Research on application of earthquake prediction based on chaos theory

Real earthquake time series were analyzed and studied by chaos theory. Through the quantitative identification of time series, the results show that seismic time series performance deterministic chaotic characteristics. Considering the problems of slow convergence speed and low efficiency and local optimum caused by Gradient descent method, a method which momentum term with adaptive momentum factor was introduced into the Gradient descent method to modify the parameters of RBF neural network. Combined with the improved RBF neural network model, the seismic time series were predicted. Through the simulation of real data, the results show that the one-step prediction results were good, and when the multi-step prediction steps over a range of forecasts, forecast performance quickly declined.