Spatiotemporally explicit earthquake prediction using deep neural network

Due to the complexity of predicting future earthquakes, machine learning algorithms have been used by several researchers increase Accuracy forecast. However, concentration previous studies has chiefly on temporal rather than spatial parameters. Additionally, less correlated variables were typically eliminated in feature analysis and did not enter model. This study introduces investigates effect parameters four ML algorithms' performance for magnitude earthquakes Iran as one most earthquake-prone countries world. We compared performances conventional methods Support Vector Machine (SVM), Decision Tree (DT), a Shallow Neural Network (SNN) with contemporary Deep (DNN) method biggest upcoming earthquake next week. Information Gain analysis, Accuracy, Sensitivity, Positive Predictive Value, Negative Specificity measures exploited investigate outcome using new parameter, called Fault Density, calculated Kernel Density Estimation Bivariate Moran's I, prediction, comparison other commonly discussed behavior models while dealing different combinations classes magnitudes. The results showed promising proposed parameter high magnitudes, especially SVM DNN models.