摘要
This paper presents a new method to estimate the time of important earthquakes in Hormozgan region with magnitude greater than 5.5 based on the Radial Basis Function （RBF） Neural Network （NN） models. Input vector to the network is composed of different seisrnicity rates between main events that are calculated in convenient and reliable way to create optimized training methods. It helps network with a limited number of training data to estimation. It is common for earthquakes modeling by data-driven methods in this case. In addition, the proposed method is combined with Rosenberg cluster method to remove aftershocks events from the history of catalog for NN to better process the data. The results show that created RBF model successfully estimates the interevent times between large and sequence earthquakes that can be used as a tool to predict earthquake, so that comparison with other NN structure, for example Multi- Layer Perceptron （MLP） NN, reveals the superiority of the proposed method. Because of superiority proposed method has higher accuracy, lower costs and simpler network structure.
This paper presents a new method to estimate the time of important earthquakes in Hormozgan region with magnitude greater than 5.5 based on the Radial Basis Function （RBF） Neural Network （NN） models. Input vector to the network is composed of different seisrnicity rates between main events that are calculated in convenient and reliable way to create optimized training methods. It helps network with a limited number of training data to estimation. It is common for earthquakes modeling by data-driven methods in this case. In addition, the proposed method is combined with Rosenberg cluster method to remove aftershocks events from the history of catalog for NN to better process the data. The results show that created RBF model successfully estimates the interevent times between large and sequence earthquakes that can be used as a tool to predict earthquake, so that comparison with other NN structure, for example Multi- Layer Perceptron （MLP） NN, reveals the superiority of the proposed method. Because of superiority proposed method has higher accuracy, lower costs and simpler network structure.
