Research Of Time Series Addition Model Decomposition-forcasting

One of the main purpose on statistics analysis of time series is a forecast problem,so it is necessary and meaning for establishing a reasonable system method of decomposing,estimating and forecasting time series.Generally,a time series can be decomposed an addition or a multiplication model of trend item,season item,cycle item and random item or among them several some items through function transfomation or directly.But multiplication modle can be conversed an addition modle by logarithms transformatin,therefore,this paper reseaches on the estimate and forecast problem of trend item,season item,cycle item and random item of time series addition model.In the decomposing-forecasting of trend item,with the purpose of improving the accuracy of simulation and forecasting of trend item,this paper brings forward a new thinking of removing season item again to simulate trend item,and gives the mothod of removing season item; also,in this part researchs the estimate and forcast problem of trend item by difference of higher order method in the disturbance of cycle item and random item,and exemplifies that the difference of higher order method is better than minimum two mutiplication at individual of estimate.In the forcasting of season item, this paper improves trend extrpolate decomposte-forcaste method which takes correspondig simulation and estimate according to the characteristics of season item data,and exemplifies the method of improvement can enhance the accuracy of approching and estimating indeedly;in the meantime,in the seasonal index number smooth method,gets the estimate algorithm which declines the estimate error of season item to minimum of two season parameter by non-linear program of operation research. In the estimate and forcasting of cycle item,this paper explains the theory basis and calculate steps of numerical approximation seasonal extrpolate method put forward by literatures [5],[6],[7] further,and exemplifies that numerical approximation seasonal extrpolate has better estimate effect.The forecast of random item be the forcast of steady process namely,this paper sums up "new info" recursion arithmetic got by hilbert space theory.