Forecasting flood disasters using an accelerated genetic algorithm: Examples of two case studies for China

This article discusses a rescaled range analysis model, titled AGA-R/S, that is based on an accelerated genetic algorithm. The parameter a, Hurst index of rescaled range analysis, and the recurrent time of disaster in the next time-period, were directly computed using an accelerated genetic algorithm developed by the authors. As case studies, using the AGA-R/S model, a forecast was made of the tendency for change in a time series of annual precipitation for the city of Jinhua, China. The model also forecast flooding-disaster in the city of Wuzhou, China. Results indicate that it is a relatively efficient technique to forecast the change-tendency of flood and disaster time series using the AGA-R/S model. When time series is utilized, forecasted error of the AGA-R/S model is less than with a linear least square method. The Hurst indexes of the two cities are from 0.23 to 0.24, which indicates that these time series are fractal and relatively long-term. Their fractional Brownian motion shows anti-persistence. AGA-R/S has application in forecasting the change-tendency of other natural disaster for specific time series. Copyright Springer Science+Business Media B.V. 2008