Self-organized maps based neural networks for detection of possible earthquake precursory electric field patterns

Some physical parameters of the nature are known to have precursory correlation to the preceding earthquakes. Unfortunately, due to several uncertainties and structural variations of the geophysical faults a definite relationship that repetitively indicates occurrence time, magnitude and epicenter of a probable earthquake could not been constructed yet. Some recent investigations focus on distinction of earthquake precursory anomalous geo-electric field patterns. Highly sensitive monopolar electric field probes are used in measurement of them for more than four years, within a network of 15 stations located in Northwestern Anatolia. In this study, a neural network approach is proposed for detection of hazard precursory anomalous signal patterns. Correlation between the spatio-temporal electric field data measured by different stations and the regional seismicity is processed before being applied into the model for classification purposes. Main concern of this study is solving the classification problem of acquired signal patterns where time scale variation of the pattern window is high.