FORECASTING MODEL OF ANNUAL AND SEASONAL NUMBERS OF AFFECTING TROPICAL CYCLONE BASED ON PROJECTION PURSUIT REGRESSION

The projection pursuit algorithm is a kind of up-to-date statistic algorithms which deals with the height dimensional problem, especially the negate normality problem. The forecasting factors are screened out by stepwise regression through general overhauling correlation coefficient between the 500 hPa、100 hPa geopotential heights of the Northern Hemisphere、 the sea surface temperature(SST) of the North Pacific Ocean and the circumfluent features of 500hPa geopotential heights and annual, seasonal numbers of tropical cyclone affecting Fujian. Then, a forecasting model for annual and seasonal numbers of tropical cyclone affecting Fujian is proposed using the original idea and its implement algorithm of projection pursuit regression (PPR). The results show that the forecasting precision of the PPR model is much better than that of stepwise regression (SR) model.