Pattern Recognition on Seismic Data for EarthquakePrediction Purpose

Earthquakes has been known as a destructive natural disaster. Due to high human casualties and economical losses, earthquake prediction appears critical. The b-value of Gutenberg Richter law has been considered as precursor to earthquake prediction. Temporal variation of b-value before earthquakes equal or greater than Mw = 6.0 has been examined in the south of Iran, the Qeshm island and around of this from 1995 to 2012. Clustering method by the k-means algorithm has been performed to find pattern of variation of b-value. Three clusters are obtained as optimum number of clusters by the Silhouette Index. Before all mentioned earthquakes greater than Mw = 6.0, cluster 1, which is known as a decrease in b-value has been seen. so decreasing b-value before main shocks as distinctive pattern has been considered. Also an approximate time of decrease has been determined. Keywords— earthquake prediction, long-term seismic hazard analysis, pattern recognition, clustering, seismicity rate, b-value.