Research and application of a hybrid system based on interpolation for forecasting direct economic losses of marine disasters. (July 2019)

• Record Type: Journal Article
• Authors: Zhao, Xin; Li, Hui; Ding, Lili; Liu, Mengxi
• Is Part Of: International journal of disaster risk reduction. Volume 37(2019)
• Publication Date: 2019-07
• Subjects: Interpolation -- Adaboost algorithm -- BPNN -- DELMD -- Virtual point; Emergency management -- Periodicals; Risk management -- Periodicals; Disaster relief -- Periodicals; Hazard mitigation -- Periodicals; 363.34
• DOI: 10.1016/j.ijdrr.2019.101121
• Languages: English
• ISSNs: 2212-4209

Abstract

Abstract: The accurate prediction of the direct economic losses of marine disasters (DELMD) is critical for allowing policy makers to take proper measures in managing a marine disaster. The prediction of the annual DELMD is hindered by its characteristically small samples, nonlinearity and indetermination. Developing ways to create an efficient system to forecast DELMD is challenging work. For this reason, this paper introduces a hybrid forecasting system that uses an adaptive boosting (Adaboost) algorithm and a back propagation neural network (BPNN) based on interpolation to predict DELMD. In this paper, four interpolations are employed to expand the original small sample with virtual points, then the Adaboost algorithm is used to optimize the results obtained by the weak predictor, BPNN, and produce the final forecasting results. Furthermore, to verify the prediction performance of the developed forecasting system, traditional models are used as comparisons to the new forecasting system. The experimental evidence shows that (a) the cubic spline interpolation is the most effective way to solve the small sample problem for forecasting DELMD, (b) the proposed hybrid forecasting system not only outperforms other traditional models but also is robust for other samples.