Improvement Research Of Time Series Exponential Smoothing

In recent years,the establishment of non-linear model of economic forecasting in time series analysis,widely attention and research.Among them,exponential smoothing as an important branch of time series forecasting analysis,has become a classic model for the prediction and control,this is mainly because of its own with excellent performance,easy to operate,widely used features.However,the traditional exponential smoothing model there are still some problems is not well resolved:The first is smoothing parameter values of the majori- ty of people still rely on everyday experience to acquire,difficult to achieve the optimum value;Followed by the smoothing parameter is difficult static and change over time;Last is smoothing once again difficult to determine the initial value problem.This article is from these problems,and to make dynamic optimization of the three exponential smoothing model,professor Li Suoping using the parameters proposed by improved methods,by the exponential smoothing parameterαreplaced by a static dynamic exponential smoothing parameter,and use C language programming,through four of one-dimensional search algorithm for parameter optimization to solve these problems,and determine the smoothing parameterαwhat makes SSE is the smallest,to solve the problem which difficult to select the smoothing parameter.In this article last,use two examples to be verified,to make forecasts is worth to the optimization,further proof of the advantages of the new model.