Research On Earthquake Prediction Based On Genetic Neural Networks

Presently,earthquake predicton is a very important task in the field of time series study.Because there are many factors which arose earthquake,it is very difficult to establish physics model.Neural Network technology has unique advantage in constructing and predicting unknown object theoretical model,so it has been widely applicated in prediction. This paper discusses earthquake time series analysis method and theory,studies genetic neural network's application in earthquake forecasting.Its facility and validity are proved by experiments by using MATLAB.In this paper,the main content includes earthquake time series,analyzing some forecast models,genetic neural network's advantage in earthquake forecasting,the development of the neural network technology and the neural network's aplication in non-linear forecasting;also includes BP network's structure,model,which is commonly used in neural network,the improving of algorithm,the steps to achieve and the basis of genetic algorithm,and a brief description of MATLAB toolbox.In this paper,the integration of genetic algorithm and BP neural network are analyzed. Considering unit's bias and initial weight value are the most important factors in forecasting's precision of genetic neural network,in order to escape local minimum and accelerate the convergence speed of BP networks,genetic algorithm is used to choose a more reasonable frame of neural network.Lastly,earthquake time series are predicted,in which,the genetic neural network model is utilized.The results are ideal,and illustrates the earthquake time series forecasting model based on genetic neural networks have good capability of popularization and forecasting,validity and suitable quality.