in the target area. In this paper, a strong-motion earthquake prediction model, called C-ELM, has been developed by applying the integration of the extreme learning machine and convolutional neural network. The prediction model applies the convolutional neural network method to extract features from the initial P-wave. The extreme learning machine method is applied to minimize the loss function. The proposed prediction model has been evaluated by comparing its predicted results to the traditional convolutional neural network. The result shows that the strong-motion prediction accuracy of the C-ELM is close to the result of the traditional convolutional neural network. However, the C-ELM can extensively reduce the computation training time.