Nonlinear Autoregressive Network with the Use of a Moving Average Method for Forecasting Typhoon Tracks

摘要

Forecasting of a typhoon moving path may help to evaluate the potential negative impacts in theneighbourhood areas along the moving path. This study proposed a work of using both static and dynamicneural network models to link a time series of typhoon track parameters including longitude and latitude ofthe typhoon central location, cyclonic radius, central wind speed, and typhoon moving speed. Based on thehistorical records of 100 typhoons, the performances of neural network models are evaluated from theindices of a correlation coefficient and a mean square error. The dynamic model or the so-called nonlinearautoregressive network with the use of a moving average method proved to forecast the ten types oftyphoon moving path more effectively in Taiwan region. The new and simply approach developed in thisstudy for solving studied typhoon cases may be applicable to other areas of interest worldwide..