The Earthquake Magnitude Prediction Using Multilayer Perceptron Neural Network

Theoretical and empirical equations developed for calculating the magnitude of earthquakes are affected by a lot of parameters. Most of these parameters need to be measured and entered in the equations accurately, while, in many areas, due to the lack of required equipment, these parameters mostly are measured approximately and with low precision or even sometimes assumed. Moreover, these equations usually are exclusive of a specific region or state, so they are not reliable enough for other new regions.
On the other hand, neural networks have been proven to be one of the most practical effects in modelling and forecasting. There are three major advantages of neural networks. First, neural networks are able to learn any complex non-linear mapping. Second, they do not make a priori assumption about the distribution of data. Third, they are very flexible with respect to incomplete, missing and noise data (Vellido et al, 1999). Moreover, neural networks, regardless of the region and country, are a general solution in all areas.
The aim of this paper is to use a kind of neural network system named Multilayer Perceptron (MLP), which is one of the most influential neural network models, to predict the magnitude of the earthquakes.
This method consists of several layers of nodes. It includes an input layer, an output layer, and a hidden layer, each of which contain input node(s), output node(s), and hidden node(s), respectively.
The input nodes are based on some variables. In the current research, six independent variables including three spatial variables, one time variable and two variable related to physical characteristics are defined. The output nodes of neural networks are the prediction outputs or labels.
The seismic data that have been used in the research are got from the whole instrumentally recorded earthquakes occurred in Iran.
From whole data, 85% are used for network training and 10% for network testing and revising. The remained 5% is dedicated to derive the final prediction of the magnitudes of earthquakes. Then, these predictions have been compared with exact values to assess the network prediction ability.
In the hidden layer, as there is no method to decide the optimal number of hidden nodes directly, four different numbers of hidden nodes are chosen, including 8, 12, 16 and 20. Moreover, a well-known concern with neural networks is ‘‘overtraining”. To ease this problem, a set of four different learning epochs are used, including 1, 2, 4 and 8. Moreover, in training part, two different training methods, named Batch and Online, were applicable. In order to reach to more comprehensive results, both of these methods are applied. As a result, we setup 32 different groups of parameters and models.
After all, the results of the study indicate that MLP network has a good capability for predicting the magnitude of earthquakes. The average correct prediction of the models is about 70%.
To conclude, according to the results, the network is a functional device in predicting the magnitude of the earthquake of a region in an arbitrarily considered time.