Hybrid Neural Network Methodology to Detect and Predict Seismic Activities

Abstract

The prediction of earthquakes, which can be devastating calamities, has proven to be a challenging research area. Because it involves filtering data to disturbed day changes, the contribution from multi-route effects and typical day-to-day fluctuations even on quiet days, the extraction of earthquake-induced features from this parameter requires intricate processing. Nevertheless, many researchers have successfully used several seismological concepts for computing the seismic features, employing the maximum Relevance and Minimum Redundancy (mRMR) criteria to extract the relevant features. The Artificial Neural Network (ANN) and the Adaptive Neuro-Fuzzy Inference System (ANFIS) are the primary soft computing tools that can be collaborated to detect and estimate earthquakes positively. The model in ANFIS is developed using subtractive clustering and grid partitioning procedures. The outcome shows that compared to ANFIS, ANN is more effective at predicting earthquake magnitude. Furthermore, it has been discovered that using this method to estimate earthquake magnitude is highly quick and cost-effective. Compared to earlier prediction studies, the acquired numerical findings show enhanced prediction performance for all the regions considered.