Time series estimation of earthquake occurrences in Bali and its surroundings using NARX network model

Abstract

Bali and its surrounding is earthquake hazardous as the geographical position is flanked by two major earthquake generators (subduction zone in the Southern and back arc thrust zone in the Northern) that have different tectonic structures. In this paper, a dynamic artificial neural network with NARX (Nonlinear Autoregressive Network With eXogenous Inputs) model will be used for time series estimation of earthquake occurrences in Bali and its surroundings area. NARX model has capability to represent time varying systems which is of interest in dynamical system estimation problem. Simulation results showed that NARX model gives better performances than ANFIS (Adaptive Network Fuzzy Inference Systems) and Layer Recurrent Network.