Author(s): Lixing Yang; Zuguo Zheng
Linked Author(s):
Keywords: No Keywords
Abstract: Ice-jam flood is a kind of complex system on large scale, which includes many factors related to hydrology, meteorology and geography etc. Many of them are hardly measured correctly, for they are complex, variable, nonlinear and nonnormal. The results of the present prediction methods are not good enough, based on the parametric model and the normal hypothesis. The new nonparametric method, Projection Pursuit Regression technique, doesn't give any hypothesis such as normal distnbution to the original observations and any pretreatment to the data, such as the exchange of data. Using computer, the method analyses multi-dimentional data through directly projecting them to low dimentional space, and uses numerical function to describe and predict the statistical law of data. In the case of the ice-jam over thirty six years, we form the model with the series of 1 / 3 or 1/2 data and test them with the series of the remaining 2/ 3 or 1/2 dataset respectively. All of them get the perfect results which the regular methods can't get.
Year: 1996