Error Analysis Of Sea Level Prediction And Its Analysis System

With the increasing global warming,sea level rise has gradually attracted people’s attention.A large number of studies focus on the anaylsis of sea level trends,in contrast,the researches about local sea level fluctuations are in small quantities.In this paper,local sea level fluctuations are predicted,prediction results are analyzed and sea level prediction system is developed.The highlights of this paper are as follows.Firstly,linear and nonlinear models namely Autoregressive(AR)model,Support Vector Machine(SVM)and Back Propagation Neural Network(BPNN)are used to predict sea level time series.The results show that prediction errors generated by SVM and BPNN are much lower than that generated by AR model.Besides,prediction errors are still lower than that of linear model when the prediction step is increased.Therefore,the results show that forecasting cost of sea level time series will be saved when SVM and BPNN are used.Secondly,this paper gets the quantitative relationship between prediction error and prediction step in the research of sea level time series.The results exhibit that prediction error of sea level in terms of prediction step follows power function,which provides a guideline for choosing the length of prediction step and controlling prediction error.Last but not the least,a system which is used to predict sea level fluctuations is developed on the MATLAB GUI platform.It has three main functions:predicting sea level with any given samples using three kinds of prediction models,generating the quantitative relationship between prediction error and prediction step and generating the curve between prediction error and sample size.Users can understand the methods to predict sea level and the prediction results in depth via this system.