PERFORMANCE INVESTIGATION OF NEURO-FUZZY SYSTEM FOR EARTHQUAKE PREDICTION

In this paper, the ability of Adaptive Neuro-Fuzzy Inference System (ANFIS) for the seismic moment prediction of the next earthquake has been investigated. To do so, two seismic indicators were used as inputs which were computed from the data related to the earthquakes recorded during last 63 years in a region which is located in southern Iran with 1° N latitude and 4° E longitude. The first indicator is the logarithm of the mean annual rate of exceedance for each record which is determined based on definition of Gutenberg-Richter law and the second indicator is time duration from the specified time origin. The output indicator is the logarithm of cumulative amount of seismic moment between the origin event and the future earthquake of each event. The test results with correlation factor 98% showed that ANFIS can be a useful tool for earthquake prediction.