Predicting Time Series Data at Floodplain Area using Chaos Approach

Floods are natural disaster that can cause substantial losses of lives and property. Flood risk can be minimized if an early warning can be issued. In this regard, daily river flow prediction was analyzed at a river flow station in Ladang Victoria, Malaysia which is located in a floodplain area. Prediction using chaotic approach that involves the reconstruction of phase space and prediction have been employed in this research. The reconstruction of phase space involves a single variable of river flow data to m-dimensional phase space in which the dimension (m) is based on the optimal values of method of Cao and the variation of m for inverse approach. The results from the reconstruction of phase space have been used in the prediction process using local linear approximation method. From our investigation, river flow at Muda River is chaotic based on the analysis from Cao method. Overall, prediction results for both methods can provide a good prediction based on a high correlation coefficient. However, the combination of the preliminary parameters for the inverse approach yields better prediction. Therefore, the inverse approach can be proposed for predicting daily river flow data for the purpose of providing important information about the flow of the river system in floodplain area especially in Sungai Muda.