The Statistical Modeling And Predicting About The Increased-value In Industry Of China

Abstract: The paper collected the history data of the increased-value in industry of China by month from January.2001 to December. 2005, based on which it sets up an appropriate model by using the analytical theory and methods. First, the series is resolved into three parts according to its scatter plot: the trend item, the seasonal item and the random item. We find that the data series' plot is similar with that of the quadratic function and that of the exponential function by the method of seasonal adjust provided by Eviews software, hence the article imitates these two functions respectively, and by comparison it gets the result that the ' quadratic function is more better, meantime, it also gets the predicted value of the trend item. Second, the article gets the seasonal factor of each month by carrying on the seasonal adjustment to the data series, which eliminates the trend item. Through observation of plot of the random item, we find that the random item is not steady. Therefore one-degree difference method is adopted to remove the instability. Through the analysis of autocorrelation and partial correlation plot, the random item is imitatedby three models: ARMA(1,1)model, ARMA(2,1)model and ARMA(2,0)model.By statistical test, the ARMA(1,1)model is the best one. Last, thearticle predicts the random item' s predicted value of January and February. 2006, and then by introducing the predicted value of the trend item and the seasonal factor, it gets the predicted value of increased-value in industry. By contrasting the predicted value with the real value observed, the predict error is under 5%. So the conclusion got from this paper is that the model whose trend item imitated by the quadratic curves and whoserandom item imitated by the ARIMA(1,1,1) model is the best model we attempt to find.