Research On Model Of Fire Based On Box-Cox Transformation

In this paper, Box-Cox nonlinear transformation (abbreviated by BCT) onsome monthly fire and meteorological data are made from1997to2007bymaximal likelihood method.Applying the BCT data, we analyze the correlationsbetween the monthly times of fire and meteorological factors including monthlyaverage temperature, monthly average relative humidity, monthly average windspeed, monthly average rainfall, and build several linear regression models. At thesame time, we chose the best linear regression forecast model from forecast errorand mean square error of cross-validation. At last, we use the selected model topredict the times of fire in2008, and the predicted results show that the model setby BCT data can make the better prediction on the number of fire.The first chapter is preface, in which the theory and phylogeny of the city fireare introduced, the background, the purpose, the meanings of the study and theinnovation point are stated.In chapter2some fundamental information of City fire and methods of themodel are introduced, which will be used in the3thchapter. Scatter diagramanalysis, correlation analysis, linear regression analysis are common statisticalanalysis methods, Box-Cox transformation plays a key role in this paper.Chapter3is the empirical part, the prediction models of fire is built after thedata processing, we use the selected model to predict the times of fire in2008, andthe predicted results show that the model set by BCT data can make the betterprediction on the number of fire.In the last chapter, the conclusion and further work are talked out.